DAVID BAUCKMANN

PROGRAMMATIC

SUPPLY
CHAIN
IN CONTEXT
A practical textbook
 1

PROGRAMMATIC

IN CONTEXT

David Bauckmann
A practical textbook
2

© David Bauckmann, 2023

Cover and design © Programmatic Media s.r.o., 2023
www.Bauckmann.info
ISBN 978-80-11-02821-3
 3

Acknowledgements
This book could not have been written I have also had great help from people
without the support of my colleagues and in the programming field who have read
friends in my professional community. this book in various stages of completion
and provided invaluable comments and
First and foremost, I must thank my
feedback. Thanks to Lukáš Šmol (cpex.
colleagues in the publishing and monetiza-
cz), Matěj Kolarovský (GroupM, IAM
tion team at Impression Media who have
Slovakia), Eugenu Cirák, Jorg Vogelsang
worked tirelessly with me for several years
(BVDW e.V.), Jan Sunavec (FILMZIE), Petr
to document new trends and practices
Mareš (Invibes.com) a Isabel Montalvão
in the field. It can be very challenging at
(Pubstack.io).
times, and I appreciate that they have the
energy to constantly innovate and seek And a big thanks to Martin Picek (GroupM)
out new practices. Thanks to Veronika for his support and great marketing advice,
Buriánková, Petronella Ördögh, Marcela without which this book would be much
Hofmanová, Mark Jeníček and Martina more boring.
Ševčíková. You are a great team with great
And last but certainly not least, thanks to
know-how that complements one another
my family for their support.
perfectly.

Furthermore, I would like to thank Roman

Strolejda, CEO at Impression Media, who
constantly supports us and gives the whole
team space to find new solutions. Thanks
also to other colleagues who have read this
book and provided feedback, especially
Martin Holeček and Marcela Pecková.

Special thanks to Beatrice Trojan (beartcz.

art) for the comprehensive illustrations and
graphic design. I know it is not easy for
creatives to do technical illustrations and
endure several rounds of comments and
corrections.
4

Contents

Acknowledgements 3
Foreword 9
Who this book is for 9
How to work with the book 9
1. OpenRTB protocol 10
Relationship between RTB and OpenRTB 10
History of OpenRTB 10
The principle of OpenRTB 10
Bid request 13
Sample bid request 15
Bid response 16
Sample bid response 18
RTB 3.0 19
Chapter conclusion 21
2. Basic principle of supply chain 22
Basic building blocks of supply chain 23
Resales within ad exchanges 25
Impact of header bidding on supply chain 27
Fees in supply chain 28
Average number of participants in supply chain 31
Average number of SSPs used by publishers 33
Average number of DSPs used by advertisers 37
Loss of information in supply chain 40
Cookie Matching (Cookie Syncing) 41
Other factors affecting supply chain 42
Chapter conclusion 42
3. Supply chain disruption: Ad fraud 44
What is ad fraud? 44
Magnitude of ad fraud 45
Types of ad fraud 46
 5

Malicious bots 47
Click farms 48
Cookie stuffing 48
Domain spoofing 49
Malvertising 50
Ad injection 50
AdWare 51
How to recognize ad fraud 51
How to fight ad fraud 52
Chapter conclusion 53
4. Protection – Transparency projects 54
Ads.txt 54
Seller.json and SupplyChain object 58
Ads.cert 60
Buyers.json and DemandChain Object 60
SupplyChain and DemandChain Objects 61
SupplyChain validation 63
How to check if the settings are correct 64
Support for individual projects 66
Chapter conclusion 67
5. SPO and DPO 68
Supply path optimization (SPO) 68
Automated purchase path optimization 69
What purchasing platforms can automate 70
Manual optimization of the purchase path 70
What SPO optimizes 71
Setting up SPO in four steps 71
Demand path optimization (DPO) 74
Automated sales path optimization 74
Manual optimization of the sales path 75
6

What DPO allows to optimize 76

Seven useful steps for an effective DPO 76
Bid throttling 77
When bid throttling occurs 78
How to limit bid throttling 78
How to spot bid throttling 79
Chapter conclusion 79
6. Sustainable Programmatic Supply Chain 80
General framework 80
Global warming impacts 82
CO2 versus other greenhouse gasses 82
Net zero 83
Calculation of CO2 production 84
CO2 production 85
Carbon credits 86
Carbon credit life cycle 87
Carbon credit markets 87
How the advertising industry produces emissions 89
Electricity 89
Energy consumption by the user 91
Online advertising is comparable to the airline industry 92
Video is many times worse than banners 94
Attention economy 94
Frameworks 95
ADGREEN 96
Sustainability and supply chain 96
Chapter conclusion 97
Conclusion 98
Terms 100
Sources 104
7
8
 9

Foreword
The programmatic approach to online such as DSP, SSP, header bidding, and
ad buying has significantly benefited our ad exchange and assume that the reader
industry, but as time has shown, it has also understands them. We comment on these
brought some significant challenges. Grad- terms only briefly.
ual layering and packaging has resulted in
abuse of trust and ad fraud, and one of the How to work with
biggest hurdles today is transparency in
supply chains. the book
In this book, we will attempt to explain all This book is published in print and PDF.
the essential components of programmatic If you are working with the printed version,
supply chains. We start with the basic build- then links to all the resources can be found
ing blocks of OpenRTB protocol on which a at the end of the book in the form of a QR
supply chain is based (see page 10). code.
The main content of this book covers the If you are working with the PDF version,
principles of supply chains (see chapter then all links are active hyperlinks.
22). We then go through how it is abused
(see page 44) and what tools we use to try If you have any recommendations or
to prevent this (see page 54). comments, please feel free to contact the
author. The world of online advertising is
In addition to theoretical information, we undergoing a very rapid evolution, and this
outline some ways for advertisers and pub- book too will gradually become obsolete. If
lishers to use supply chain knowledge to it’s within our capabilities, we’d be happy to
optimize their campaigns (see page 68). release an update.
Finally, we discuss the impact of supply
chains on greenhouse gas emissions (see
page 80).

Who this book is for

This book is primarily intended for profes-
sionals in the field of programmatic buying,
both on the side of the advertiser and the
publisher.
In general, however, the information pre-
sented here will also be relevant to online
advertising executives for whom an under-
standing of the whole principle of program-
matic buying is essential.
This book is not intended for beginners and
assumes knowledge of basic concepts and
procedures. Many times we refer to terms
10

1.
OpenRTB protocol
Before we get into the supply chain itself, into the details, it is recommended to read
it is necessary to understand the principle the entire documentation of the current
of OpenRTB protocol on which an entire version of OpenRTB 2.61.
supply chain is based.
At the time of this book’s publication,
OpenRTB protocol defines the basic build- OpenRTB protocol version 2.6 is widely
ing blocks of the supply chain: deployed, but there already exists a ver-
sion 3.02 which adds new capabilities, par-
● bid request
ticularly in the area of ad fraud protection
● bid response (see more on ads.cert on page 60).
As it is useful to understand this protocol
in principle, it is discussed in more detail
later in this chapter, but only to a degree
that is relevant to marketers. For devel-
opers and those who want to delve deeply

Relationship between RTB

and OpenRTB
Real-time bidding (RTB) is a business
model for selling based on auctioning every
single bid (e.g. an impression).
OpenRTB is a particular communica-
tion protocol designed to allow different
systems to send bids to each other for
auction, thus creating a common language
for all systems.

History of OpenRTB
The text in this section is taken from the
OpenRTB protocol specification:

1 https://iabtechlab.com/wp-content/uploads/2022/04/OpenRTB-2-6_FINAL.pdf
2 https://github.com/InteractiveAdvertisingBureau/openrtb/blob/master/OpenRTB%20v3.0%20FINAL.md
 11

OpenRTB was launched as a pilot project between three demand-side plat-

forms (DataXu, MediaMath, and Turn) and three sell-side platforms (Admeld,
PubMatic, and The Rubicon Project) in November 2010. The first goal was
to standardize communication between parties for exchanging block lists.
Version 1.0 of the OpenRTB block list specification was released in December
20103.

After a positive response from the industry, Nexage approached the OpenRTB
project with a proposal to create an API specification for OpenRTB focus-
ing on the actual real-time bid request/response protocol and specifically to
support mobile advertising. The mobile subcommittee was formed between
companies representing the buy-side (DataXu, Fiksu, and [X+1]) and compa-
nies representing the sellside (Nexage, Pubmatic, Smaato, and Jumptap). This
project resulted in the OpenRTB Mobile 1.0 specification, which was released
in February 2011.

Following the release of the mobile specification, a video subcommittee was

formed with video ad exchanges (BrightRoll and Adap.tv) collaborating with
DataXu and ContextWeb to incorporate support for video. The goal was to in-
corporate support for display, video, and mobile in one document. This effort
resulted in OpenRTB 2.0, which was released as a unified standard in June
2011.

Due to very widespread adoption by the industry, OpenRTB was adopted as

an IAB Tech Lab standard in January 2012 with the release of version 2.1.
Governance over the technical content of the specification remains with the
OpenRTB community and its governance rules.

3 https://www.iab.com/guidelines/openrtb/
12

The principle of OpenRTB

The following figure shows an interaction 5. All bidders who have delivered a bid
between a seller and a buyer in an OpenRTB response are informed:
environment. a) The winner of the winning bid and
To summarize what is happening in the notification of the fees owed.
picture: b) Others of the failure.

1. A request is generated for the publisher 6. The publisher receives an advertisement

to deliver the ad. to display.

2. On this basis, a bid request is generated

from the seller towards the buyer.
3. The interested party (buyer) responds
with a bid response.
4. The seller selects the winner of the
auction.

Figure 1.1: OpenRTB protocol model

 13

Bid request
A bid request comprises objects that supply
it with all necessary information which
is delivered to the advertisers who then
decide – on the basis of the information –
whether and how much they are willing to
pay for the advertising position.

Figure 1.2: Object model of bid request in version 2.6

The figure above shows what information the bid request contains.
14

We can think of a bid request as a salesperson with a briefcase in which they carry
all the sales materials related to their product. Once this salesperson (bid request)
meets the potential customer (DSP), they present everything they have from the
briefcase and the customer can decide if they like it and are willing to submit their
bid.

Bid
request
DSP
size:
300x250
px
location:
Prague

device:
UserID: iPone13
YWxhZG

Figure 1.3: Visualization of a bid request

For the sake of clarity, here is an outline ● minimum acceptable price – floor price,
of the main information the bid request
● publisher restrictions – for example,
contains:
white or blacklisted advertisers,
● details about the user to whom the ad will
● legislative restrictions – whether the user
be displayed – their device, geo-location,
can be targeted according to different
UserID or other additional information,
legislation (GDPR, coppa, us_privacy),
● type of advertisement offered – banner,
● type of auction that will be used to select
native, video or audio,
the winner – first or second price model,
● format of the ad – dimensions, and many other parameters.
 15

Sample bid request

How a simple bid request can look is shown
in this example taken directly from the
OpenRTB documentation:

{
“id”: “80ce30c53c16e6ede735f123ef6e32361bfc7b22”,
“at”: 1,
“cur”: [ “USD” ],
“imp”: [ { “id”: “1”, “bidfloor”: 0.03, “banner”: { “h”: 250, “w”: 300, “pos”: 0 }
} ],
“site”: {
“id”: “102855”,
“cat”: [ “IAB3-1” ],
“domain”: “www.foobar.com”,
“page”: http://www.foobar.com/1234.html “,
“publisher”: {
“id”: “8953”,
“name”: “foobar.com”,
“cat”: [ “IAB3-1” OpenRTB 2.6 © 2021 IAB Technology
Laboratory iabtechlab.com/openrtb Page 47 of 85 ],
“domain”: “foobar.com”
},
“user”: { “id”: “55816b39711f9b5acf3b90e313ed29e51665623f” }
}
}
16

Bid response
Bid response is much simpler and comprises
fewer objects. It contains all the information
the advertiser sends to the publisher.

Response to this Request

Bid Response Bid Request

SeatBid

Bid for this Impression

Bid Imp Pmp

Bid under the term of this Deal

Deal

Figure 1.4: Bid response object model in version 2.6

The figure above shows what information

the bid response contains.
 17

We can think of a bid response as a bidder submitting an offer to a suitcase

merchant in the form of an envelope containing a price offer and information on
how to respond.

My account
ID: 123

device: Bid
iPone13 request
DSP

URL of my
price: ad: www.
$2,5

Figure 1.5: Visualization of bid response

For the sake of clarity, here is an outline

of the main information the bid response
contains:
● price offered,
● URL of the ad that will be displayed in
case of winning,
● URL that will be called in case of a win –
notification of the win,
● URL that will be called in case of failure –
notification of loss.
18

Sample bid response

How a simple bid response can look is
shown in this example taken directly from
the OpenRTB documentation:

{
“id”: “1234567890”,
“bidid”: “abc1123”,
“cur”: “USD”,
“seatbid”: [ {
“seat”: “512”,
“bid”: [ {
“id”: “1”,
“impid”: “102”,
“price”: 9.43,
“nurl”: “http://adserver.com/winnotice?impid=102”,
“iurl”: “http://adserver.com/pathtosampleimage”,
“adomain”: [ “advertiserdomain.com” ],
“cid”: “campaign111”,
“crid”: “creative112”,
“attr”: [ 1, 2, 3, 4, 5, 6, 7, 12 ]
}]
}]
}
 19

RTB 3.0
Ever-complicating supply chains, new tech- ● Ads.cert now requires all bid requests
nologies on the market, and always-evolv- be cryptographically signed, giving
ing header bidding multiplying bid requests advertisers better visibility of their
have all led to increased opportunities to impressions (see page 60).
abuse the system, as we will more fully
● OpenRTB 3.0 requires strict use of
explore in later chapters.
HTTPS protocol. HTTP is no longer
OpenRTB protocol has decided to counter sufficient.
this in version 3.0, which incorporates many
● Ad Management API6 now allows
innovative changes compared to previous
advertisers to have their creatives pre-
versions. The main goals in 3.0 are restora-
approved to save time during real-time
tion of transparency and protection against
bidding.
ad fraud. Unlike previous versions, in this
one we assume a more complex supply
chain and define relationships not between
buyer and seller, but between two entities
in the supply chain in general regardless of
whether they are buyer, seller, or one of the
many intermediaries.

The main changes are summarized

briefly in the following points:
● OpenRTB has split into several separate
parts, which, while working together
inseparably, will continue to be
standardized in their own ways. The
whole concept is covered by a layered
principle called OpenMedia4.
● OpenRTB will continue to deal primarily
with bids, auctions, auction types, etc.;
other accompanying information will be
handled by other standards in the future.
For example, ads, users, and devices
will be handled by the new standard
Advertising Common Object Model
(AdCOM) standard5.

4
https://iabtechlab.com/standards/openmedia/
5
https://github.com/InteractiveAdvertisingBureau/AdCOM
6
https://github.com/InteractiveAdvertisingBureau/AdManagementAPI
20

Figure 1.6: OpenMedia model

OpenRTB 3.0 (including the other Open-

Media components) was released in 2018,
but as it is not backwards compatible with
version 2.x, it will take some time before it
is fully integrated by all market players.
 21

Chapter conclusion
In this chapter we:
● described the building blocks of OpenRTB
protocol,
● showed how it will evolve in the future.

This chapter is slightly more technical than

the others, however, it is still meant for pro-
gramming professionals. If we want to get
a good understanding of the whole system,
we should understand the basic building
blocks on which it is based.
22

2. Basic principle of
supply chain
In programmatic advertising, the term This intricacy presents several disad-
“supply chain” refers to the path of goods vantages, the main two being:
between the advertiser and the publisher
● Ad fraud – Fraudsters abuse the complex
through intermediaries. Indeed, this term
system directly by stealing part of the
is used in industry in general, for exam-
money that advertisers spend or by
ple, to describe the path of a product from
planting technology, allowing them to
the manufacturer through a network of
steal again in the future more easily (see
resellers, suppliers, and dealers to the end
page 44 on fraudulent practices).
consumer.
Within the limits of programmatic buying,
we can think of a product as a final adver- ● Overcharging – Increasingly, no one
tisement that has been commissioned by knows how many nodes of the supply
an advertiser and displayed to a user on chain an impression has passed through
a website. Along the way, it has passed and whether they are all necessary and
through various technologies such as DSP, beneficial. This, of course, often results
SSP, and ad exchange among various other in the involvement of unnecessary
players in the programmatic ecosystem like oversellers and a reduction in the
Agency Trading Desks, ad servers, and ad amount of funding that the publisher
networks. We call these places that the ad actually receives. Or, conversely, the
passes through sequentially “nodes”. There advertiser will not buy the target space
are usually multiple nodes within a supply from the publisher because the publisher
chain. will sell it to another advertiser from
whom it will receive a larger volume of
Programmatic buying is often not simple
funding.
from a supply chain perspective. On the
contrary, it could be said that with new ● Over time, these drawbacks have
technologies and market development, it is proven so pervasive that the market
becoming increasingly complicated, all the has decided to address them. Some
while more and more actors are involved. methods are purely technical and must
be implemented by technology systems
(e.g. DSPs and SSPs), but some are
organizational and must be incorporated
by publishers and advertisers (see page
54).
 23

Basic building blocks

of supply chain
To further understand supply chains, we
have broken it down below from the very
basics to real-life examples of all their com-
plexity.

The simplest possible supply chain in programmatic buying might look like this:

Bid request
Ad SSP
Ad
DSP
exchange

Publisher Bid response Advertiser

Figure 2.1: Basic supply chain

In this example, there are only buying and Gradually, marketplaces (ad exchanges)
selling platforms and a single intermediary have evolved, and today they are mostly
– the ad exchange marketplace. already directly part of SSP or DSP technol-
ogies. Thus, we often do not think of ad ex-
The publisher uses a SSP as a platform to
change as a separate technology but rather
sell its advertising. The advertiser uses
as a role played by a familiar technology.
a DSP as a platform to buy advertising.
Between the SSP and the DSP is an inde-
pendent marketplace (ad exchange) that
brings together many publishers (their SSP)
and advertisers (their DSP) and conducts ad
auctions between them.
24

We can observe similar developments in largest companies such as Google, Ama-

DSPs, SSPs, or DMPs. Many technologies zon, or Facebook provide a comprehensive
today are bundled and provided in a single solution that includes all the named roles
package, and only in it are the roles then and are also closed to surrounding technol-
divided. We can go even further to the field ogies.
of the so-called “walled garden” where the

Imagine a natural gas producer in the US and a European customer who wants to
buy this gas (for example, an industrial plant). The producer has their gas auctioned
on an exchange and sells it to the highest bidder. In our case, the industrial plant in
Europe will pay the most. This exchange also acts as a transporter, transporting the
gas from the producer to the customer.

Between the publisher and the advertiser, However, there may be more than one
bids flow in both directions as defined by bidder (advertiser). Each publisher receives
OpenRTB protocol: the initial bid request and sends its bid
response to it. The ad exchange chooses
● Bid request to the advertiser – This
the highest bid response and sends it to
comes from the publisher (from their
the publisher’s SSP as the only one. The ad
SSP) and informs advertisers that they
from this bid response is then displayed on
have the option to display an ad to
the publisher’s website.
the user and provides all the relevant
information available. The whole process, however complex,
occurs in milliseconds.
● Bid response to the publisher – Once
the advertiser receives the bid request
(on its DSP), it decides whether it is
interested. If so, it sends a reply in the
form of a bid response. This contains,
in addition to the ad itself, the price the
advertiser is willing to pay.
The bid response copies the same path
back to the advertiser as the bid request
took.
 25

Resales within ad exchanges

This first example has been greatly simpli-
fied and does not reflect the reality of most
bids. Realistically, several resales (we call
companies “intermediaries”) happen in the
middle of the chain as exemplified in the
graphic below:

Ad Ad exchange
SSP exchange 1 DSP
3

Ad Ad Ad exchange
exchange 2 4
Publisher Advertiser

Ad exchange
5

Bid request

Bid response Ad exchange

6

Figure 2.2: An evolved supply chain

26

Our simple example of a natural gas producer in the USA and a customer in Europe
can be expanded. Say now that there are multiple producers represented by one
group on the exchange so that as a larger entity, they are in a better trading
position. There is not one exchange, but several, and the group of producers offers
their gas on most of them. In addition, the exchanges speculate and trade not only
between gas producers and customers, but also between themselves.

To explain why this happens, we need to This is analogous to the offline world.
understand the relationships between the Every seller wants to have as many buyers
actors involved in this chain (DSPs, SSPs, bidding for their products in the auction as
ad exchanges, and others). These entities possible. Likewise, every buyer wants to
usually collect a percentage fee for each have bids from as many suppliers as possi-
bid response they successfully clear. Their ble when selling their goods.
motivation is therefore to be connected
to as many partners as possible and thus
maximize the volume of bid responses they
can clear.

In our scenario, the exchanges are paid, for example, 1% of the trades executed.
Let’s say the exchange doesn’t have any gas from a producer, but has a customer
willing to buy gas. At the same time, it has a contract with another exchange that
currently has plenty of gas. So in this case, the exchanges sell gas to each other.

However, the technological interconnec- ternal timeout and various other elements
tion of the entities is not always 100%, so influence the decision making process. So
one entity (SSP 1) has different links to it follows that even if the same advertiser
other entities (ad exchange) than anoth- and publisher are at the beginning and
er (SSP 2). In addition, some companies end of the chain, the bid responses that
operate more than one part of the system, eventually get back to the publisher are
i.e. one entity operates both DSP and SSP different. Even if they came from the same
technologies. The transfer of bids between place and arrived at the same place, they
them is then generally faster, more fre- went through a different path.
quent, and omits the ad exchange.
In general, it makes sense to link all sys-
The type of auction the systems internally tems together as much as possible. For a
run determines which bid request to send long time this chaining worked well, but
on. (They have a choice of first-price or eventually the negative effects became
second-price auctions, although the first- apparent (see page 44).
price model already clearly dominates.) In-
 27

Impact of header bidding

on supply chain
Header bidding brings significant complexity Header bidding gave publishers the oppor-
to the supply chain. Before header bidding, tunity to present their bid requests in sev-
publishers used a waterfall system. This eral SSPs at the same time. They all broad-
gave a single SSP a chance to sell their cast bid requests to their partner networks,
space (inventory), and if that SSP failed, thus multiplying the number of potential
then the next SSP came along, and then the paths between advertiser and publisher. In
next. In this case, there was one system contrast to Figure 2.2, the possible paths in
(one SSP) at the beginning of the chain. header bidding look like this:

Publisher Advertiser

Ad

HB Ad exchange Ad exchange
wrapper SSP 1 DSP
1 4

SSP 2 Ad exchange Ad exchange

2 5

SSP 3 Ad exchange Ad exchange

3 6

SSP 4
Ad exchange
7
Bid request

Bid response

Figure 2.3: Supply chain including header bidding

We could complicate our example by having a gas producer connected not just to
one network representing the producers, but perhaps three, and making the same
offers to all three. All three would try to sell the gas through their contacts and the
producer would wait to see who offers the highest price.
28

But that’s not all. Today it is common for In the end, due to the many combinations
publishers to use multiple header bidding of paths, it can happen that an advertiser
solutions (wrappers). I discuss this issue competes with itself in several nodes of
in more detail in my book Programmatic the supply chain with its next offer coming
Header Bidding in Context, but here I’ll from another node.
briefly mention that the main wrappers
used are Google’s Open Bidding, Amazon’s
UAM/ATM, and Prebid Wrapper. The first
two take advantage of their exclusivity, so
if a publisher wants to have Google or Am-
azon bids involved in their auctions, then
they must use their wrapper as well. Each
wrapper can then use the same SSP (the
Google and Amazon wrappers are largely
represented in auctions by the same SSP
and ad exchange). This multiplies the com-
plexity and possible combinations within
the whole system.

Fees in supply chain

The goal of the different systems in a sup- requests, as many of them end up unan-
ply chain is to charge a fee for participating swered (i.e. no fee earned).
in the path between the advertiser and the
As a rule, unsuccessful bid responses are
publisher.
not charged. Only a single winning bid is
The common business model for all sys- charged. So in total, there is a set of bid
tems (whether SSP, DSP or ad exchanges) requests, a smaller set of bid responses,
is to charge a percentage of the winning and only one winning bid response:
bid response.
It is necessary to bear in mind that the fee
is applied to bid responses, which include
the ad bid and the price the advertiser is
willing to pay, unlike bid requests, which
contain the publisher’s ad placement bids.
The fee is applied to the winning bid, i.e.
the intermediaries who brokered the bid
response get the fee. This results in the
intermediaries’ tireless forwarding bid
 29

one
winning bid
all bid responses response

all bid requests

Figure 2.4: Subsets of a bid request

In our example, imagine that a producer drops the information that they have gas
to sell, at first among the sellers and afterwards, the exchanges. All entities start
exchanging information among themselves and search for a target customer. There
is a certain administrative burden on all of them. However, only those who succeed
in making the sale get a commission on the sale.

Let’s say the average fee is 10% and four

systems are involved in the supply chain.
We can simplistically say that the total
amount of fees between the advertiser
and the publisher will be 40%. Thus the
publisher receives only 60% of the amount
paid by the advertiser, or the advertiser
has to pay 40% more than what the pub-
lisher is asking for.
 30

This is just hypothetical, of course. Different studies have tried to uncover the reality of
the situation.

In 2020, ISBA conducted a study on this topic in the UK market7. The study included
15 advertisers, 12 agencies, five DSPs, six SSPs, and 12 publishers, representing
approximately £0.1bn of UK programmatic ad spend and nearly two-thirds of AOP
(premium publisher) digital ad revenues. You can see the result in the chart:

Supply chain findings

%
100
7%
8%

10%
75
100% 15%

8% 1%
50

51%
25

Figure 2.5:
0
Advertiser Agency DSP Technology Uknown SSP Technology Publisher
spend fee fee fee delta* fee fee revenue Supply chain
(demand side) (suply side)
findings Source: ISBA
In 2022, this study was repeated8 with noteworthy differences. Publisher revenue
increased from 51% to 65%, and there was a much better connection between the
sides of the publisher and the advertiser.
According to another study, “US Programmatic Fees 20199“, a full third of the
total spend within the US market in 2019 went to the area between publisher and
advertiser. We get similar data from other sources. For example, Magna Global’s
private network10 from 2017 shows system-wide fees at 55%, or a chart from
Pubmatic11 from 2010, which is even less hopeful for the publisher, shows that only
$1.8 of a $5 ad went to the publisher.

Over the years, you can see that the chain market, and gradually, greater transparen-
fees have been gradually decreasing. This cy throughout the industry.
is due to increased competition, a larger

7
https://www.isba.org.uk/system/files?file=media/documents/2020-12/executive-summary-programmatic-
supply-chain-transparency-study.pdf
8
https://www.isba.org.uk/knowledge/second-programmatic-supply-chain-transparency-study
9
https://www.emarketer.com/content/us-programmatic-fees-2019
10
https://www.emarketer.com/content/why-tech-firms-obtain-most-of-the-money-in-programmatic-purchases
11
https://pubmatic.com/blog/ad-tech-tax-or-a-reality-of-doing-business/
 31

These findings have one thing in common: where they really need them and where
Both the advertiser and the publisher see they bring value, or rather, whether they
the systems that feed off of them along need them all.
the way. They can and should consider how
to work with them and be selective about

Average number of participants

in supply chain
The studies in the previous chapter show for example, we see the possible paths to a
a framework of the fees for each of the single publisher for different advertisers as
technologies used. But how many systems outlined in the aforementioned ISBA study
do bid responses actually pass through? and also presented in the IAB webinar
Supply Chain Transparency & Sustainability
Simply put, each bid response has its own
in 202212:
length and if we want data, we need to
approach the task using statistics. Below,

1,400
1,231
1,200

1,003 1,019
1,000
854

800
663
586 598
600 541
490 503 493
432 444 449 453 438
399 406 411
363
400 342
289
236
201 204
200 165

53

0
Advertiser 1

Advertiser 2

Advertiser 3

Advertiser 4

Advertiser 5

Advertiser 6

Advertiser 7

Advertiser 8

Advertiser 9

Advertiser 10

Advertiser 11

Advertiser 12

Advertiser 13

Advertiser 14

Advertiser 15

Advertiser 16

Advertiser 17

Advertiser 18

Advertiser 19

Advertiser 20

Advertiser 21

Advertiser 22

Advertiser 23

Advertiser 24

Advertiser 25

MEAN

MEDIAN

Figure 2.6: Advertiser supply paths on 1 publisher Source: IAB Europe

12
https://www.youtube.com/watch?v=3iNOf69me5Y
 32

However, Well-known.dev’s analysis13 of all possible supply chains for the Marieclaire.fr
domain14 showed about 40 million possible path combinations:

Figure 2.7: 40mil possible supply chain combinations Source: well-known.dev

The above figure depicts general data and

all possible combinations. You’ll have to
find out what the specific supply chain is
for your bids yourself using the analytical
tools designed for this.

13
https://twitter.com/braedon/status/1567882264431988737
14
https://well-known.dev/resources/ads_txt/sites/marieclaire.fr
 33

Average number of SSPs used

by publishers
As time goes on, more and more demand HBIX tracker, the average number of
suppliers (SSPs and ad exchanges) are demand suppliers involved per publisher
interested in getting involved in header increased between 2017 and 2020, with one
bidding on the side of the publisher. They publisher having an average of seven de-
create connectors to the Prebid wrapper mand suppliers participating in the header
and other extended wrappers. This is well bidding auction in 2020 and only five three
represented by a comparison posted on years earlier.
adpushup.com15. According to the Adzerk

Figure 2.8: Average number of enquiry suppliers involved Source: Adzerk

It’s also interesting to look at the more

detailed layout provided by AdPushup’s
“AdTech Insights Report for Publisher”16:

15
https://www.adpushup.com/blog/header-bidding-adoption-three-years-in-review/
16
https://adpushup.wistia.com/medias/swzqclb2xx
34

Distribution of the number of demand partners

among publishers
32% 32%
28%

8%

1–3 Partners 4–6 Partners 7–9 Partners 10+ Partners

Figure 2.9: Distribution of the number of demand suppliers among publishers Source:
adpushup.com

In 2023, there were over 400 potential de- discuss this section in detail separately
mand partners that publishers could engage on page 17, but it should be mentioned
in header bidding, but they used only seven here that the more demand suppliers
on average. Logic says we should use the you use, the more paths you open up
maximum possible sources to increase com- between publisher and advertiser (supply
petition in the auction, so why only seven? chain). Each added path brings nega-
tives, most importantly reduced reve-
There are several reasons:
nues.
1. Zero-value added of demand part-
3. Difficulty of engaging a new source
ner – Many demand partners have no
– Even though the wrapper makes it
unique offering and only combine other
significantly easier to add a new demand
resources that the publisher already
partner, it’s still not insignificant. You
has engaged. We recommend engaging
should have a contract with each partner,
large demand partners like Amazon,
do an initial implementation, verify their
AppNexus/Xander, Pubmatic, Criteo,
data regularly, perform monthly billing,
Adform, etc., plus demand partners that
analyze their runtime, and edit settings if
have a clear value-add that makes sense
necessary. Doing this for seven partners
for you. Don’t be afraid to ask a new
is a lot less work than for twenty.
partner where their enquiries are coming
from.
2. Stacking the combination of bid paths
between publisher and advertiser – We
 35

4. Inappropriate technical integration by the

demand partner – Some demand part-
ners have very unconventional ways of
technical integration, which leads to the
fact that their wrapper connector is sig-
nificantly larger than your wrapper with
all the other demand partners or they
internally conduct other slow auctions
and don’t manage to respond to your
demands within the time limit.
5. Non-support of some other solution you
use – Header bidding is no longer about
the auction itself. Wrappers are often
connected to other systems and connect-
ing them can be a necessity that some
ad sources can’t handle. For example, it
may be a connection to GDPR or User ID
modules.
6. Technical limits on the number of web
browser connections and the impact on
page load speed – if you ask for five or
twenty-five suppliers, this will be reflect-
ed in the loading speed of your site.
7. Increasing demands for legal data
protection – GDPR in Europe and other
legislation around the world place clear
demands on data protection. Publishers
have regulations on how they can handle
data and an obligation to report on it.
Any newly involved demand partner
means not just more technical work, but
also more legal ramifications.
 36

The world of advertising technology is Take a look at which demand partners were
constantly changing and it is necessary to most used according to the HBIX tracker in
follow the current developments and adapt 2021:
to them. Even if you are currently using,
say, seven demand partners, keep an eye
on the market and test. No one can tell you
exactly which demand partner they are and
what number is optimal for you.

What adapters are header bidders using? (Top 10, 2021)

Amazon: 78%
Xandr: 75%
Index: 68%
Magnite: 65%
Pubmatic: 59%
OpenX: 49%
TripleLift: 38%
Criteo: 31%
Sovrn: 32%
District: 19%

This looks just at sites doing header bidding and shows what % of those sites are

Figure 2.10: Most commonly used adapters Source: Adzerk

It is important to note that this chart does

not apply regionally. The share of each
technology varies substantially across
markets.
Research on a similar topic was presented
on eMarketer.com17 in 2021 about specifi-
cally what demand partners US publishers
used in the last 12 months.

https://www.emarketer.com/chart/248339/which-supply-side-platforms-ssps-have-us-publishers-used-past-12-
17

months-of-respondents-march-2021
 37

Which SSP have US publishers used in the

past 12 months (March 2021)
Google ad manager 72%
Amazon publisher services 46%
Verizon Media ad platform 44%
PubMatic 44%
Xandr Monetize 42%
OpenX 42%
SpotX 42%
Index Exchange 40%
Magnite 36%
TripleLift 31%

Figure 2.11: SSPs that have US publishers used in the past 12 months Source: eMarketer

Average number of DSPs used

by advertisers
While from a supply chain perspective it
would be ideal for each advertiser to use
only one DSP, this is not generally the case
for several reasons. Nevertheless, there
is a gradual consolidation in the number
of DSPs used by advertisers. Research
presented on eMarketer18 from 2018 shows
a decline in the average number of DSPs
in use:

18
https://www.insiderintelligence.com/content/advertisers-cut-their-dsps-40-over-two-years
38

6.1

5.6

4.2

3.4 3.4

2016 2017 2018 2019 2020 2021

Figure 2.12: Average Number of DSPs Used by US Advertiser, 2016-2021

Source: Pathmatics, eMarketer

There are several reasons why advertisers DSP provide better features (i.e. ad fraud
use multiple DSPs: and supply chain optimization) that will
save you money in the end?
1. Differences in functionality – Each DSP
has certain strengths and weaknesses. 3. Differences in connected data – The
Some have very strong reporting tools, programmatic ecosystem is an open and
others stand out for their quick connec- connected environment. Most players try
tion to many other systems, and each to link to as many partners as possible,
has a different degree of dealing with ad but even then the linking is not compre-
fraud and supply chain validation. Due to hensive and is often done with different
high numbers of bid requests, DSPs use techniques.
bid throttling (see page 74), so they don’t
4. Walled garden – Some companies direct-
respond to every bid impression and
ly restrict other technologies’ access to
sometimes don’t get to their interme-
their space. Such companies act as DSP,
diaries. Occasionally some features are
SSP, ad exchange, and DMP all in one.
not well documented, and you have to
META (Facebook), Amazon, and Google
ask for support, and even that varies in
play a major role here, although Google
quality. This is also true in reverse. SSPs
restricts connections only in certain
have their own optimizations and decide
areas.
which partners to send the bid request to
and which not to. 5. Differences in reach (especially mul-
ti-publishers) – Each DSP tries to connect
2. Differences in fees – There is a differ-
to the maximum number of publishers
ence if one DSP asks you for a 15% fee
either directly or indirectly, but they
and another only 8%. But is the differ-
are not always successful, especially
ence justified? Can the more expensive
when connecting different devices. The
 39

moment a client wants to control cam- These, among other reasons, cause some
paigns in one DSP not only on the web advertisers to use multiple DSPs at once.
but also in apps and Connected TV, their If they do, they should be careful to pay
choice is limited, then. special attention to their supply chain. Ulti-
mately, they could be creating competition
6. Differences in the supply chain itself
for themselves and driving up the price of
– Each DSP is connected to different
their own campaigns.
systems, so bids go through different
paths. At different points, a bid’s path In general, when you drop something into
is affected by its price or connected OpenRTB, you never know exactly what
data. The type of auction (first-price or you’re going to get. Therefore, it is better
second-price) used by intermediaries in to carefully manage your supply chain
the supply chain and any data loss that whenever possible.
occurs in the process play significant
roles. Last but not least, the technical
integration on the side of the publisher
is also important, specifically which SSPs
are connected and in what way (waterfall
or header bidding, how many and which
wrappers are used, the connection to the
ad server). All of this can mean that one
SSP can perform much better with the
first publisher than another SSP with that
publisher, but conversely another SSP
performs better with a second publisher.
40

Loss of information
in supply chain
We’ve explored how the longer and more the way, but complexity can also bring
complex a supply chain is, the more likely negatives in terms of incremental loss of
it is that ad fraud or unnecessary fees information.
and technology will come into play along

Most kids know the game of silent mail. In a chain of several people, each whispers
information only to the next person who then passes it to the next until the
information has passed from the first to the last. The last player speaks out loud
what they heard from the previous person. In general, this is a very funny activity
where the original information is quite different from the information the last player
relays.

It’s the same in a supply chain. At first glance, it would seem that we are
eliminating human error with computers passing data to each other, but errors are
introduced at the level of the algorithm design of each system.

All it takes is one system in the chain to

omit the SupplyChain Object for the DSP
system at the end of the chain to discard
that purchase path as untrusted (even
though perhaps the other systems in the
chain were completely transparent).
Information loss could also occur as a
result of inaccurate or incomplete imple-
mentation of any standards or unintention-
al errors at the ads.txt or sellers.json level
inserted by an inexperienced publisher
employee.
As an example of possible complications,
let’s discuss data loss at the cookie match-
ing level.
 41

Cookie Matching
(Cookie Syncing)
Each advertising technology stores its Each time an ad request is made, the DSP
own cookies with a unique user identifier buying platform also retrieves these URLs
on each domain where it can advertise. from other technologies. It compares these
However, these cookies are only readable URLs to a table where user IDs with these
by their creator (or the domain that stored URLs are stored long term.
them). For example, if a demand partner
If technologies are able to “match” unique
stores a cookie on TopPublisher.com under
identifiers to their users in this way, it
the domain doubleclick.net, then this file is
means they are also able to match them
inaccessible to any other domain.
with the user data they collect.
Each demand partner therefore stores
This is a very simplified description. In
Cookie Matching (Cookie Syncing) its own
reality, it is important to note that cookies
set of cookies in the TopPublisher.com
are time-varying, in the sense that some
domain and cannot read other partners’
users disable them entirely in their brows-
cookies. This is crucial because the ad-
er, or occasionally delete them, or use
vertising ecosystem is made up of many
some blocking software. In these cases,
entities (SSPs, DSPs, DMPs, etc.) that pass
matching has no chance of taking hold.
information between each other.
The problem with cookie matching is also
In order to be able to target advertising to
that there is some drop-off each time a
users and to use multiple technologies in
match is made, because the matching is
the purchase chain, Cookie Matching, or
never 100%. See the example in the figure.
cookie syncing, must occur.
If the match rate in the pairing was 80%,
The whole process is based on each then after the third step you only have a
technology storing its own user ID for the 40% match rate.
domain in a cookie with a unique URL.

Data loss during retransmission

100M users 80M users 64M users 44,8M users

START Sync 1: Sync 2: Sync 3:

Advertiser’s With DMP 20% With DSP 20% With SSP 30%
target audience audience loss audience loss audience loss

Active cookies Lost cookies

Figure 2.13: Data loss during retransmission Source: The Trade Desk
42

Other factors affecting

supply chain
We have studied only some of the di- For more insight into these topics, please
rect actors in the supply chain, but by no look to other publications. For example,
means all of them. Our aim in this publi- how to work with GDPR and related legis-
cation is to explain the basic principles of lation is partially explained in our second
operation, not to describe every detail. book, Programmatic Header Bidding in
Context.
However, we will briefly mention two addi-
tional significant factors:
● Working with data, targeting, and DMP
platforms are closely related to supply
chains as we explained in the previous
chapter on cookies. For any targeting
to be meaningful, it must be fully
communicated through an entire supply
chain.
● Legislation around working with data is
important to take into account. If you
don’t have consent from the user, you
shouldn’t provide their data to other
players in a supply chain.

Chapter conclusion
In this chapter we: This chapter is the core of this entire
book. Knowledge of supply chain princi-
● explained what a supply chain is,
ples is very important for understanding
● explored why it can be very complicated other principles of programmatic buying
and who is involved, for every participant in the programmatic
● reviewed the motivations of the different ecosystem.
actors involved in a supply chain,
● described how many participants are
typically involved in a supply chain
on the sides of the advertiser and the
publisher,
● described how information is lost in
transmission.
43
 44

3. Supply chain
disruption: Ad fraud
Programmatic buying is characterized by But with huge opportunities come huge
the fact that it allows advertisers to buy risks. Because it’s such a big and open sys-
advertising on any domain or use any tem, it’s not difficult to abuse it. If we want
entity’s data available in an online market- to avoid exploitation, we need to know how
place. This can be done through almost any to defend ourselves.
technology, as the IT world is incredibly
accessible.

In our example, imagine a fake salesman claiming to represent a well-known

mining company. They make a deal and sell gas to a customer. But instead of gas,
they send them something else that is much cheaper and of lower quality, or even
replace the gas with air. This illicit exchange is only a small part of the total volume
of gas that the customer buys and therefore often goes unnoticed by the customer.

What is ad fraud?
According to Wikipedia19, “ad fraud is advertising don’t recognize this, then you
concerned with the practice of fraudulently risk buying advertising on a completely
representing online advertisement impres- different site.
sions, clicks, conversion or data events in
The very first major document to mention
order to generate revenue”.
ad fraud was “On the security of pay-per-
Ad fraud is any practice whereby one party click and other Web advertising schemes”20
(the fraudster) tries to deceive another back in 1994.
party and sell them something that looks
These abuses are not exclusive to buying
different from what it really is. This is done
banner ads. Ad fraud is now widespread in
to enrich the fraudster.
all areas of online advertising, and you can
For example, let’s say as an advertiser, you find fraudsters able to deliver fake clicks
want to buy advertising on a particular do- or orders (leads) in affiliate programs (see
main, XYZ.com, and you target your ad on page 46).
that domain. The fraudster in the program-
matic advertising ecosystem is running a
different site masquerading as the XYZ.
com domain. If the systems you use to buy

19
https://en.wikipedia.org/wiki/Ad_fraud
 45

The acronym IVT, which stands for “invalid traffic”, is sometimes used in connection
with ad fraud. These are visits that are not human, but they are not necessarily
fraudulent visits (ad fraud). Sometimes IVT is simply a bot that has visited a
website and makes no attempt to hide the fact that it is a bot. GIVT – “general
invalid traffic” – is used for these types of visits. The problem tends to be SIVT –
“sophisticated invalid traffic”. In this case, the bot tries to behave like a human and
is more difficult to detect.

Magnitude of ad fraud
In 2004, Google CFO George Reyes stated21 estimating that it would double between
that ad fraud is the biggest threat to the 2018 and 2023, reaching 100 billion US
Internet economy. dollars.
In 2021, Statista released22 an approxima-
tion of the costs associated with ad fraud

in billion U.S. dollars

120

100
100

81
80
65
60
52
42
40 35

20

0
2019 2020 2021 2022 2023

Figure 3.1: Estimated cost of digital ad fraud worldwide (in billion US dollars)
Source: Statista.com

21
https://money.cnn.com/2004/12/02/technology/google_fraud/
22
https://www.statista.com/statistics/677466/digital-ad-fraud-cost/
 46

In 2021, Forbes described23 some of the most prominent cases in which major brands like
P&G, eBay, and UBER lost hundreds of millions of dollars to various ad fraud practices.
All cases had one thing in common: advertisers paid for fraudulent ads that did not bring
them anything. The moment they turned off these ads, they had no impact on their ad
conversions.

Figure 3.2: Impact of disabling ad fraud on conversions Source: Augustine Fou for Forbes.com

Types of ad fraud
It is important to remember that the goal This publication deals with ad fraud only
of the fraudster is to scam you, so they marginally in relation to programmat-
will always think of new ways to achieve ic buying. We will briefly describe some
this. The whole industry therefore reacts to practices.
fraudsters with a lag, and it is impossible
to avoid them completely.

23
https://www.forbes.com/sites/augustinefou/2021/01/02/when-big-brands-stopped-spending-on-digital-ads-nothing-
happened-why/
 47

Malicious bots
The term “bot” appears frequently in the following lines, so we should define it now.
Wikipedia24 tell us “an Internet bot, web robot, robot, or simply bot, is a software
application that runs automated tasks (scripts) over the Internet, usually with the
intent to imitate human activity on the Internet. Internet bots are able to perform
tasks that are simple and repetitive much faster than a person could ever do.
The most extensive use of bots is for web crawling, in which an automated script
fetches, analyzes, and files information from web servers.”

“Bot” is the name given to simple pro- Generally, bots are also used in combina-
grams that mimic human behavior, e.g. tion with other types of fraud or target-
come to a page and after some time click ed attacks. Distributed denial-of-service
on a link or an advertisement. A bot can do (DDoS) tasks are very well known. In this
that on behalf of hundreds of thousands of activity, a particular website is over-
people at a time. whelmed with so many visits from bots
that loading speeds become so slow the
But no two bots are the same. The Inter-
site is rendered virtually useless.
net industry uses many types of bots, and
some are perfectly ethical and useful. Imperva publishes regular reports on bad
A typical example is a search engine bot bots. According to the 2021 report,25 an
(like Google) that scans the content of average of 25% of the traffic to any given
a page to use in its results. site is from bad bots while 15% are good
bots.
These bots make no effort to disguise
themselves and are easily recognizable
(GIVTs). Malicious bots are those that try
to impersonate real people and make their
owner money through deception (SIVTs).

24
https://en.wikipedia.org/wiki/Internet_bot
25
https://www.imperva.com/blog/bad-bot-report-2021-the-pandemic-of-the-internet/
48

25,6%
Bad Bots
Bad Bots Amount all Website Traffic in 2020 25.6%
Percentage change in bad bot traffic from previous year 6.2%

Good Bots Traffic Percentage in 2020 15.2%

Percentage change in good bot traffic from previous year 16.0%

15,2%
25.6%
Good Bots Human Website Traffic Percentage in 2020
Percentage change in human traffic from previous year 5.7%

59,2%
Human

Figure 3.3: Bad bots vs good bots vs human traffic 2020 Source: Imperva

Click farms
We’ve covered how bots can be misused, Ad fraud doesn’t have to be targeted.
but humans are also directly involved in There are also ad formats designed so
fraud, for example by being employed to that users click on them by mistake. For
click on ads or fill in enquiry forms. Such instance, the user thinks that by clicking a
people have no real interest in the prod- certain place, they will close the ad, but in
ucts they click on. fact the X they clicked on is part of the ad
itself. These people are then taken advan-
A group of such people (or even bots)
tage of by the person who placed the ad on
is called a click farm.
the page.

Cookie stuffing
Click farms are often (but not only) used from that page to another page where, for
by affiliate programs. When a user arrives instance, they fill out an enquiry form. Up
at one affiliate program page, the owner to this point, it’s above board. The whole
of that page gets paid when the user clicks system works on storing third party cookie
information.
 49

The moment the affiliate program page then make an order (lead) on any of these
launches other pages with other affiliate other sites in the future, then whoever
programs in the background, the user’s abused their cookie will get a commission
cookie is applied to them as well. If they from that lead.

Domain spoofing
This is one of the oldest scamming tech- fact the advertising is on a completely dif-
niques which no longer works very well to- ferent site. The text “nytimes.com” is even
day due to the ads.txt standard and other written into all the advertiser’s statistics.
methods described on page 50.
There is a very famous Methbod26 case
The practice is based on the fraudster where a scammer made $3 to $5 million
creating a web page that they pass off as every day using a false domain.
another. Let’s say they spoof the nytimes.
com domain. Shoppers then think they are
buying advertising on that domain, but in

$5.00

SSP1 Domain
✘

DSP $2.00
$0.50

Spoofed
✔ SSP2 Domain

Figure 3.4: A spoofed fake domain

26
https://www.whiteops.com/hubfs/Resources/WO_Methbot_Operation_WP.pdf
 50

Malvertising
Also known as “malicious advertising”, this in the supply chain to thus gain access to
is the practice by which fraudsters pur- the entire network. The sheer volume of
chase advertising that contains malicious advertising and the fact that each of us
code on reputable sites. If a user clicks on now receives personalized, tailored ads
it, their device is infected with malware or makes identifying such malvertising very
they are redirected to another fraudulent difficult.
site. In some cases, they don’t even have
The first case of malvertising was reported
to click – simply viewing the ad takes ad-
in 2007, exploiting vulnerabilities in Adobe
vantage of a bug in the browser code.
Flash27.
Fraudsters exploit the complexity of the
programmatic ecosystem and the multi-
tude of participants they are able to attack

Ad injection
Submitting an ad to a page without the
owner’s knowledge is a subset of adware.
This can happen if a user installs a pro-
gram that includes malware that arranges
for ads to be inserted on someone else’s
site. Such ads are then only visible to the
user with the specific program installed.
The ads are real, the page they appear
on is real, the user who sees them is real.
However, the revenue from the ad does
not go to the publisher of the page, but to
the impostor who published the program
installed by the user.

27
https://www.avast.com/c-what-is-malvertising-how-to-stop-it
 51

AdWare
Adware refers to software that is placed on The difference between adware and mal-
a user’s device and generates real adver- vertising is that adware makes money by
tisements and presents them to the user. displaying ads on the device where it is
The user often takes them as real ads of installed whereas malvertising is an adver-
the domain he/she is visiting. tisement that can be displayed to anyone
and aims to attack the user’s device.
Adware must be installed on the user’s
device somehow, and this is usually done
without the user’s knowledge.

How to recognize ad fraud

Most ad fraud can be identified from sta-
tistical data that can be found in Google
Analytics or any other analytics tool. Some
analytical tools have special software for
detecting ad fraud. There are even expert
groups that update lists of known bots
such as the IAB/ABC28.
Ad fraud is usually characterized by worse
performance of the same behavior as
non-fraudulent ads. For example:
● Incoming traffic from a fraudulent
source has much worse performance
than from other sources (e.g. high
bounce rate or low time spent on the
page).
● Incoming traffic is from a single IP
address or from a limited set of IP
addresses.
● It delivers the usual results you would
expect from a real ad, but does so at a
significantly better conversion rate.
If you see any of this in your stats, then it
could be ad fraud and you should investi-
gate the case further.

28
https://iabtechlab.com/software/iababc-international-spiders-and-bots-list/
 52

How to fight ad fraud

Many preventative measures have been and others that make the entire purchas-
taken by systems that participate in pro- ing supply chain and who participates in
grammatic buying such as DSP, SSP, ad it transparent. This makes it harder for
exchange and others. In particular, fighting fraudsters to fake their bids. The Pixalate
fraud is about following standards like IVT Benchmarks Report29 shows how pure-
ads.txt, Sellers.json, SupplyChain Object, ly using app-ads.txt limits ad fraud.

40%

37%

33% 33%
30% 31%

26%
25%
20%

10%

No No No
Ads.txt Ads.txt Ads.txt Ads.txt Ads.txt Ads.txt

0%
Google Apple Google+Apple

Figure 3.5: Invalid traffic benchmark report (Q2 2022) Source: Pixalate

As an advertiser who doesn’t want to buy a DDOS attack. Since the originators are
fraudulent ads though, you have other op- usually different each time, it is very diffi-
tions. As mentioned, you can detect many cult to prevent these attacks and the only
fraudulent visits in your analytics tool, and thing you can do is to react quickly.
if you can spot a fraudster, you can elimi-
Some types of ad fraud can be detect-
nate him.
ed with special tools that usually focus
That being said, some types of fraud and not only on detecting ad fraud, but also
attacks can no longer be handled by the ad low-quality advertising like an ad in the
buyer, but must be identified and blocked footer of a website that users don’t see.
by the webmaster. A typical example is You need to pay extra for such tools, but

29
https://www.pixalate.com/hubfs/Reports_and_Documents/Ad%20Fraud%20Reports/2022/Pixalate%20-%20Q2%20
2022%20IVT%20Benchmark%20Report.pdf
 53

if ad fraud is a pertinent problem for you,

then it’s definitely a good investment. Rep-
utable tools include Moat30, Double Verify31,
Integral Ad Sciencein32, and Pixalate33.
In 2016, the World Federation of Advertis-
ers also published a document34 on how to
defend against ad fraud.
We will discuss the protection options on
the side of the advertiser in more detail
in the chapter on supply path optimiza-
tion and on the side of the publisher in
the chapter on demand path optimization
(page 68).

Chapter conclusion
In this chapter we: The aim of this chapter is to describe ad
fraud and its dangers. With this knowledge,
● explained what ad fraud is and the
you will understand the importance of the
scope of the problem,
following chapter, which describes what
● described the basic practices that ad projects aim to defend against ad fraud.
fraud uses,
● presented ways to identify ad fraud.

30
https://www.oracle.com/cx/advertising/measurement/ 33
https://www.pixalate.com/
31
https://doubleverify.com/ 34
https://www.wfanet.org/app/uploads/2017/04/WFA_
32
https://integralads.com/ Compendium_Of_Ad_Fraud_Knowledge.pdf
 54

4. Protection –
Transparency
projects
Having introduced the complexity of In order to make this system more trans-
supply chains, we can see that this level parent, the IAB consortium defined the
of convolution is almost asking for abuse ads.txt standard in 2017 and added seller.
and greatly complicates the possibilities to json and SupplyChain Object in 2019 and
detect and avoid it. later in 2021, Buyers.json and DemanChain
Object.
If we hadn’t started taking precautions
against ad fraud and the constant length-
ening of the supply chain, the entire pro-
grammatic ecosystem would be in danger
of collapsing.

Ads.txt
After the IAB introduced the ads.txt proj- Like a robots.txt file for example, ads. txt
ect in 201735, it quickly gained popularity can only be published on a domain by the
and the number of publishers that started publisher’s webmaster, making it authen-
using it skyrocketed to 50%36 in the first tic. Moreover, it is a text file that is easily
year. Today, ads.txt is already strictly editable and readable by all entities.
required by some demand partners and
The ads.txt file can be periodically scanned
without it, they won’t accept demands at
by advertisers, i.e. DSP systems, to check
all. For larger publishers, this is practically
the eligibility of sellers to offer these do-
a necessity. If you are a smaller publisher
mains.
or your site is monetized by one specific
system, you may not need ads.txt.
Ads.txt is a simple way for publishers to
declare who is authorized to sell their
inventory, increasing transparency for
advertisers.

35
https://iabtechlab.com/press-releases/iab-tech-lab-launches-assault-on-illicit-advertising-inventory/
36
https://www.emarketer.com/content/ads-txt-adoption-continues-its-steady-growth
 55

Example: Let’s say there is the domain, publisher.com. An advertiser receives an

offer through the XYZ system for an impression on this domain and is interested in
buying it. He can easily verify that the XYZ system is listed in the publisher.com/
ads.txt file. If it is not, it is probably an ad fraud and the advertiser can discard the
offer.

The following picture may help to explain:

Ads. Domain
txt A.com

Ads. Domain SSP Ad exchange DSP

txt B.com

Ads. Domain
C.com Crawled
txt ads.txt
data

Figure 4.1: Explaining the function of the ads.txt file. Source: IAB

The ads.txt file has a clearly prescribed

structure37 which is being slightly refined
over time. For example, the 2022 version
of ads.txt. 1.1 includes an extension to
define the relationship between the domain
owner and sales managers.

37
https://iabtechlab.com/ads-txt/
 56

Basic file structure:

< SSP/Exchange Domain >, < SellerAccountID >, < Pay-
mentsType >, < TAGID >

For example:
greenadexchange.com, 12345, DIRECT, AEC242
blueadexchange.com, 4536, DIRECT
silverssp.com, 9675, RESELLER

Explanatory notes for each item:

- SSP/Exchange Domain – The official name of the demand partner that is

authorized to sell advertising on the domain. This name is supplied by the
partner itself.

- SellerAccountID – A unique identifier that the seller assigns to the publisher’s

account.

- PaymentsType – Describes the type of relationship between the publisher

and the seller. It can only take the following options:

- DIRECT – Indicates that the publisher directly controls its account with
the partner.

- RESELLER – Indicates that the reseller is authorized to sell the domain.

- TAGID – An optional parameter indicating the partner ID in the Trustworthy

Accountability Group (TAG)38.

Within the world of mobile apps, there is an chain, is to make the market transparent
equivalent to ads.txt in the app-ads.txt file, and optimize buying processes, so publish-
but it has minor differences that you need ers should pay attention to the ads.txt file
to pay attention to. and keep it up to date.
For many publishers, the ads.txt file has
become a repository of text records that
demand partners dictate to them and
publishers don’t think about. The goal of
ads.txt, along with sellers.json and supply

38
https://www.tagtoday.net/
 57

Good policies include:

- checking what records have already been added. Often, demand partners
use the same ad exchange and their lines are then repeated.

- removing entity lines from your file when the cooperation has terminated.

- inserting text notes in your file to help you keep track of the records. For
example, it is useful to note who asked for particular lines to be inserted.

- being the only one who marks your direct sellers as DIRECT, marking other
resellers with the RESELLER flag.

- regularly checking your ads.txt file with one of the checking tools.

As an example of a maintained file, here is relationship with the publisher. However, it

ads.txt from the NY Times publisher39. It has several different platforms available to
contains only 25 records (as of the time of it. Other examples: theguardian.com: 55
publication of this book – 2023) and all of entries, bbc. co.uk: 32 entries, bild.de: 30
them are with the PaymentsType parame- entries... And so we could go on. Clearly
ter DIRECT. Thus, the publisher does not the big publishers are careful about who
allow any reselling of its ad space and any- they do business with.
one who wants to buy it must be in a direct

Figure 4.2: Content of the ads.txt file on nytimes.com in February 2023

39
https://www.nytimes.com/ads.txt
 58

Seller.json and
SupplyChain object
The sellers.json project is directly related of the seller (the demand partner) and de-
to ads.txt and was introduced by the IAB clares its relationship to the publisher.
consortium in 201940.
The actual path between the advertiser
The ads.txt project helps publishers to de- and the publisher (i.e. each entity on it)
clare which demand partner is authorized is stored in a SupplyChain Object41 that is
to offer them and is published on the pub- part of the RTB protocol.
lisher’s domain. The sellers.json project, on
Technically, this can be shown as follows:
the other hand, is published on the domain

Ads. Domain
txt A.com

Ads. Domain SSP Ad exchange DSP

txt B.com

Domain
Ads. Crawled
C.com sellers.json sellers.json
txt SupplyChain
Object data

Figure 4.3: Explanation of supply chain functionality

This gives advertisers an idea of how far Below is an example of how such optimiza-
their offer has to travel before it reaches tion can take place. The figure shows two
the publisher and how many participants examples, before and after optimization:
are involved in the path. With this, DSP
tools can optimize their buying paths and
eliminate certain players along the way.

40
https://iabtechlab.com/press-releases/iab-tech-lab-announces-two-new-technologies-to-build-more-transparency-
trust-in-the-programmatic-supply-chain/
41
https://github.com/InteractiveAdvertisingBureau/openrtb/blob/master/supplychainobject.md
 59

Before optimalization

Ad Ad Ad Ad
SSP exchange exchange exchange DSP
1 2 3

After optimalization

Ad SSP DSP

Figure 4.4: Difference in the bid path before and after optimization

Setting up the sellers.json record must be done by the publisher in the demand
partner system that manages their ads. In most cases, this is part of the system
administration, where they must fill in:

- Seller_id: the unique code of your account in the system

- Name: the name of the seller
- Domain: the domain of the seller
- Seller_type: indicates the type of seller. Can take values:
- PUBLISHER – if the publisher directly manages the account with the seller
- INTERMEDIARY – if it is a reseller
- BOTH – if it is a combination of both of the above

We recommend reading the entire docu-

“seller_id”: “158945”,
mentation42, including the very useful FAQ
section with examples of use. “name”: “NY Times”,
An example of a sellers.json file can be “domain”: “Nytimes.com”,
found for each participant in the program- “seller_type”: “PUBLISHER”
matic ecosystem. See, for example: https://
cdn.pubmatic.com/sellers/ data/sellers.json
and the entry for the NY Times publisher
contained there in:

42
https://iabtechlab.com/sellers-json/
 60

Ads.cert
While ads.txt, sellers.json, and Supply- Ads.cert is based on the principle of bid en-
Chain Object have reduced fraud, they cryption and two keys, one public and one
certainly haven’t eliminated it completely. private, known only to the publisher. The
Already, another standard from the IAB launch will take place with the transition to
has been announced43 to further clean up the RTB 3.0 protocol (page 10).
the market under the name ads.cert.

Buyers.json
and DemandChain Object
In 2021, the first definition of buyers.json The actual path between the publisher and
was created by the IAB 44. It is the opposite the advertiser (i.e. each entity on it) is
of sellers.json and its essence is to unique- stored in a DemandChain Object45, which is
ly identify the advertiser behind the offer. part of the RTB protocol.

Technically it can be shown as follows:

DSP 1 buyers.json

Domain.com SSP Ad exchange DSP 2 buyers.json

DSP 3 buyers.json

Crawled sellers.json
DomandChain
Object data

Figure 4.3: Explanation of supply chain functionality

43
https://iabtechlab.com/press-releases/openrtb-3-0-beta/
44
https://iabtechlab.com/wp-content/uploads/2021/03/buyers-json-1-0.pdf
45
https://iabtechlab.com/wp-content/uploads/2021/03/DemandChainObject-1-0.pdf
 61

This gives publishers an idea of how far Thus, every buyer system starting with the
their offer has to travel and how many par- DSP should have a buyers.json file stored
ticipants are involved in the journey. With on their site to identify themselves and the
this information, SSP tools can optimize entities they are authorized to buy for.
their buying paths and eliminate certain
actors along the way.

The buyers.json file should contain the following for each buyer:

- Buyer_id: the unique code of your account in the system

- Name: buyer’s name
- Domain: domain of the buyer
- Buyer_type: type of buyer. Can take on the following values:
- ADVERTISER – if the buyer is a direct customer – typically an advertiser
account in a DSP
- INTERMEDIARY – if the buyer is a reseller – typically an ad exchange or SSP
- BOTH – if it is a combination of both of the above

SupplyChain
and DemandChain Objects
SupplyChain and DemandChain Objects ● A DemandChain Object starts with the
are part of the RTB protocol definition and advertiser and progressively stores each
store each system participating in the intermediary from the reseller up to the
trade. In principle, the only difference be- publisher, and the publisher can read it
tween them is the direction in which they and optimize their sales journey based
store the records: on it.
● A SupplyChain Object starts with the
publisher and sequentially stores each
intermediary from the reseller up to
the advertiser, and the advertiser can
read it and optimize their purchase path
based on it.
62

SupplyChain and DemandChain Objects cessor and puts it into a larger box where
expand over time and get bigger the more it adds its own records. The box then gets
systems a bid goes through. Think of them progressively larger as it contains more
as a box in which each system stores its nested boxes.
records. The next system in line then
takes the box that came from its prede-

Figure 4.6: SupplyChain and DemandChain Objects on the example of nested boxes

An example SupplyChain Object might look like this:

Domain Ad net Ad Exchange

.com SSP1 DSP 1
1 1

Bid request Bid request Bid request

Figure 4.7: Example SupplyChain Object

 63

SupplyChain validation
The standards (ads.txt, sellers.json, For each reseller, the SupplyChain Object
SupplyChain Object, Buyers.json, and stores its identifiers asi, sid, and rid. These
DemanChain Object) defined by the IAB must be the same as what the reseller
Tech Lab are technical. It is imperative that declares in its sellers.json file. This verifies
publishers, advertisers, and programmatic that the reseller listed in the SupplyChain
platform developers fully integrate and use Object is indeed authorized to engage the
them to their fullest extent. given bid taken from the previous system
in the chain. But this is only part of the
This is happening gradually. At the time
story.
of this publication, for example, ads.txt is
already almost fully integrated. Sellers.json In the sellers.json file on the domain of the
and SupplyChain Object still have a way last reseller (usually the SSP of the pub-
to go, but advanced systems are already lisher), there is a seller_id record, which
working with them. On the other hand, contains the name and domain of the site
Buyers.json and DemandChain Object are that is authorized to represent and offer its
just at the beginning of their lives. advertising. This information must match
the information in the ads.txt file on the
Ads.txt, sellers.json, and SupplyChain Ob-
publisher’s domain to ensure that the data
ject are already integrated with advanced
is verifiable.
systems, and it makes a lot of sense for
both advertisers and publishers to inte-
grate these systems into the supply chain.
Now we can start checking on the side of
the advertiser in their DSP for the contents
of the SupplyChain Object and eliminate
risky incoming bid requests when neces-
sary. Thus, we cut off any incoming bid
requests that are not completely transpar-
ent, may have been attacked by ad fraud
along the way, or may hide intermediaries
that we don’t want to engage.
We introduced the contents of the Supply-
Chain Object and how it grows incremen-
tally in the previous chapter (page 58).
Now we’ll take a closer look at the bindings
to ads.txt and sellers.json.
64

The publisher then declares in its ads. information stored with the SupplyChain
txt what systems are authorized to bid Object about each reseller, this creates
on its bid requests, and each system in a verifiable series of records that can be
turn declares in its sellers.json what other checked.
systems it works with. Together with the

Ads. Domain
txt A.com

Domain Ad Ad DSPs/
Ads. SSP Advertiser
txt B.com exchange 1 exchange 2 buyers

Domain sellers.json sellers.json sellers.json

Ads. Crawled
txt C.com
ads.txt
data

Figure 4.8: Sample logic of data stored in a SupplyChain Object

How to check if the settings

are correct
If you’re a publisher, you’ll be interested ● https://tools.iabtechlab.com/
to know how you can check the accuracy
● https://iabtechlab.com/software/supply-
of your ads.txt and sellers.json records in
chain-validation/
your SSP systems.
● https://well-known.dev/resources/ads_
Fortunately, there are validators that will
txt/supply-chains/validate
not only help you detect errors, but also
point out other potential problems:
65
 66

Support for individual projects

In 2021, IAB Europe conducted a survey46
on how agencies and advertisers use
transparency technologies. It clearly shows
that ads.txt is the most widely adopted by
agencies. On the other hand, quite sur

70%
56%
56%
Yes

56%
54%

21%
Agencies

31%
29%
No

28%
26%

9%
Don‘t know

12%
16%
16%
19%

67%
65%
67%
Yes

57%
69%

26%
Advertisers

26%
25%
No

33%
21%

6%
Don‘t know

9%
9%
10%
11%

0% 10% 20% 30% 40% 50% 60% 70% 80%

Ads.txt Sellers.json Supply Path Object Buyers.json Demand Path Object

Figure 4.9: Level of support for individual transparency projects

46
https://iabeurope.eu/wp-content/uploads/2022/10/IAB-Europe-Attitudes-to-Programmatic-Advertising-Report-
2022-V3.pdf
 67

Chapter conclusion
In this chapter we: If publishers and advertisers do not want
● explained the concepts of different to become victims of ad fraud, they must
transparency projects, defend themselves. This chapter describes
exactly how they can do this and what
● reviewed how publishers and advertisers tools they have at their disposal. These
can work with them, tools need to be set up well so that we can
● described exactly how they prevent the then effectively optimize the entire supply
spread of ad fraud. chain system as we will show in the next
chapter dealing with SPO and DPO.to get a
good understanding of the whole system,
we should understand the basic building
blocks on which it is based.
 68

5.
SPO and DPO
Supply path optimization (SPO) is a prac-
tice for advertisers. Publishers will find
interesting information in the following
demand path optimization (DPO) chapter
(see page 74).

Supply path optimization (SPO)

We have extensively discussed the com-
plexity of supply chains, and we’ve ex- In our example, SPO could be used by
plained how data from the SupplyChain gas buyers to see which sales groups
Object helps us to understand it. and exchanges are making significant
So now the question for advertisers is how sales and which are only occasional
to use this knowledge. SPO has emerged to and small contributors. Buyers would
simplify the buying path, cut out unneces- also see the quality of their trades,
sary nodes in the buying chain, and reduce and based on this data, they would
costs.
terminate their cooperation with
The term SPO probably first appeared in some of them. This would give them
2016 in an article by Brian O’Kelley fewer potential suppliers but more
(AppNexus)47. room to manage and nurture their
remaining relationships which will
lead to buying more volume at the
SPO is a set of practices that allows
same price and better quality.
us to optimize the entire purchase
and achieve better results.

With the tools SPO gives us, it is possible Actions that are carried out within the
to reduce the number of bid request paths framework of SPO can be divided into two
(potential supply chains) and thus simplify types:
purchasing decisions. Simply put, instead
● Automated execution by technology
of thousands of bids, you’ll only be decid-
platforms.
ing between hundreds or dozens without
compromising the quality of the bids you ● Manual execution by campaign
receive. You cut out 80% of the volume managers.
and focus on the remaining 20% of quality
bid requests.

47
https://bokonads.com/preventing-bad-behavior-in-header-bidding-supply-path-optimization/
 69

Before we break them both down in detail, the use of specific tools. SPO is only a set
we must add a very important point. SPO of practices that allow us to optimize the
does not restrict what media or ad formats entire purchase and achieve better results.
advertisers can buy, nor does it prohibit

In 2020, IAB Europe published “The IAB Europe Guide to Supply Path Optimisation”,
which summarizes SPO and offers specific practices. The guide is regularly updated,
with the latest version released in 202248.

Automated purchase path

optimization
Most purchasing platforms (DSP systems) and suggests optimal supply chain pas-
today have tools for automating SPO activi- sages or alerts administrators to potential
ties. DSPs have data about the purchase problem nodes.
path from projects such as ads.txt, seller.
We have partially discussed this issue in
json, and SupplyChain Object. This data
the chapter on supply chain validation,
is digital and machine-readable and can
and we will further introduce bid throttling
therefore be machine evaluated and acted
optimization technique at the end of this
upon.
chapter on page 74.
While some optimizations can be done in
real time, it is extremely time consuming
and not very useful, which is why many
systems only perform them once a week,
for example. As a result, not all changes
made by publishers or other participants in
a supply chain are immediately reflected,
which needs to be kept in mind throughout
the optimization.
The entire automated SPO process is
greatly aided by machine learning, which
evaluates large amounts of historical data

48
https://iabeurope.eu/knowledge-hub/iab-europe-guide-to-supply-path-optimisation-spo-updated-june-2022/
 70

What purchasing platforms

can automate
They can: and thus they help the advertiser
(ad buyer):
● machine-verify the links between the
different actors in the supply chain, ● decide to respond to the most
transparent and credible bid,
● machine-verify who is authorized to
represent the publisher, ● remove bid requests that may be
infected with ad fraud,
● see the entire bid request path and who
is involved, ● use information from the SupplyChain
Object to help them in their bid
● see if a node in the supply chain has
throttling decision algorithm,
been disrupted and part of the path is
missing, ● find a better/shorter route through the
supply chain.

Manual optimization
of the purchase path
The document “The IAB Europe Guide to Supply Path Optimisation”49 describes well the
essence of SPO:

At its core, supply path optimization (SPO) is a method for improving advertising
outcomes by minimizing the number of exchanges with whom a buyer or advertiser
works to achieve the most efficient path to ad inventory. Such improvements
fall into four key categories: greater innovation, savings, increased trust and
transparency, and an optimized user experience. It allows buyers to assess the
effectiveness of their supply-side partners (SSP) and identify opportunities for more
transparency and innovation in their bidding processes.

Manual work on the side of the advertiser

puts demands on industry workers to try to
map the routes they take to buy advertis-
ing. They uncover the added value and fees
of all participants and then make optimiza-
tions.

49
https://iabeurope.eu/wp-content/uploads/2022/06/Guide-to-Supply-Path-Optimsation-SPO-2022-update-FINAL.pdf
 71

What SPO optimizes

● Using SPO tools, budgets can be ● SPO can significantly help the
reallocated across campaigns and advertiser choose between platforms
media, eliminating losses caused by as a whole, regardless of fees. Some
either ad fraud or an unnecessarily platforms simply don’t sufficiently
complex supply chain. This directly leads meet the parameters of transparency
to reduced advertising costs or reduced and protection against ad fraud, or
conversion costs. they are unable to meet your specific
requirements.
● Simplifying a supply chain increases
transparency to advertisers. Thanks ● SPO assists the advertiser in negotiating
to SPO, advertisers limit the number the quality of space they provide with
of possible supply chain paths and are publishers and, if applicable, at what
thus able to better monitor and optimize price and through what buying channel.
them.
● SPO puts the advertiser in a better
position to negotiate fees for the
technologies it uses. Once you remove
the unhelpful intermediators, you can
negotiate fees for the increased traffic
you will drive through the remaining
ones.

Setting up SPO in four steps

So how do you approach SPO? Below we
have outlined a possible procedure general
1. Determine KPIs
enough to give you a foundation on which Think about your media buying up until
you can make your own adjustments. It today. What are the objectives and meas-
consists of four steps: urable criteria (KPIs)? Ask yourself the
following questions:
● What is my buying model? (CPC, CPT,
1. Determine KPIs
CPA, etc.)
2. Analyze of the current state
● How do I measure performance?
3. Limit the number of partners (number of impressions, number of site
4. Consolidate and optimize visits, number of conversions, cost, etc.)
● How do I measure quality? (visibility, IVT
ratio, etc.)
Let’s break down these steps in more detail.
72

● Do I have preferred vendors? analytics tools to assess the ratio of ad

(publishers, SSP, ad exchanges, DSP, fraud for each vendor and the average
etc.) length of each supply chain.
● What percentage of my budget am I able
to spend on technology fees and how
much on humanizing my resources?
● Establishing clear KPIs is key for 3. Limit the number
any business activity, and SPO is no
exception.
of partners
Start with the tools you know on your side,
then move to the publishers’ tools. The
first one is usually DSPs. Ask yourself, how
many are you using, and do you really need
2. Analyze the current them all given your KPIs and current state
situation analysis? Ideally, do a current state analysis
for each DSP separately and compare them.
Analyze your current supply chain using
Can you fully replace one DSP with another?
the knowledge presented in the previous
If you want to keep a DSP, define clearly
chapters. To get started, ask yourself the
what the reason is and how it differs from
following questions:
the others.
● Can you describe your supply chain for at
Once you’ve gone through the DSPs, take
least 80% of bid requests? If not, review
the same look at ad exchanges, SSPs, and
the reasons why.
other technologies your DSPs use.
● What are your target publishers? Rank
SPO is based on the principle of simplifica-
them according to the size of your
tion, so it’s a good idea to reduce your num-
spread. Then classify and categorize
ber of partners. Before you do this, discuss
them. For example, what are your top
your findings with concerned partners (DSP,
ten with the:
SSP, ad exchanges, etc.) to ensure you
– most expensive CPT? know their strengths and weaknesses, what
– best CPA, visibility, number of other systems they are connected to, how
conversions, etc.? they are unique, and what fees they apply.

– shortest supply chain? Always try to find the most direct route to
your preferred publisher. Within the part-
– subjectively most important to you? ners, focus on technologies that can set up
● If you use special formats, do this such an ideal path for you. For example:
analysis separately only within the media ● Use whitelists to define and blacklists
that support them and see the format for to exclude where you want or don’t
a particular publisher – sometimes there want your ads to appear. For example,
are multiple options under one name. many advertisers gradually expand their
● Analyze not only the performance data blacklist with domains that produced
for individual SSPs and ad exchanges below average results or are even
from your DSP, but also use specialized suspected of ad fraud. Other advertisers
 73

only define a whitelist with verified It is advisable to make the first reduction
domains and IP addresses where they carefully and to exclude only some part-
want to advertise. ners. Gradually, and after further analysis,
this limitation should continue until you
● Build programmatic deals with publishers
reach the minimum meaningful number of
or participate in limited Private
balanced partners that can still meet your
Marketplaces (PMPs). With these tools,
KPIs.
you control your supply chain and can
prevent undesirables from penetrating it. Even here the process does not end. As we
said, the entire programmatic ecosystem is
changing and evolving over time, and you
will occasionally have the need to try out
additional technologies or partners or to
4. Consolidate and return to some that you have used in the
optimize past.

Once you’ve gone through the three steps

above, make the first reductions in the
SPO is a never-ending process.
technologies and partners you use.
The following figure shows the optimization
Know, though, that a programmatic envi-
of a purchase path where one SSP with a
ronment is an ever-changing ecosystem,
fraudulent domain and one SSP with the
so you will need to monitor and optimize it
highest price are excluded:
regularly. Conduct regular analyses based
on the KPIs you’ve set.

DSP
✕ ✕ $2
5 .30
.0
$2.35

$2
$5.05

SSP 1 SSP 3
Spoofed SSP 2 High Price
SSP 4
Domain

Publisher Publisher Publisher Publisher

Figure 5.1: Example of a manual SPO

74

Demand path optimization (DPO)

Demand path optimization (DPO) is essen- For example, they have auction data from
tially the opposite of SPO – it is primarily a header bidding such as win rate, response
technique for publishers to optimize reve- time (timeout), CPM of all bid responses,
nue from advertisers by clearing the buying etc. This data allows the publisher to dis-
path. cern the quality of bids from each supplier’s
enquiries and thus understand the contribu-
The goal of DPO is not just to increase
tion of a particular supplier.
eCPM, but to bring transparency and trust
to the buying process and make sure the In contrast, publishers do not see the data
entire system is set up optimally, meaning from the advertiser’s perspective. That is,
that advertisers have a direct straight path they lack an overview of all bid requests
to you with a minimum of oversellers. and are forced to decide which bid response
to send without the advertiser’s perspec-
As we have shown, knowledge of SPO is es-
tive.
sentially important for publishers, but pub-
lishers have different tools than advertisers.

In our example, DPO could be used by a gas producer to see which sales groups
and exchanges are making significant sales and which are only occasional and small
contributors. It would also see the quality of their trades and, based on this data,
terminate cooperation with some of them. This would give them fewer potential
sellers but more room to manage and nurture them, which will lead to them earning
more income for their gas and creating better relationships with their customers.

Automated sales path

optimization
Unfortunately, the DemandChain Object to advertisers. That is, they will be able
and Buyers.json projects (see descrip- to see the entire supply chain and make
tions on page 60) have a relatively small optimizations to it, including employing
percentage of coverage so far. Their mass machine learning algorithms.
adoption is yet to come.
However, publishers have a slightly dif-
When that happens, publishers will have ferent position vis-à-vis advertisers. The
supply chain automation capabilities similar number of advertisers a publisher wants
 75

to satisfy is always much higher than the

number of publishers a particular adver-
tiser wants to use. Publishers will have to
work with much more robust solutions in
DPO and limit the number of partners very
carefully.

Manual optimization of
the sales path
In SPO, the advertiser tries to minimize the
number of partners in order to maximize its
KPIs. The publisher is in a slightly differ-
ent position. The publisher’s goal is not to
reduce the number of partners or adver-
tisers, but to increase the revenue they
generate by leveraging their supply chain
with the largest advertisers and redirecting
bid responses.
Therefore, any reduction in partners must
be done very cautiously, and before any
reduction takes place, it is necessary to
assess the loss and whether it will be offset
by other partners.
76

What DPO allows to optimize

● DPO tools can be used to reallocate ● DPO will allow the publisher to identify
advertiser spends to media and limit the key players in the market (such
losses caused by either ad fraud or an as advertisers or agencies) and try to
unnecessarily complex supply chain. This establish a direct supply chain link with
directly leads to an increase in publisher them.
revenue (yield).
● DPO will enable you to find bulk problem
● Cleansing and simplifying a supply chain nodes with the supply chain. This is
leads to increased transparency for both not only from an ad fraud perspective,
publisher and advertiser. but also from a delivered ad quality
perspective. There exist DSPs or ad
● DPO improves position in negotiating
networks that collect ads of questionable
fees for the technologies the
quality or large data volume.
publisher uses. By removing unhelpful
intermediaries, you can negotiate
fees with the remaining ones for the
increased traffic you will be providing
with your advertisers.

Seven useful steps for

an effective DPO
Below is an overview of a few steps that ● Analyze your suppliers (SSP, ad
every publisher should take into account exchange) and their overlaps. Remove
when setting up DPO: the redundant ones gradually and
carefully.
● Make sure you and your partners have
set up the ads.txt and sellers.json files ● Do not include demand suppliers
correctly. Perform regular validation. (SSP, ad exchange) that do not have
something unique. It is unnecessary (and
● Make sure your bids contain maximum
sometimes downright harmful) to bring
additional information, such as user
in a partner that has 99% of the same
identity tools, proper linking to CMP
offerings as your current suppliers.
tools, and forwarding approvals to all
demand partners. ● Invite major advertisers to a Private
Marketplace (PMP) or arrange a direct
● Analyze not only the performance data of
programmatic deal. This minimizes the
each system (SSP, ad exchanges, etc.), but
number of systems between you, and
also use specialized analytical tools to show
the price is clearly agreed upon including
the ad fraud ratio of each supplier and the
technology fees.
average length of each supply chain.
 77

● Keep an eye on the quality of the ads that route, but at the same time you won’t
delivered. If a bidder regularly delivers block specific advertisers because they still
poor quality ads, then say goodbye. have many ways to reach you.
Each publisher may see the term “poor
You should always analyze carefully before
quality” as something different, but
you cut off any major demand partner.
common indications of poor quality
Generally speaking, you should start cutting
include promotion of services on the
back on smaller suppliers and gradually
edge of ethics, ads exceeding certain
move to larger ones, as smaller suppliers
data limits, or overly aggressive look or
usually copy the demand of larger ones and
feel of ads.
rarely have their own added value.
Finally, always ask suppliers about the
If your analysis shows that a supplier is not
volume of unique enquiries in their system,
beneficial to you, don’t be afraid to part
that is, anything that cannot be obtained
ways with them. By removing their entry
other than by engaging with their system
from ads.txt, you’ll let the market know that
only.
your space is no longer available through

Bid throttling
Bid throttling is a method in which a DSP or This of course has huge demands on the
SSP does not handle all bid requests, but technical infrastructure of supply chains.
selects only some. Technologies have come up with a solution
called bid throttling in which they decide
When discussing bid throttling, you will
whether to send (in the case of SSPs) or
come across the term “queries per second”
respond (in the case of DSPs) to a given bid
(QPS). This term has a general meaning
request.
outside of programming50, but here it refers
to the number of bid requests sent to the Their decision depends on many factors,
server. all of which are reflected in a probability
calculation of how likely it is that a given
With the advent of header bidding, the
bid request will be successfully sold. If this
number of bid requests that go to demand
probability is highly below average, then bid
partners has increased massively. While in
requests will gradually start to be throttled.
waterfall one bid request went to all part-
ners in turn, in header bidding all suppliers
get a bid request immediately and simul-
taneously. So, supposing you have seven
demand partners, you are sending seven
times more bid requests than in waterfall.
 78

When bid throttling occurs

DSP limits its responses if it has purchased on mobile ad spaces when the end
a below-average number of bids in the past. consumer is viewing the desktop version.
Let’s say a DSP keeps statistics that show ● There could be flawed implementation of
that on an average domain, advertisers lazy loading where auctions occur even
buy 5% of the bid responses they send. If though the ad may never be viewed.
they find that on an average domain they
● The ad server buys most impressions,
only buy 0.1% of the bid responses, then
so even if an auction takes place and the
logically advertisers will try to limit their
result is submitted to the ad server, the
responses because they are inefficiently
ad server discards it in most cases and
serving them.
prioritizes the internal campaign.
Let’s look at some typical causes of bid
throttling:
● On the side of the publisher, there may
be a flaw in the auction implementation
where bid requests will go unanswered.
For example, auctions are conducted

How to limit bid throttling

It is possible to limit bid throttling by in- start limiting their bids, or they can even
creasing the chances that bidders succeed employ artificial intelligence to speed up
in the auction. The key is to eliminate the the optimization process. While this point
aforementioned technical errors that lead to relates to visibility, bid throttling is more
unnecessary auctions. about whether lazy loading is performed
correctly, i.e. the auction should occur just
Another step that can reduce bid throttling
before the site visitor scrolls to the relevant
(although removing technical errors is ab-
ad position.
solutely the most important) is bid caching.
This allows us to use a bid we have from The number of demand partners being
a previous auction. Its validity makes it approached also plays a role. If you are
unnecessary to approach the supplier of the reaching out to five partners per ad unit in
requests again. each auction, their chance of winning the
auction is 20%. If you have ten partners
Another area that we touched on is the
involved in the auction, then their chance is
correct implementation of lazy loading. If
only 10%. In addition, some percentage of
you’re not using it, then you are probably
the impressions will use the ad server, so
running an auction for areas on a web page
the chances of the demand partners win-
that the user will never see. An advertis-
ning decrease again.
er’s visibility stats can help them gradually

50
https://en.wikipedia.org/wiki/Queries_per_second
 79

There are also significant differences in the There are many other parameters that play
paths bids have to travel. Most technologies a role in limiting bid throttling. The specif-
have a much more efficient path if they ic algorithm for how bid throttling works
run both an SSP and a DSP and have set is handled differently by each technology.
up a server-to-server connection between In 2015, the folks at DSP Turn presented
them. For this reason, it is advisable for the an insightful technical study on how they
publisher to set up programmatic deals in approached this topic51.
the SSP technology in which the advertiser
uses the DSP.

How to spot bid throttling

If you are a publisher, no one is likely to call tive, but you can get a good estimate by
you and say, “We’ve started bid throttling comparing bid response statistics from
your domain.” You have to find out for your- different request suppliers. If they differ
self, and probably the best way to do this is strongly somewhere, then this may be a
to look at auction statistics. case of bid throttling.
If you’re sending a lot of bid requests but
the bidders are not responding well, then
it’s probably bid throttling on their part. The
term “under-responsive” is highly subjec-

Chapter conclusion
In this chapter we: Most of the previous chapters explained
● explained the concepts of SPO and DPO, how an entire supply chain works and what
adjustments to make to reduce ad fraud.
● introduced the options publishers and It is only in this chapter we described the
advertisers have to reduce ad fraud, options available to specific actors and how
● presented best practices for publishers they can use their knowledge of how the
and advertisers. whole system works to optimize it.

51
http://www0.cs.ucl.ac.uk/staff/w.zhang/rtb-papers/turn-throatling.pdf
 80

6. Sustainable
Programmatic
Supply Chain
In recent years, sustainability and the re- In this chapter, we will first present the
duction of greenhouse gas emissions have general framework of international agree-
been increasingly discussed in the context ments and standards that underpin this
of advertising. The programmatic supply area and then look in more detail at the
chain is as much a part of this issue as any advertising industry and its supply chain.
other product supply chain.

General framework
The Paris Agreement52 is a legally binding Each country has committed to different
international treaty on climate change that deadlines. India, for example, has set a
replaced the Kyoto Protocol. It was adopted goal of zero emissions by 2070, while Saudi
by 196 Parties (States Parties) in Paris on Arabia and China have committed to zero
12 December 2015 and went into action on emissions by 2060. The US, the EU, the UK,
4 November 2016. Japan, and most other UN greenhouse gas
emitters have committed to reducing their
Among other things, the agreement
emissions by 50% by 2030 and to achieving
● articulates a long-term climate protection zero emissionsl by 205053.
goal to hold the increase in the global
average temperature to well below 2°C
above pre-industrial levels and to pursue
efforts to limit the temperature increase
to 1.5°C,
● brings about a significant change in
terms of greenhouse gas reduction
commitments. It includes commitments
from not only developed but also
developing countries to set nationally
determined contributions to reduce
emissions and to achieve the goal of the
agreement.

52
https://unfccc.int/process-and-meetings/the-paris-agreement/the-paris-agreement
53
https://press.un.org/en/2020/sgsm20411.doc.htm
 81

The goal is to limit global warming to well

below 2°C, preferably to 1.5°C above
pre-industrial levels. Scientists have also
expressed these goals graphically:

NET NET
LIMITING GLOBAL ZERO ZERO
CO2 GHG
WARMING TO 1.5°C ENTAILS
2020 2030 204025% 2050 2060 2070 2080 2090

LIMITING GLOBAL
WARMING TO 2°C ENTAILS
NET NET
ZERO ZERO
CO2 GHG

Figure 6.1: Global timeline to reach net zero emissions Source: IPCC

The figure shows that net zero CO2 emis-

sions must be achieved by 2050 (1.5°C)
or 2070 (2°C) at the latest. For non-CO2
emissions, this means we must reach net
zero by 2060 (1.5°C) or by the end of the
century (2°C).
 82

Global warming impacts

In support of the Paris Agreement, a The AR6 report outlines mitigation strate-
report54 was produced on the impacts of gies that cover all major sectors including
global warming of 1.5°C compared to the energy, transport, buildings, cities, industry,
pre-industrial era. forestry, and agriculture.
The Intergovernmental Panel on Climate IPCC scientists believe that if these climate
Change (IPCC) has released the third and change mitigation measures are implement-
final part of its AR6 report55 on climate ed, the world could cut emissions by at least
change. Together with the previous two half by 2030.
AR6 reports, this report is of fundamental
While rapid reductions need to happen in all
importance for scientific climate analysis.
sectors, the energy sector has the biggest
In general, it provides scientific data on impact. This means a significant reduction
annual greenhouse gas emissions between in fossil fuel consumption and large-scale
2010 and 2019. It also presents possi- electrification will be necessary. It also re-
ble solutions to reduce emissions. These quires greater energy efficiency and the use
solutions are the consolidated ideas of 278 of alternative fuels such as hydrogen.
scientists from 65 countries.

CO2 versus other

greenhouse gasses
Along with CO2, five other gasses are
designated as greenhouse gasses (GHGs):
methane, nitrous oxide, hydrofluorocar-
bons, perfluorocarbons, and sulfur hexa-
fluoride.
The IPPC report shows that CO2 comprises
64% of greenhouse gasses and its elim-
ination is therefore the most important
objective. It is clear that most human be-
havior results in the production of CO2, and
as industrialization progresses, the volume
of CO2 in the atmosphere increases. This
comes with many implications, such as the
warming of the planet.

54
https://www.ipcc.ch/sr15/
55
https://www.ipcc.ch/report/ar6/wg3/
 83

World

35 billion t

30 billion t

25 billion t

20 billion t

15 billion t

10 billion t

5 billion t

0t
1750 1800 1850 1900 1950 2021

Figure 6.2: Annual CO2 emissions Source: Ourworldindata.org

Net zero
The only way to reach the Paris Agree- ● Analyze your current emissions.
ment’s aims is through reducing green-
● Aim to reduce your emissions as much
house gas emissions. Many companies are
as possible, ideally to zero.
beginning to pursue a “net zero” strategy.
This refers to the effort to be neutral, i.e. ● Offset the emissions you fail to eliminate
to either not produce any emissions or to by purchasing carbon credits (page 84).
compensate for those emitted in some All such procedures must meet public
way, for example, by planting new trees. standards and be audited by an indepen-
Over 70% of the world’s economy has dent organization. It should be repeated
already committed to this program, but regularly, for instance every year.
everyone understands net zero differently, Therefore, many activities aim to reduce
hence the creation of the Science Based CO2 production. Most activities start with
Targets initiative (SBTi56). The SBTi is the calculating how much CO2 we actually
world’s first science-based standard for produce.
setting net zero targets for businesses
(the Net-Zero Standard57). There are three
steps to get there:

56
https://sciencebasedtargets.org/
57
https://sciencebasedtargets.org/net-zero
 84

Calculation of CO2 production

Calculating CO2 production is not a simple electric cars. Clearly, someone produced
matter. To simplify the process, several this energy by burning fossil fuels, for
standards have been developed. One of the instance.
best known is the Greenhouse Gas Proto-
● Scope 3 emissions – These are emissions
col58.
that we cause indirectly and do not have
The Greenhouse Gas Protocol divides emis- a direct influence on, for example, the
sions into three categories, or scopes: emissions that are produced during the
manufacturing of the car you drive.
● Scope 1 emissions – These are emissions
we produce directly, for example, by
burning fuel in cars or traveling by plane.
● Scope 2 emissions – These are emissions
that we produce indirectly, for example,
by consuming the energy we buy in

CO2 SF6 CH4 N2O NF6 HFCS PFCS

DIRECT
INDIRECT EMISSIONS
EMISSIONS FROM OTHER INDIRECT
PURCHASED ENERGY EMISSIONS
BUSINESS
TRIPS

PURCHASE OF TREATMENT OF
ELECTRICITY, HEAT PURCHASED
OR STEAM MATERIALS AD
SERVICES
WASTE
MANAGEMENT

COMBUSTION
OF FOSSIL
FUELS PROCESSED RENTING OR
PRODUCTS LEASING
COMPANY VEHICLES
Scope 1 VEHICLES OUTSOURCED
Scope 2 ACTIVITIES
Scope 3

Figure 6.3: Three scopes of emissions

58
https://ghgprotocol.org/
 85

For some companies, the calculation of companies, especially large ones, to report
greenhouse gas emissions is mandatory. emissions. This includes Scope 1 and 2
For one, the Securities and Exchange Com- emissions, as well as Scope 3 emissions if
mission has issued its own GHG disclosure they are found to be significant.
rules59.
The proposed regulation requires public

CO2 production
One tonne of CO2 can be thought of as a Everyone can calculate their own carbon
cube with a side length of 8.2 meters (27 footprint in the online calculator61.
feet). The table below shows the distance
we can travel by different means of trans-
port and produce 1 tonne of CO2:

Flight 2,527

Train 17,500

Electric Car 14,600

Large Car 3,556

Small Car 5,737

0 2.000 4.000 6.000 8.000 10.000 12.000 14.000 16.000 18.000 20.000

Miles

Figure 6.4: Distance traveled for 1 tonne of CO2 Source: Emsmastery.com60

59
https://carboncredits.com/sec-carbon-emissions/
60
https://emsmastery.com/2022/07/15/what-is-1-ton-of-carbon-dioxide-and-how-it-relates-to-daily-life
61
https://www.carbonfootprint.com/calculator.aspx
 86

Carbon credits
Have you already analyzed how much CO2
your activities release into the air, optimized Typical greenhouse gas emitters
them, and still can’t get to zero? If so, are airlines. In their current form,
you’re not carbon neutral. That’s common, they will never be able to reduce
as most industrial activities inevitably lead
greenhouse gas emissions to zero,
to an increase of CO2 in the air.
but they can become carbon neutral
On the other hand, there are activities that by financially supporting socially
have exactly the opposite effect by taking beneficial projects.
CO2 out of the air, but these actions are
in other fields unrelated to your business.
Typical examples would be agriculture and
forestry.
It takes 50 trees growing for one year
Often these projects cannot be imple- to remove 1 tonne of CO2 from the
mented because they have no commercial air 62.
benefit. The solution? They are actually
subsidized by the CO2 producers in a system
of tradeoffs. Those who emit more CO2 can
buy “credits”, also called offsets, from those
who take CO2 out of the air.
Carbon credits (carbon offsets) are issued
by companies that remove CO2 from the
atmosphere. These offsets are then sold to
companies that emit (or have emitted) CO2
into the atmosphere.
They are usually traded in terms of tonnes
of CO2. Buying one credit will subtract one
tonne of CO2 you have produced. Through
this system, you can become a carbon
neutral company even if you are a net CO2
producer (see page 87).

62
https://www.climateneutralgroup.com/en/news/what-exactly-is-1-tonne-of-co2/
 87

Carbon credit life cycle

The life cycle of a carbon credit is made up ● Registration & issuance – If a project
of four phases: passes the first two stages successfully,
it is listed in a register and credits are
● Development – Each project wishing
assigned. The project can then dispose
to sell credits must undergo a
of them, meaning sell them to interested
methodological assessment according
parties.
to the Verified Carbon Standard (VCS)
methodology, for example. On this basis, ● Retirement – Issued credits can be
the exact environmental impact of the sold or traded until they are “used”, or
project is calculated. implemented in projects to reduce the
company’s carbon footprint. From this
● Validation/verification – At the next
point on, used credits are marked in
stage, the project must be assessed by
the registry and cannot be disposed of
an independent auditor to verify that
further.
the project actually does what it claims
to do. For this purpose, there are many
carbon credit ratings companies – e.g.
Sylvera63 – that provide ratings of
projects, some using letter grades such
as AAA, AA, or A.

Carbon credit markets

There are two types of carbon credit mar- ● A voluntary market exists because
kets: some companies consider it socially
responsible to reduce their emissions or
● A compliance market is a market
even to be neutral. Whether it is because
regulated by national or international
of shareholder pressure or because it
regulations. It operates on a cap-and-
is a good PR move, they are not forced
trade basis. Regulators issue a certain
to do so by the regulators. They do it
number of permits and allocate them
voluntarily.
to polluters. However, there are always
fewer allowances than the polluters Voluntary markets are made up of environ-
need and this forces them to reduce mental projects that aim to reduce emis-
their emissions. If they do not do this sions or promote other socially beneficial
effectively enough, they have to buy activities. Examples include reforestation,
these allowances from their successful regenerative agriculture, recycling, solar
colleagues who have an excess of energy, and socially beneficial projects in
allowances. The current price of credits developing countries. More examples of pro-
can be monitored online64. jects can be found at GoldStandard.org65.

63
https://www.sylvera.com/
64
https://carboncredits.com/carbon-prices-today/
65
https://marketplace.goldstandard.org/collections/projects
 88

These projects are grouped together into Voluntary markets are growing. Globally,
larger units (programs), which give them there is a growing demand that industrial
a common framework. The framework polluters who are not regulated voluntarily
includes (but is not limited to) exactly how contribute to beneficial projects, and these
the project’s contribution to CO2 reduction projects are expanding as well. Below is
will be calculated and also defines how a visualization of voluntary carbon credits
and by whom the results will be audited to issued since 2005 in metric tons of CO2.
ensure they are credible. One of the most
Beneficial projects can also receive public
well-known programs is Verra – The Verified
funding. For example, the UK is investing
Carbon Standard66.
£60 million in 15 successful projects69.
Trading of credits is possible on exchanges
such as Xpansiv67, or can be carried out
directly by trading between the polluter and
the beneficial project, however there are
often traders or intermediaries in between.

300

250
Offsets Issued (MtCO2e)

200

150

100

50

0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 August
2021

Figure 6.5: Voluntary carbon offsets issued Source: carboncredits.com68

66
https://verra.org/
67
https://xpansiv.com/
68
https://carboncredits.com/the-4-best-carbon-offset-programs-for-2022/
69
https://carboncredits.com/uk-carbon-removal-rechnologies/
 89

It should be added that not everyone is a

fan of voluntary markets. The most vocal
critic is Greenpeace , which argues that
these activities will not ultimately work and
that the only thing that will save humanity
is an immediate reduction in fossil fuel con-
sumption regardless of economic impact.

How the advertising industry

produces emissions
Electricity
Within the advertising industry, electricity
is the most important source of energy. It
powers servers, computers, and lights in
offices. Therefore, most of the CO2 emis-
sions caused by our industry are attributed
to electricity generation.
But electricity generation is not the same
globally. Some countries rely on nuclear
power, which has minimal CO2 emissions,
or renewable energy. Others, on the other
hand, have historically produced most of
their electricity by burning fossil fuels.
Energy prices, method of production, and
the resulting emissions can be tracked on-
line. You can use the map to find out which
sources your country’s energy comes from
and the emissions it produces on average.
The differences between locations are
significant.

70
https://carboncredits.com/the-greenpeace-revolt/
71
https://app.electricitymaps.com/zone/PL
90

< 50 gCO2eq/kWh
50-75 gCO2eq/kWh
75-100 gCO2eq/kWh
100-200 gCO2eq/kWh
200-300 gCO2eq/kWh
300-500 gCO2eq/kWh
> 500 gCO2eq/kWh

Figure 6.6: Electricity maps emissions Source electricitymaps.com

We have selected only a few countries to

demonstrate the substantial differences.
Notice a 33-fold increase in emissions be-
tween Iceland and Poland:

Countries gCO2eq/kWh
Iceland 27
France 104
Czech Republic 562
Germany 693
Poland 910
 91

Servers are the factories of the 21st centu- greener in places where electricity is gen-
ry. Take a look at these interesting findings erated with low CO2 emissions.
on how much CO2 just one server in Nordic
For the reasons described above, Google
countries72 emits based on the type of
switched to a fully renewable energy sup-
electricity used:
ply in 2017. Amazon plans to achieve this
● Cloud server using 100% green by 202573. By 2030, Google is aiming to run
electricity: 160 kg CO2e / year its business on carbon-free energy every-
where, at all times74.
● Cloud server using non-green electricity:
487 kg CO2e / year
● On premise or data center-server using
100% green electricity: 320 kg CO2e /
year
● On premise or data center-server using
non-green electricity: 975 kg CO2e / year
It must be taken into account that the
Nordic countries have some of the smallest
carbon footprints. In other countries, these
emissions will be much higher. Running
websites and online advertising is much

Energy consumption by the user

Another significant influence on CO2 emis- For developers, GreenFrame.io is a tool that
sions is the energy consumption of the calculates energy consumption and asso-
target devices. For example, a simple web ciated greenhouse gas emissions. This is
page without images, ads, or other embed- done based on the energy consumption of
ded scripts and analytical tools consumes each part of the site being analyzed – page
much less energy than other “bulky” pages. code uploads, ads, analytics tools, etc.
You can check how green your website is by From the authors’ point of view, one of the
visiting websitecarbon.com, for example, most interesting pieces of information is
which uses available data such as page size that up to 70% of the energy consumption
and data center information to estimate (and therefore carbon emissions) caused by
the amount and type of electricity used by visiting media sites is triggered by ads and
viewers of your webpage and converts it statistics75.
into carbon emissions.

72
https://www.goclimate.com/blog/the-carbon-footprint-of-servers/
73
https://sustainability.aboutamazon.com/environment/the-cloud?energyType=true
74
https://sustainability.google/progress/energy/
75
https://marmelab.com/blog/2022/01/17/media-websites-carbon-emissions.html
 92

Online advertising is comparable

to the airline industry
It is estimated that the internet accounts for ● Displaying an ad on the target site – The
2-4% of global CO276 emissions, comparable target site is a server and also consumes
to the aviation industry. energy for its operation.
Let’s walk through the energy consumption ● Measurement – We also need to take into
of an average advertising campaign: account the energy consumption of the
other tools we attach to the ad, like the
● Production of materials (banners,
analytics tools on both the advertiser’s
videos, etc.) – For this we need energy
and publisher’s side.
to run the computer. The bigger and
more complex the advertisement is, the As data from the Scope3 platform shows,
more energy is needed. For example, there are significant differences in how
producing a static banner is many times much emissions an ad impression produces
cheaper than making a 15-second video. depending on which domain it was displayed
on:
● The flow of advertising through the
supply chain – Each server consumes
energy to run. The more servers a bid
request goes through, the more energy
is consumed.

Figure 6.6: gCO2e/imp – grams of CO2 per 1,000 impressions Source: Scope3.com

76
https://www.adexchanger.com/the-sell-sider/advertisers-have-an-environmental-responsibility/
 93

The difference is due to many factors. One line advertising, has also set up the Green
of the most important is the cost of elec- Media Products (GMPs) initiative. Its aim is
tricity generation in a given location and to get advertisers to buy advertising only
how it is produced. from publishers that are net zero.
Scope3, which has developed a tool to
calculate greenhouse gas emissions in on-

Scope3 and Ebiquity published a study77 examining display impressions from 43

brands totaling over $375 million. As part of this study, they also introduced a
new CO2PM metric – an abbreviation of gCO2ePM, or grams of carbon dioxide and
equivalent greenhouse gas emissions per 1,000 advertising impressions.

Domains 55.2 4,782.8

Markets 345.7 1,206.3

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

CO2PM rangers

Figure 6.7: Variance in CO2PM Source: Scope3 & Ebiquity

The results present a variance of CO2PM generated by the domain impressions from
55 to 4782 with an average of 680g CO2PM. There is a huge difference in emissions
between domains with long load times, complex content, and complicated ad supply
chain versus optimized domains with optimized supply chains. But the specific
market in which the campaign is running is also important for many reasons, the
most important being the way electricity is generated in that location. This is why
data for the US and France, for instance, are not comparable, because most of
France’s energy comes from nuclear power plants, which have a much smaller
carbon footprint.

77
https://www.ebiquity.com/news-insights/viewpoints/the-hidden-cost-of-digital-advertising/
 94

Video is many times worse

than banners
Online video generates 60% of the world’s
traffic, and thus more than 300 million
tonnes of CO2 per year78. Video is therefore
one of the most powerful contributors to
online CO2 emissions.
Naturally, the same applies to video adver- All non
video uses
tising. The carbon footprint of a video ad Online
20 %
is many times larger than that of a banner. video
Video is larger in volume, so it is much Other 60 %
more challenging to forward between nodes video
20 %
in the supply chain, then much more chal-
lenging to physically display (play) on the
end user’s device.

Figure 6.6: gCO2e/imp – grams of CO2 per

1,000 impressions Source: Scope3.com

Attention economy
The quality of advertising also plays a big one-third of ads get the full attention of
role in the relationship between emissions their audience.
and advertising. Here we refer to quality in
An IAS study80 also found that ads placed
the sense of:
within high-quality editorial content attract
● Was the ad visible to the end user (was it up to 3x more attention compared to regu-
in-screen)? lar digital ads.
● How long did the user actually see the Simply put, an ad plugged into the content
ad, and did they pay attention to it? of a page and displayed when the user is
paying attention is far more effective than
● How many other ads were displayed
other forms of advertising. This fact drives
around the ad in question, or conversely,
the effectiveness of special ad formats com-
was the ad surrounded by page content?
bined with content. These are then more
In fact, Dentsu Aegis Network’s The Atten- effective in terms of brand KPIs (brand
tion Economy79 survey revealed that only uplift, attention, viewability, engagement,

78
https://theshiftproject.org/en/article/unsustainable-use-online-video/
79
https://www.dentsu.com/attention-economy
80
https://go.integralads.com/rs/469-VBI-606/images/IAS_TheContextEffect_Final.pdf
81
https://www.invibes.com/insights/smart-scenarios-article.html
 95

etc.) according to research by Invibes81. But it’s not just about MFA sites. Even the
basic viewability metric that most advertis-
Now let’s translate these findings into the
ers track has an unspoken component re-
language of emissions.
lated to CO2 emissions – ads that were not
The aforementioned Ebiquity research viewable not only cost unnecessary money,
(Figure 6.7) indicates that a significant but also emitted unnecessary CO2.
proportion of ads (15.3%) were displayed
Thus, the mere fact that an advertiser cares
on Made for Advertising (MFA) sites, which
about maximizing the visibility of its ads
are unsuitable for advertisers and even emit
means that it is not only behaving economi-
more CO2 than quality sites.
cally, but also environmentally.

Frameworks
Even though net zero is a relatively new tising industry are working with greenhouse
idea, some frameworks and guides have gas reduction activities on its website85.
already been created to help you get there.
They also run a Media Carbon Calcula-
The French framework for calculating the
tor86 (available to IPA members only) to
carbon footprint of digital campaigns from
help planners calculate the greenhouse
SRI France is relatively detailed and very
gas emissions that planned campaigns will
useful82.
produce.
In 2020, the Advertising Association estab-
Note that the calculator only figures the
lished two working groups focused on the
emissions produced directly by the adver-
path to net zero83. Out of these, The IPA Cli-
tising itself. The next step the UK adver-
mate Charter was created, giving all charter
tising industry has taken is to calculate
members access to the tools and resources
the emissions that are produced by the
to make a real impact in their own agencies.
products that are produced in the advertis-
Ad Net Zero84 is an initiative of the British ing itself. ADGREEN87 was created for this
Advertising Association to coordinate, mon- purpose.
itor, and promote action across the industry
to achieve true net zero carbon emissions
from ad development, production, and me-
dia placement by the end of 2030.
The IPA has also presented information and
examples of how companies in the adver-

82
https://www.sri-france.org/wp-content/uploads/2021/11/SRI_Calculating-the-carbon-footprint_VF.pdf
83
https://adassoc.org.uk/resource/advertising-association-launches-two-climate-action-groups/
84
https://adnetzero.com/
85
https://ipamediaclimatecharter.co.uk/media-owner-showcase/media-owner-showcase/
86
https://ipamediaclimatecharter.co.uk/media-carbon-calculator/how-to-use-the-calculator/
87
https://weareadgreen.org/
 96

ADGREEN
The advertising industry is one of the The aim of these efforts is to calculate the
biggest polluters in the world. Ads reach emissions produced by promoted products
almost everyone on the planet through TV, not only at their creation, but also during
mobiles, computers, billboards, radios, and their lifetime and subsequent disposal. If we
other media and influence directly whether can do this, then we can start to consider
consumers buy a product with a higher or whether it is right to promote higher-emit-
lower impact on the environment. ting products and whether we should pro-
mote lower-emitting ones instead. We can
Gradually, an awareness is spreading that
calculate both options. The mere publication
we advertisers have an impact not only on
of this data puts pressure not only on the
the emissions that we produce, but also on
producers of the products but also on the
those that are indirectly produced by the
entire supply chain they go through, includ-
advertisements we participate in.
ing the advertising industry.
This idea is already making its way into
regulation. In Amsterdam88 and France89,
ads promoting fossil fuels are banned. The
biggest agencies, like WPP90 and others in
the UK, are proactive and are voluntarily
embarking on the path to carbon neutrality
themselves.

Sustainability and supply chain

The information above in relation to supply example, advertising in countries with
chains shows the following: coal-fired power plants is significantly
less environmentally friendly than in
● The length of the programmatic supply
countries with renewable or renewable
chain is directly proportional to energy
resource-based power plants.
consumption.
● In general, the shorter the supply chain,
– The more nodes a bid passes through,
the lower the CO2 emissions.
the more energy it consumes, and the
more CO2 emissions it produces. ● If you focus on showing only relevant ads
to relevant users, you will produce less
– CO2 emissions depend on the type of
emissions. The targeting process should
electricity used and how it is pro-
also include tools to eliminate fraud and
duced.
impressions with low visibility.
● The method of production of the
energy consumed as it passes through
the supply chain plays a key role. For

88
https://www.euronews.com/green/2021/05/20/amsterdam-becomes-first-city-in-the-world-to-ban-this-type-of-advert
89
https://www.euronews.com/green/2022/08/24/france-becomes-first-european-country-to-ban-fossil-fuel-ads-but-
does-the-new-law-go-far-e
90
https://www.wpp.com/sustainability/wpp-net-zero
 97

● Each impression shown can have

completely different CO2 production
values, and due to the number of
impressions, it is not possible to calculate
them easily, but it is necessary to do
your due diligence.
If there is one lesson to learn from the
above, it’s this: the shorter the supply
chain, the greater the financial savings, and
the less CO2 emissions.

Chapter conclusion
In this chapter we: If publishers and advertisers do not want
● discussed the main documents and to become victims of ad fraud, they must
standards in the field of greenhouse gas defend themselves. This chapter describes
emissions, exactly how they can do this and what
tools they have at their disposal. These
● showed the most important emission tools need to be set up well so that we can
factors in the advertising industry, then effectively optimize the entire supply
● discussed carbon credits and ways to chain system as we will show in the next
achieve carbon neutrality, chapter dealing with SPO and DPO.to get
a good understanding of the whole system,
● showed that effective advertising is we should understand the basic building
ecological. blocks on which it is based.
98

Conclusion
I hope that the information in this book has
helped you understand the entire ecosys-
tem of the programmatic supply chain.
To provide further reading on specific top-
ics, I have referred to and provided cita-
tions of additional materials and resources
that you may find useful.
The main message of this book is to make
your supply chain as optimal as possible.
By doing so, you will:
● save and/or make more money from
campaigns,
● reduce ad fraud,
● increase transparency for everyone
involved,
● reduce CO2 emissions.

I welcome any comments or feedback.

Please do not hesitate to contact me.

This book is available free of charge from

professional organizations, educational
projects, and universities. I welcome you
to register on my website www.Bauck-
mann.info if you would like to be contacted
about other publications and supplementa-
ry materials.

If you find the book useful, please feel

free to comment on Amazon or your social
networks. Not only will you be helping me,
you’ll also help other professionals in our
field gain access to useful material.
99
100

Used terms
Term Description

A programmatic marketplace where DSPs and SSPs communi-

Ad exchange
cate.
Ad fraud Abbreviation for online advertising fraud.
Ad Injection An ad fraud that uses false advertising.
A network of several publishers that acts as a single unified
Ad network
entity.
Ad server A system publishers use to easily manage ads.
Ads.cert The successor of the Ads.txt file within the RTB 3.0 protocol.
A simple text file that publishers use to declare who is autho-
Ads.txt
rized to sell their space.
AdWare Malicious software placed on the visitor’s computer.
Agency Trading Desk
A technology or set of services provided by a media agency.
(ATD)
A request for a new impression offer generated by the publish-
Bid request
er’s system and sent to bidders.
Bid response A response sent by a prospective advertiser to a bid request.
An algorithm that optimizes which bid requests to respond to
Bid throttling
on the purchasing side.
Blacklist A list of domains that we exclude from ad targeting.
Bot A program (robot) whose goal is to crawl a website.
The opposite of sellers.json, its essence is to uniquely identify
Buyers.json
the advertiser behind an offer.
Carbon credit / carbon A tradable unit that represents the amount of emissions re-
offset moved from the atmosphere.
Click Farms An ad fraud practice of generating fraudulent clicks.
CO2PM, gCO2ePM Grams of CO2 produced by 1,000 impressions.
Consent management
Platforms that manage users’ consent to the use of their data.
platform (CMP)
Cookie matching/ A process of how information in cookies is matched between
syncing different domains.
Cookie stuffing The practice of ad fraud using cookies planted on another site.
Short text files created by a website that allow data to be
Cookies
stored on the browser side.
Cost per mille (CPM) Price per ad impression per 1,000 impressions.
 101

Term Description

Data management
A tool designed to process data and create audiences.
platform (DMP)
Demand path optimi- A set of procedures on the side of the publisher to optimize the
zation (DPO) paths through the purchase chain to the advertiser.
Demand-side platform A tool to represent the buyer’s (advertiser’s) interests in pro-
(DSP) grammatic buying.
Part of the OpenRTB protocol where the supply chain for bid
DemandChain Object
response is stored.
Distributed deni-
A forceful attack on a site to overwhelm it.
al-of-service (DDoS)
Domain spoofing An ad fraud practice using a false domain.
First price model A method of auctioning advertising where the winner pays the
auction price they bid.
Floor price The minimum price a publisher asks for its advertising space.
General Data Pro-
Designation for the legal protection of Internet users in the
tection Regulation
European Union.
(GDPR)
General invalid traffic Site visit caused by a robot that does not mask itself in any
(GIVT) way.
A method of online auctions in a programmatic environment
Header bidding that ensures equal conditions and time for all auction partici-
pants.
Interactive Advertising An organization established to standardize, research, and pro-
Bureau (IAB) mote the online advertising industry.
Intermediaries Middlemen between publishers and advertisers.
Invalid traffic (IVT) Site visit caused by a bot.
Intergovernmental
An international body established to help the world understand
Panel on Climate
more about climate change.
Change (IPPC)
The time it takes to perform a particular operation, for exam-
Latency
ple, to load a page (page latency).
Malicious bots Ad fraud practice using SIVT.
An ad fraud practice that uses advertising to plant malicious
Malvertising
code on a visitor’s computer.
The goal of offsetting produced emissions to reach net neu-
Net zero emissions
trality.
102

Term Description

Open Bidding Header bidding solution from Google.

A consortium founded in 2010 to standardize Real Time Bid-
OpenRTB Consortium
ding.
OpenRTB protocol Standard for the transmission of bids in RTB.
A binding document under which countries have committed to
Paris Agreement
reducing greenhouse gas emissions.
Prebid The most widely used open source header bidding solution.
Private marketplace
Marketplace with offers limited to invited participants.
(PMP)
A commercial agreement between buyer and seller in an RTB
Programmatic deal
environment.
Programmatic buying An automated way of buying online advertising.
A form of programmatic buying based on the sell/buy method
Programmatic direct
for fixed, agreed volumes.
Programmatic guar- One form of programmatic buying based on the sales/pur-
anteed chase method for fixed, agreed volumes.
Real-time bidding One form of programmatic buying based on the auction sell/
(RTB) buy method.
Science Based Targets The first global science-based standard for setting net zero
initiative (SBTi) emission targets for businesses.
A method of auctioning advertising where the winner pays not
Second price model
the price they bid, but the price of the second highest bid plus
auction
one cent.
A file stored on the side of the demand partner used by pub-
Seller.json
lishers to declare who is authorized to sell their space.
Greenhouse gasses Gasses that cause global warming.
Sophisticated invalid
Site visit caused by a robot impersonating a real visitor.
traffic (SIVT)
A set of procedures, especially on the side of the advertiser, to
Supply path optimiza-
optimize the paths through the purchase chain to the publish-
tion (SPO)
er.
Supply-side platform A tool to represent the seller’s (publisher’s) interests in pro-
(SSP) grammatic buying.
Supply chain A series of intermediaries between buyer and seller.
 103

Term Description

Part of the OpenRTB protocol where the supply chain for a bid
SupplyChain Object
request is stored.
A condition where no response to a bid request arrives by the
Timeout
requested time.
Timeout rate Ratio of timeout responses to all bid request responses.
Transparent Ad Mar-
Amazon’s header bidding solution for larger publishers.
ketplace
Unified Ad Market-
Amazon’s S2S header bidding solution for smaller publishers.
place (UAM)
104

Sources
https://iabtechlab.com/wp-content/uploads/2022/04/
1 1
OpenRTB-2-6_FINAL.pdf

https://github.com/InteractiveAdvertisingBureau/open-
2 2
rtb/blob/master/OpenRTB%20v3.0%20FINAL.md

3 https://www.iab.com/guidelines/openrtb/ 3

4 https://iabtechlab.com/standards/openmedia/ 4

5 https://github.com/InteractiveAdvertisingBureau/AdCOM 5

https://github.com/InteractiveAdvertisingBureau/AdMan-
6 6
agementAPI

https://www.isba.org.uk/system/files?file=media/docu-
7 ments/2020-12/executive-summary-programmatic-sup- 7
ply-chain-transparency-study.pdf

https://www.isba.org.uk/knowledge/second-program-
8 8
matic-supply-chain-transparency-study

https://www.emarketer.com/content/us-programmat-
9 9
ic-fees-2019

https://www.emarketer.com/content/why-tech-firms-ob-
10 10
tain-most-of-the-money-in-programmatic-purchases

https://pubmatic.com/blog/ad-tech-tax-or-a-reality-of-
11 11
doing-business/

12 https://www.youtube.com/watch?v=3iNOf69me5Y 12
 105

https://twitter.com/braedon/sta-
13 13
tus/1567882264431988737

https://well-known.dev/resources/ads_txt/sites/marie-
14 14
claire.fr

https://www.adpushup.com/blog/header-bidding-adop-
15 15
tion-three-years-in-review/

16 https://adpushup.wistia.com/medias/swzqclb2xx 16

https://www.emarketer.com/chart/248339/which-supply-
17 side-platforms-ssps-have-us-publishers-used-past-12- 17
months-of-respondents-march-2021

https://www.insiderintelligence.com/content/advertisers-
18 18
cut-their-dsps-40-over-two-years

19 https://en.wikipedia.org/wiki/Ad_fraud 19

https://www.sciencedirect.com/science/article/abs/pii/
20 20
S1389128699000237

https://money.cnn.com/2004/12/02/technology/goo-
21 21
gle_fraud/

https://www.statista.com/statistics/677466/digital-ad-
22 22
fraud-cost/

https://www.forbes.com/sites/augustinefou/2021/01/02/
23 when-big-brands-stopped-spending-on-digital-ads-noth- 23
ing-happened-why/

24 https://en.wikipedia.org/wiki/Internet_bot 24
106

https://www.imperva.com/blog/bad-bot-report-2021-the-
25 25
pandemic-of-the-internet/

https://www.whiteops.com/hubfs/Resources/WO_Meth-
26 26
bot_Operation_WP.pdf

https://www.avast.com/c-what-is-malvertising-how-to-
27 27
stop-it

https://iabtechlab.com/software/iababc-internation-
28 28
al-spiders-and-bots-list/

https://www.pixalate.com/hubfs/Reports_and_Docu-
29 ments/Ad%20Fraud%20Reports/2022/Pixalate%20-%20 29
Q2%202022%20IVT%20Benchmark%20Report.pdf

30 https://www.oracle.com/cx/advertising/measurement/ 30

31 https://doubleverify.com/ 31

32 https://integralads.com/ 32

33 https://www.pixalate.com/ 33

https://www.wfanet.org/app/uploads/2017/04/WFA_
34 34
Compendium_Of_Ad_Fraud_Knowledge.pdf

https://iabtechlab.com/press-releases/
35 iab-tech-lab-launches-assault-on-illicit-advertising-inven- 35
tory/

https://www.emarketer.com/content/ads-txt-adoption-
36 36
continues-its-steady-growth
 107

37 https://iabtechlab.com/ads-txt/ 37

38 https://www.tagtoday.net/ 38

39 https://www.nytimes.com/ads.txt 39

https://iabtechlab.com/press-releases/iab-tech-lab-an-
40 nounces-two-new-technologies-to-build-more-transpar- 40
ency-trust-in-the-programmatic-supply-chain/

https://github.com/InteractiveAdvertisingBureau/open-
41 41
rtb/blob/master/supplychainobject.md

42 https://iabtechlab.com/sellers-json/ 42

43 https://iabtechlab.com/press-releases/openrtb-3-0-beta/ 43

https://iabtechlab.com/wp-content/uploads/2021/03/
44 44
buyers-json-1-0.pdf

https://iabtechlab.com/wp-content/uploads/2021/03/De-
45 45
mandChainObject-1-0.pdf

https://iabeurope.eu/wp-content/uploads/2022/10/
46 IAB-Europe-Attitudes-to-Programmatic-Advertising-Re- 46
port-2022-V3.pdf

https://bokonads.com/preventing-bad-behavior-in-head-
47 47
er-bidding-supply-path-optimization/

https://iabeurope.eu/knowledge-hub/iab-eu-
48 rope-guide-to-supply-path-optimisation-spo-updat- 48
ed-june-2022/
108

https://iabeurope.eu/wp-content/uploads/2022/06/
49 Guide-to-Supply-Path-Optimsation-SPO-2022-update- 49
FINAL.pdf

50 https://en.wikipedia.org/wiki/Queries_per_second 50

http://www0.cs.ucl.ac.uk/staff/w.zhang/rtb-papers/
51 51
turn-throatling.pdf

https://unfccc.int/process-and-meetings/the-par-
52 52
is-agreement/the-paris-agreement

53 https://press.un.org/en/2020/sgsm20411.doc.htm 53

54 https://www.ipcc.ch/sr15/ 54

55 https://www.ipcc.ch/report/ar6/wg3/ 55

56 https://sciencebasedtargets.org/ 56

57 https://sciencebasedtargets.org/net-zero 57

58 https://ghgprotocol.org/ 58

59 https://carboncredits.com/sec-carbon-emissions/ 59

https://emsmastery.com/2022/07/15/what-is-1-ton-of-
60 60
carbon-dioxide-and-how-it-relates-to-daily-life
 109

61 https://www.carbonfootprint.com/calculator.aspx 61

https://www.climateneutralgroup.com/en/news/what-ex-
62 62
actly-is-1-tonne-of-co2/

63 https://www.sylvera.com/ 63

64 https://carboncredits.com/carbon-prices-today/ 64

https://marketplace.goldstandard.org/collections/proj-
65 65
ects

66 https://verra.org/ 66

67 https://xpansiv.com/ 67

https://carboncredits.com/the-4-best-carbon-offset-pro-
68 68
grams-for-2022/

https://carboncredits.com/uk-carbon-removal-rechnolo-
69 69
gies/

70 https://carboncredits.com/the-greenpeace-revolt/ 70

71 https://app.electricitymaps.com/zone/PL 71

https://www.goclimate.com/blog/the-carbon-footprint-
72 72
of-servers/
110

https://sustainability.aboutamazon.com/environment/
73 73
the-cloud?energyType=true

74 https://sustainability.google/progress/energy/ 74

https://marmelab.com/blog/2022/01/17/media-web-
75 75
sites-carbon-emissions.html

https://www.adexchanger.com/the-sell-sider/advertis-
76 76
ers-have-an-environmental-responsibility/

https://www.ebiquity.com/news-insights/viewpoints/
77 77
the-hidden-cost-of-digital-advertising/

https://theshiftproject.org/en/article/unsustain-
78 78
able-use-online-video/

79 https://www.dentsu.com/attention-economy 79

https://go.integralads.com/rs/469-VBI-606/images/IAS_
80 80
TheContextEffect_Final.pdf

https://www.sri-france.org/wp-content/uploads/2021/11/
81 81
SRI_Calculating-the-carbon-footprint_VF.pdf

https://adassoc.org.uk/resource/advertising-associa-
82 82
tion-launches-two-climate-action-groups/

83 https://adnetzero.com/ 83

https://ipamediaclimatecharter.co.uk/media-own-
84 84
er-showcase/media-owner-showcase/
 111

https://ipamediaclimatecharter.co.uk/media-carbon-cal-
85 85
culator/how-to-use-the-calculator/

86 https://weareadgreen.org/ 86

https://www.sri-france.org/wp-content/uploads/2021/11/
87 87
SRI_Calculating-the-carbon-footprint_VF.pdf

https://www.euronews.com/green/2021/05/20/amster-
88 dam-becomes-first-city-in-the-world-to-ban-this-type-of- 88
advert

https://www.euronews.com/green/2022/08/24/france-
89 becomes-first-european-country-to-ban-fossil-fuel-ads- 89
but-does-the-new-law-go-far-e

90 https://www.wpp.com/sustainability/wpp-net-zero 90
112
David follows up with a great book on header bidding that every yield
manager, ad server admin, and developer should read. He very clearly
describes the technology itself, possible solutions, and their applica-
tion to everyday practice. In this book you get a great overview of not
only how the whole supply chain works, but also how to avoid ad fraud.
Petr Mareš (Invibes)

David provided a truly comprehensive introduction to the AdTech

industry. It's not necessary to read multiple blogs anymore. David has a
unique ability to explain all aspects of AdTech with context. It's the
ultimate resource for anyone in an AdOps role.
Jan Sunavec (FILMZIE)

I recommend this book not only to specialists on the publishers' side

but also to every programmatic specialist. In order to get the most out
of programmatic campaigns, it is important to understand how the
system works in detail and, unfortunately, the systems that abuse and
parasitize this ecosystem. I am very glad that the author is one of the
first in our region to address the topic of sustainability and carbon
neutrality in programmatic.
Matěj Kolarovsky (IAB Slovakia)

The programmatic ecosystem is getting more complex every year. The

journey of an ad impression from a publisher's website to an
advertiser's buying system can be long and complex. At the same time,
it's important for both parties to understand the process and be able
to work with it and possibly optimize it. That's where this book can help
– it provides a sufficiently detailed yet easy-to-understand description
of the relationships in a supply chain, along with concrete examples
and tips that both publishers and advertisers can use.
Lukáš Šmol (CPEx)