A Short Tutorial on Guesstimation

Much of the material in this tutorial are extracted from Wienstein L. and J.
Adam, "Guesstimation," Princeton University Press, 2008.
Guesstimation as one will notice immediately is made up of two root
words: namely guess and estimation. Wikipedia suggests that a guesstimate
may be a first rough approximation pending a more accurate estimate. Proponents of guesstimation do not claim accuracy, but a quick estimate that
is an order of magnitude from the answer. In engineering, more often than
not, an estimate that is an order of magnitude away from the actual value is
adequately reasonable.
In this short tutorial, we do not plan to make you an expert of guesstimation. But rather, we will like to introduce to you guesstimation. In the course
of this module, we will use guesstimation often to arrive at ballpark values.
In engineering, an order of magnitude, is referred to within a factor of 10.
For example, estimating 100 for a value of 1000 is within a factor of 10 and
therefore also within an order of magnitude. The first step in guesstimation
is to make plausible guesses for the value. We shall label them as lower and
upper bounds. The lower bound is the minimum guess of the value, beyond
which the guess will seem unreasonable. Likewise, the upper bound is the
maximum guess of the value, and similarly, if one guesses beyond this value,
the guess may seem absurd. The important point to note here is that we are
not making any accurate estimates. Guesstimation is about making informed
guesses from very general information. You will be quite surprised at how
close the guesstimate is from the actual value.
Before we look at some examples, it will be useful to appreciate a few
simple techniques:
1. Once we have established the lower and upper bounds of an estimate, we
will use the mean of these values for the estimate. If we choose to use the
arithmetic mean for this purpose, it will provide an answer that is not the
same factor above and below the bounds. For example, suppose our lower
bound is 1 and our upper bound is 100. An arithmetic mean will require
us to add the bounds and divide it by two to arrive at the mean. This will
result in about 50. The arithmetic mean is 50 times larger than the lower
bound but only two times smaller than the upper bound. This is not a good
practice. The alternative is to use the geometric mean. To compute the

ENGINEERING PRINCIPLES

& P R AC T I C E I

geometric mean, we take the product of the bounds and apply the square
root to the product. For the same example as above, the geometric mean
will be the square root of 100 (1 100) which is 10. Note that 10 is 10
times larger than the lower bound and also 10 times smaller than the upper
bound. It lies in the sweet spot in terms of factors of the bounds.
2. To facilitate the calculation of the geometric mean, it is advisable to
express the bounds in either scientific or engineering notations. In this
tutorial we will use scientific notation exclusively, but the discussion will
also apply to engineering notations.
3. Suppose the lower bound is xl 10yl and the upper bound is xu 10yu .
xl and xu are referred to as the coefficients in the scientific notation,
while yl and yu are referred to as the exponents. The geometric mean is
calculated by finding the arithmetic means of the coefficients and exponents. For example, is the lower bound is 1 100 and the upper bound is
0+2
1 102 , then the geometric mean is calculated as 1+2 1 10 2 or 1 101 .
This will always work for as long as the sum of the exponents is even.
If the sum of the exponents are odd, we will need to subtract the sum
of exponents by 1 and multiply the result by 3. For example, suppose
the lower bound is 1 100 and the upper bound is 1 103 , then the
0+31
geometric mean is calculated as 3 1+2 1 10 2 or 3 101 . Note that
the square root of 1000 is about 31.62 which is close to the 30 that we
have approximated. Remember, we are not trying to be accurate - we are
trying to arrive at a reasonably approximate value.
4. The advantage of using scientific or engineering notations for very large
and and very small numbers will help in reducing errors. For example,
suppose we write 10000000 as 1 107 , the scientific notation is nifty and
requires us to count the number of zeros only once for the conversion.
Imagine using 10000000 for mathematical operations, the chances of miscounting the number of zeros is quite great. The same argument could be
applied to a very small number like 0.0000000001. Counting the number
of decimal places after each mathematical operation is very likely to lead
to errors.
5. In the scientific notation, when we express a number as x 10y , it is important to note that the exponent, y, is more important than the coefficient
x. The reason for this statement is that the error in x is not as significant as
the error in y. For example for the number 2 108 , if 2 is changed to 3,
this will lead to a change of the result by a factor of 1.5 times, while if 8
were changed to 9, the result will change by a factor of 10 times.
6. To mulitpy two very large numbers of two very small numbers using
the scientific notation, we simply multiply the coefficients and add the
exponents. For example, (2.5 105 ) (4 106 ) will yield (2.5 4)

A S H O RT T U T O R I A L O N G U E S S T I M AT I O N

10(5+6) or 1 1012 . Likewise, for division, we divide the coefficients and

subtract the exponents. For example, (4 105 ) (2 106 ) will yield
(4 2) 10(56) or 2 101 .
7. Addition and subtraction of very large or very small numbers can only be
performed if the exponents are of the same value. Therefore, we express
both numbers until their exponents are of the same value. For example,
(3 107 ) + (4 108 ) may be expressed as (0.3 108 ) + (4 108 )
which then yields 4.3 108 .
8. Since we are estimating values to a factor of 10, we only need to keep
the coefficient to a single digit. There are two reasons for this practice.
First, since we are estimating values, these are not accurately known and
carrying precision in the numerals is not necessary. Second, when dealing
with single digits, it is easier to perform mathematical operations, such as
multiplication, division, addition or subtraction.
With preliminary considerations discussed above, we can now discuss examples of guesstimation. In this tutorial we will consider two examples.

Example 1
What is the surface area of a typical bath towel? Include the surface area
of the fibers of the towel in your estimation. We will solve this problem by
following the following steps:
1. Estimate the dimensions of the towel.
2. Calculate the physical area of the towel. Remember to include the area of
both sides of the towel.
3. Estimate the number of fibers per square cm of the towel.
4. Estimate the dimensions of a typical fiber.
5. Establish a reasonable model for the fiber and calculate the surface area of
a typical fiber.
6. Multiply surface area of a fiber by the number of fibers per square cm and
then multiply this value by the area of the towel in square cm.
We shall assume that a typical bath towel is 1 2m in dimension. Therefore, the surface area of the towel without considering the fibers of the towel
is (1 2) 2 sides which is equal to 4m2 .
How many fibers are there in a square cm? The accurate answer is not
possible unless we manually mark out squares of 1cm2 and count the number
of fibers in these squares and establish an average value. We are not going
to do that. We will perform a guesstimation instead. We can safely say that
the low limit of the number of fibers in a square cm is 10 and the upper limit

ENGINEERING PRINCIPLES

& P R AC T I C E I

is 1000. Therefore the geometrical mean is 100 fibers per square cm. Note 1
square cm is equivalent to 0.01 0.01m2 or 1 104 m2
A towel at my home had fibers that were about 5mm long and 1mm wide.
You may assume that a fiber can be modeled as a rectangular box of sides
1mm each and height of 5mm. We may ignore the surface area of the top of
the box as being small compared to the surface area of the sides of the box.
The bottom of the box is flush with the towel and should not be included in
our calculations. Since the box has 4 sides, we estimate the surface area of a
fiber as 4 (5 1)mm2 or 20mm2 which is equivalent to 20 106 m2 or
2 105 m2 .
Therefore surface area of towel, A is given by,
A = area of towel fibers per area surface area of a fiber
2 105 m2
100fibers

1fiber
1 104 m2

5
800 10
m2
=
1 104
= 80m2

= 4m2

This is 20 times the area of the towel without the fibers. Please also pay
attention to the dimensional analysis of the previous equation. You will
notice that the dimensions are consistent.

Example 2
How far does a soccer player travel during the course of a game?
We will solve this problem by following the following steps:
1. Make some assumptions.
2. Estimate the average speed of a player in the game.
3. Distance = speed time.
We make the following assumptions:
A soccer game lasts 90 minutes.
Players rarely stay still. We will estimate the distance traveled for a typical
player. We will exclude the goalkeeper from the estimations.
Players run and walk during the game.
A typical player runs for part of the game and walks for the rest. We shall
assume that the player never keeps still. Think of the running speed as the
upper bound on the speed of the player and the lower bound as the walking
speed. To estimate the running speed, let us use a middle distance runners
speed. A middle distance runner is able to complete 1500m in about4 mins.

A S H O RT T U T O R I A L O N G U E S S T I M AT I O N

Lets increase the distance slightly to 1600m in 4mins. This will yield a speed
of 400m per min or roughly 7m/s. (Remember we need to estimate the speed
to just one digit in the scientific notation.)
The walking speed must certainly be lower than the running speed. Let
us assume that the walking speed is about 2m/s. (Note you can use whatever walking speed that you are comfortable with. We will not arrive at an
accurate answer - all we want is a reasonable estimate.)
Therefore the average speed that a player assumes in a soccer game is the
geometric mean of 7 and 2 m/s, which works out to be about 4m/s. Note that
for this problem, the arithmetic mean is also about 4m/s.
Therefore for a game of 90 mins, a typical player will have travelled, D
D = average speed time

= 4m/s 90min
= 4m/s (90 60)secs
= 21600m

= 22km.

Please apply dimensional analysis to prove to yourself that the units in the
estimation are valid.

Example 3: Problem extracted from "Guestimation"

How much energy is needed to get a spaceship from Earth to Alpha
Centauri, which is about 4 light years away. The trip must be made
before the passengers die of old age. We are also required to estimate the
fuel that is required. Assume 1 ton of rocket fuel provides 4 109 joules.
We will solve this problem by following the following hints:
1. How large should the spaceship be?
2. An aircraft carrier has mass of 1 105 tons.
3. The speed of light is c = 3 108 m/s. A light year is distance that light
travels in a year. A year is about 3 107 secs.
4. What is the bound on the travel time before the passengers die of old age?
5. What will be the required speed of the spaceship?
6. What is the kinetic energy of the spaceship at that velocity?
7. Estimate the fuel required?
The kinetic energy of the spaceship will be provided by the rocket fuel.
To calculate the kinetic energy we will require the mass of the spaceship
and its velocity. We first calculate the speed required to make the trip with
the passengers not dying of old age. Let us assume that 40 years will be a
reasonable estimate considering the lifespan of a typical human.

ENGINEERING PRINCIPLES

& P R AC T I C E I

If light takes 4 years to travel to Alpha Centauri, we will need to travel at

the speed of light if we want to reach there in 40 years. Therefore the
speed of the spaceship is
1
10 th

1
c
10
= 0.1 3 108

v=

= 3 107 m/s

We will require a spaceship of a reasonable size to carry the supplies

necessary for a 40 year trip. Therefore let us assume that our spaceship will
be the size of 10% of an aircraft carrier. Note the larger the mass, the larger
is the kinetic energy and therefore more fuel is required. So lets start with the
assumption that the mass of the spaceship is 10% that of an aircraft carrier,
i.e., m = 1 104 tons, or 1 107 kg. (Note 1ton = 1000kg.)
Therefore the Kinetic Energy of the spaceship at this mass and velocity is
1 2
mv
2

2
1
= 1 107 3 107
2
= 5 1021 J

KE =

Verify that the units of Joules (J) are consistent with the units of the quantities
that we have used.
Next, we will estimate the fuel required, FR
FR =

KE
tons
Enegy from one ton of rocket fuel

5 1021
tons
4 109
= 1 1012 tons

The fuel is more than the mass of the spaceship by an order of 8 (108 ). With
the power provided by the rocket fuel, we will not be able to reach our destination in the required time. Note that in our estimation we have not accounted for the losses in the rocket engines and the energy required for accelerating and decelerating the spaceship. Space travel will require a big leap in
development before travel to the stars is a reality.
Finally, we terminate this tutorial with four tables that offer you a glimpse
about the variation of quantities in terms of magnitudes. It provides you
with a framework to make references to various sizes and dimensions. These
tables are summarized in Figures ?? to ??.

A S H O RT T U T O R I A L O N G U E S S T I M AT I O N

Figure 1: Length Order of Magnitudes

ENGINEERING PRINCIPLES

& P R AC T I C E I

Figure 2: Area Order of Magnitudes

Figure 3: Density Order of Magnitudes

A S H O RT T U T O R I A L O N G U E S S T I M AT I O N

Figure 4: Mass Order of Magnitudes