1

Business Analytics in Real-World Case Studies on Inventory Management and

Forecasting.

Student Name: SUN CHENXU

Student Number: 2022123074

Wednesday, December 25, 2024

 2

Contents
Introduction.................................................................................................................................................2
Walmart Inventory Management.................................................................................................................2
Predictive Analytics.................................................................................................................................3
Machine Learning....................................................................................................................................3
Forecasting in Amazon................................................................................................................................4
Predictive Analytics.................................................................................................................................4
Simulation...............................................................................................................................................5
Zara Inventory Management and Forecasting..............................................................................................6
Predictive Analytics.................................................................................................................................7
Procter & Gamble (P&G) Forecasting.........................................................................................................7
Predictive Analytics.................................................................................................................................8
Machine Learning....................................................................................................................................9
Starbucks Inventory Management and Forecasting.....................................................................................9
Predictive Analytics.................................................................................................................................9
Impact of Business Analytic Techniques...................................................................................................10
 3

Introduction
Globalization and technologies entail more analysis of data and information and make the usage
of analytics more necessary in current organizations. These specific findings specifically review
the role of business analytics related to product inventory for the world’s giant organizations
including Walmart, Amazon, Zara, P&G, Starbucks, and many others. Through the use of
predictive analytics, machine learning, prescriptive analytics, and simulation, these organizations
determine the appropriate stock levels for their inventory get the right demand forecasts and thus
make their undertaking more effective. Different cases and examples attempted in this paper
portray the field of business analytics as an essential source of change management for
organizations, and how the subject is central to maintaining competitive advantage as well as a
consistent theme of continuous transformation for different industries.

Walmart Inventory Management.

Walmart Inc. was founded by Sam Walton in the year 1962 and is a multinational retail company
located in Bentonville, Arkansas. Having a very long chain of hypermarkets, discount
departmental stores, and grocery stores, it is quite famous. Walmart is the largest retail store in
the United States offering affordable prices for virtually everything it sells. Currently, it has tens
of thousands of outlets and tens of millions of workers, millions of its permanent clientele is
measured in hundreds of millions of people. Walmart’s business strategy is centered on key ideas
such as transaction-level selling, large store coverage, and sophisticated methods of supply chain
management that employ strategies such as big data analysis and artificial intelligence.

Predictive Analytics
There are several methods that Walmart uses involving forecasting the store’s requirements and
these are best summed up by the term stock proximity and relevance. This includes quantitative
models such as time series and regression analysis. The time series models assist the
organization in establishing the pattern of sales data across the different seasons The regression
analysis tool used in the research assists the organization in making a forecast of sales under
some conditions given a relationship between the two variables.

Additionally, Walmart targeted predictive analytics solutions as well as the solution that is
integrated with a Walmart Luminate data platform to work with data (Joji, 2024). One of how it
can function is as a system that provides knowledge about the actions of shoppers and the trends
 4

in the market, thus, helping a brand understand what consumers might require in the future. It
involves the use of artificial intelligence and analytical prediction to transform the acquired data
into useful student performances. For instance, Walmart Luminate can be integrated with other
data management systems and the predictions are capable of having all the data that make a
proper prediction of the demand for a certain product.

Figure 1: Data sources for Walmart's predictive analytics model. The figure shows various data
sources (historical sales data, web search terms, macro weather conditions, etc.) and their
integration with the Walmart predictive analytics model.

Walmart also uses artificial intelligence based on supply chain management for efficient
automatic placement and positioning of stock atm its distribution and fulfillment centers and its
stores (Decking the aisles with data, 2024). Such predictive models are backed up by historical
sales, web-based search terms, and page visits. In addition, they incorporate ‘future data’, such as
macro-weather conditions, macro-economic data, and local demographic trends, to estimate
demand and likely fulfillment more accurately.

Walmart has created a strategy that focuses on detailed information obtained from sources
ranging from the stores to the World Wide Web and their patrons’ feedback. The type of ML
algorithms that Walmart employs are designed to review such figures with the view of
identifying inherent patterns and trends that are easily decipherable by humans. For example,
these algorithms may uncover things such as, during holiday periods some products are bought
more than others or particular promotions are known to make employees quit the company in
large numbers (Krogerus & Tschäppeler 2020).
 5

The ML algorithms can help Walmart in another way of forecasting future trends in sales by
improving the stock inventory. It forms a crucial part of Artificial intelligence as it can make
such predictions as the number of customers expected in the weather and items such as when it is
raining more ice creams are sold worldwide. This ability allows Walmart to adapt to the plans
needed meaning they are well prepared when it comes to any unpredictability that may be
expected. From these predictive statements, a far-reaching outcome has been identified which is
the ability to determine the stock level per store. The use of the ML algorithms provides
recommendations on the required stock levels that eliminate situations where there is
overstocking and thus wastage of food products, and cases of stock outs which deprive the
business of potential sales (Harbert, 2019). Such optimization also contributes to the
effectiveness of Walmart’s inventory by guaranteeing the availability of products to enable
shoppers to have their desired overall shopping experiences, while, at the same time, contributing
to a reduction of holding costs that would be irrelevant, thus making Walmart more cost-
efficient.

Machine learning provides the features of adaptive learning and also an incremental
improvement to Walmart data. One of the key characteristics of the ML systems implemented at
Walmart has to do with the adaptability of learning. Whenever new data is incorporated into the
system, the ML models that underpin the system are recalibrated and hence the accuracy
increases with time. Such a continuing process of learning and improvement is proof of an
inherent characteristic of ML and the need for such an approach in addressing the intricate task
of inventory control in one of the largest retail outlets in the world, Walmart.

Forecasting in Amazon
Amazon, which operates as the largest e-commerce and cloud services company in the world
founded by Jeff Bezos in 1994. It was founded in Seattle, Washington as an internet media seller
and expanded extremely fast, thus it is often called “The Everything Store”; today, Millions of
people all over the world use various services provided by Amazon, including but not limited to
Amazon Prime, AWS, and Alexa.
 6

Predictive Analytics
It is critical to Amazon's success and demands forecasting functions. Demand forecasting is a
sophisticated process that requires assessing consumer data and feedback in addition to being a
fantastic mathematical formula that can be filled in with numbers and values from sales history
and other market indicators. In this manner, Amazon can use this data to make suggestions about
how much inventory to keep on hand and how to buy the products so that clients don't receive
more than they need (Ghaffari, 2021).

The predictive analytics of Amazon are mostly based on Machine Learning and Cloud
Computing. An ML model, as the name suggests, is an algorithm that learns from patterns
present in a large dataset. AWS fulfills the power requirement of these algorithms and allows the
execution of these models. The ML algorithms are capable of analyzing the purchasing data,
browsing history, and other factors related to the customers that enable recommendations of the
right products to the consumers as well as the prediction of the market demand with high

precision. The models used by Amazon in their forecasting methods should be able to integrate
changes in the market and adjust for the same. For example, during the COVID-19 pandemic, it
was observed that the company’s AI-driven predictive forecasting instantly adapted to the
sudden increase in demand for products such as toilet paper and the system does not struggle
with the operationalization of new demands (Amazon Web Services, 2021).

Figure 2: This bar chart represents the various components of Amazon's machine learning and
cloud computing infrastructure, highlighting the percentage of usage for each component.

Amazon has moved from a mechanical to a cartographic method of forecasting. Known

suggestions of this approach include that it enables Amazon to think through different possible
futures and their likelihoods. In this way, Amazon can get ready for all the possible scenarios and
 7

thereby increase the stability of the supply chain. An integration of deep learning, static image
recognition, and natural language processing has allowed a single forecast model to be created at
Amazon, which is capable of efficiently predicting the demand for various kinds of products.
This model explains the high volatility of customer demand in the future, when shopping on
Amazon, customers’ search results are as desired (Torkkola & Madeka, 2021). The organization
that deals with supply chain optimization technologies (SCOT) at AMAZON has handled most
of the forecasting automation to be a set of fully automated wheels. This automation also
includes distribution of inventories in fulfillment centers which is essential in the provision of the
promise of fast shipment by Amazon.

Simulation
Simulations on the other hand are advanced and complex algorithms that mimic the actual events
to come up with an outcome of a given scenario. Amazon uses these models to build working
avatars of both the actual supply chain and the complicated, global network they both operate
within and attempt to understand. In this way, it is possible to compare the efficiency of various
strategies using mere simulations, and the realistic fundamentals of organizational processes will
not be compromised.

It is possible to state that the simulation models that Amazon possesses are focused on the
prediction of the demand for millions of products in different countries. The context involves the
identification of future demand for the targeted product by the company’s forecasting team and
feeding that information to simulation models with assumptions on the variables that may affect
the inventory and the supply systems of the company. Another major strength of simulation
models is versatility. The models applied by Amazon to fulfill the consumers’ demand are
constructed to be flexible when in the market and this was proven by their response to the
COVID-19 pandemic where some products such as toilet paper had a higher demand than
expected. The simulation models enabled Amazon to react quickly to these previously
unrecognized demand signals, which in turn enhanced the firm’s ability to respond to shifts in
the external environment.

Flexible improvement supports another important requirement that can also be seen in Amazon’s
example – the evolution of algorithms used for forecasting. In some cases, the simulation models
of the company have been strengthened through the development of deep learning, image
 8

recognition, as well as natural language processing; these tools have evolved into a unified
forecasting model. This model can produce precise decisions on several products so that the
shoppers are able to find exactly what they are looking for when shopping on Amazon (Torkkola
& Madeka, 2021).

Zara Inventory Management and Forecasting

Zara, being a leading clothing brand of the Spanish Inditex group, is defined by their policy of
‘fast fashion,’ selling clothes that are trendy but cheap. Spain-based Zara is one of the biggest
success stories in the fashion industry It was established in 1975 by Amancio Ortega and Rosalía
Mera. Currently, Zara possesses more than 2000 shops all over the world and targets people of
different ages, always responding to current fashion tendencies while preserving the company’s
environmental responsibility. It evolved as a perfect intersection of fashionable apparel produced
by integrating modern design elements, well-maintained supply chain management, and
consumer appeal.

Predictive Analytics
Zara’s key approach to PPS relies on analyzing sales data, social media information, and the
outputs of fashion show data in real time. For example, using Jet lore’s AI algorithms, Zara
increased inventory turns by 30%, reducing excess inventory and greatly improving supply chain
efficiency. Thus, using data intelligence, Zara receives immediate information from its stores
regarding the favorite styles, sizes, and regions’ preferences. The former is imperative for Zara to
forecast the customers’ needs and ensure adequate stock control across its shops (Hickey, 2018).
Currently, Zara has employed the service of Jetlore, which is an AI-backed consumer behavior
prediction firm, that uses algorithms to determine clients’ buying patterns. Jetlore’s platform
helps Zara to determine which particular items will be popular, which helps the company to
correlate its stock.

Apart from Jet lore, Zara has partnered with another Spanish big data Solutions provider El Arte
de Median. This improves the capacity for the organization to collect and analyze large volumes
of information that helps Zara advance its predictive models for stock control. Zara uses RFIDs–
Radio Frequency Identification technology incorporated in garment tags and used to help track
stock inventory and/or reorder. This technology organizes the functioning of fitting rooms and
 9

gives real-time statistics on stock that are crucial for analysis in stock planning. Zara uses POS
systems for billing and transactional purposes to extract current data regarding the transactions
made by customers. Such systems enable the sales, returns, and consumers’ behaviors to be
monitored and later used to determine future inventory needs.

Procter & Gamble (P&G) Forecasting

Procter & Gamble Company is a large transnational corporation from America, which operates
in the field of consumer goods and is based in Cincinnati, Ohio. This multinational corporation
that has operations in 70 countries around the world was established in 1837 by William Procter
and James Gamble. It deals in personal health, personal care, and hygiene goods. Such products
are categorized under beauty and grooming, health and personal care, fabrics and home care, and
baby, feminine as well as family products. The company focuses on the superiority of product,
package, brand communication, retailing, and added value (Hunt, 2021). In regard to this
strategy, P&G has embraced E2open’s next-generation demand planning solution that applies the
use of machine learning algorithms to enhance the demand planning processes.

Predictive Analytics.
Predictive analytics is used in various ways to involve analysis of different kinds of data to
determine the significant features for Sales forecasts which include: The firm has created an in-
house conversational AI platform, chatPG, using OpenAI’s interface, which is utilized in more

than 35 cases, such as demand forecasting. There is a significant focus on the intellectual
property protection of this tool and additional internal data feeds to improve the tool’s accuracy.
There is a particular model that P&G employs for carrying out Predictive Analytics in supply
chain costing: the KNIME platform. KNIME is an open-source data analytics reporting and
 10

integration tool that P&G utilizes to generate supply and demand forecasts. In the initial stages of
the pandemic, KNIME helped P&G integrate and perform analyses on large supply data thereby
cutting down on the time taken to respond to inquiries and at the same time offering cost savings
(DeNittis, 2024).

Figure 3: This bar chart depicts the various predictive analytics tools and platforms used by
Procter & Gamble and shows how often each tool is used.

Another innovation adopted by P&G is the data visualization models, which are predictive and
provided in the form of Operational Dashboards, charts, and/or supplier analytics. These models
provide what-if analysis of the overall plans and strategies since they imply the general range of
outcomes and they are always fine-tuned at the end of every quarter. This allows P&G to better
see and understand large data sets and make decisions about inventory and procurement hence
fulfilling the first objective of the strategy. In addition, analyzing the results, P&G has worked
out several spin-offs connected with the KNIME platform.

Among them are the Raw and Packing Materials Inventory Model, XIM Symbolizing the
Extended Inventory Model that is designed to model more complex distribution networks, and
RIM, which stands for Retailer Inventory Model, which models inventory up to the level of store
shelves. These models are used specifically for those segments that are applicable in P & G’s
various product ranges as per its requirements. Along with the above-stated models, P&G also
uses visual analytics from Spotfire to place visual presentations of information named ‘Decision
Cockpit’ on the computer interfaces of more than fifty thousand employees. Combined, all these
have greatly boosted decision-making as it has been empowered with easily understandable and
interpretable data.

Machine Learning.
Currently, P&G utilizes a set of machine learning algorithms to increase the accuracy of the
corresponding demand forecasting (Nowadly & Jung, 2020). These are specific algorithms
capable of learning from new numbers that come in and can tweak the forecast that has been
made in such a way that P&G is able to open up to the changes in the market demand. In this
work, Random Forest (RF) which is an ensemble learning method is used for classification and
regression. It operates by building up numerous decision trees throughout the training phase and
 11

providing the most popular one in the classes in the case of classification or the average in the
case of regression.

ANNs are systems of computations based on the biological neural networks that form animal
brains (Loukili et al., 2024). ANNs are then employed by P&G to extract high-level information
hidden within the data that cannot be easily observed using simple statistical analysis Linear is a
fundamental statistical method, whereby P&G forecasts a response variable given one or more
explanatory variables. This technique helps the company to anticipate the demand for the
products they sell and other predictor variables. Support Vector Machine is a machine learning
model intended for classification as well as regression analysis of the data. SVM is used at P&G
to categorize the products to different demands and to estimate the amount of the product that
will be sold.

Starbucks Inventory Management and Forecasting

Starbucks is a global leader in the coffeehouse business with its origin in Seattle, United States
of America in 1971. It gained worldwide acknowledgment after Howard Schultz purchased it in
and grew into a chain of more than 35000 shops. Starbucks specializes in coffee drinks and also
sells tea and pastries; the cafes are modern and comfortable. It is focused on ethical sourcing and
the connection with the community while striving to offer constant excellent quality coffee all
over the globe.

Predictive Analytics
Starbucks uses a very advanced form of predictive analytical approach in the management of
supply and inventory. Digital Flywheel Program is the recommended field of supplying the exact
Food and Drink items, including suggesting new products based on the customer’s purchase
pattern and creating a personalized experience based on factors such as weather and customer
inclination. In addition, Deep Brewing plans to apply AI technology to real-life applications,
such as predicting peak staffing needs and the probability of equipment failure, thus indirectly
improving operational efficiency (Dignan, 2017). It also relates data on weather and customers’
tendencies to order thus creating personalized or even targeted experiences and offers. Deep
Brew Initiative is one of the most important solutions that Starbucks has mastered and used to
implement AI architectures in the company. Deep Brew enhances the process of division and
 12

identification of the target market with a level of accuracy that was unthinkable earlier; it tackles
practical issues such as how many people are required to work during certain periods and when
different equipment is likely to fail.

Impact of Business Analytic Techniques

The latter business functions have transformed due to the application of sophisticated methods of
business analytics in merchandising, inventory management, and forecasting for Walmart,
Amazon, Zara, Procter & Gamble, and Starbucks. Through the use of predictive analytics, these
companies can assess the demand of the customer accurately thereby avoiding incidences of
stock out. Machine learning technology further optimizes inventory management by analyzing a
complex set of data on customer preferences and the relationship between buying trends, seasons
and promotions. For example, Walmart's machine learning system continuously improves
probability estimates to develop the best daily replenishment programmer in the world. Going
the extra mile improves the satisfaction and loyalty of the customers because the products are
well-stocked to meet the market demands. Other machine learning techniques enhance stock
management since they can read complicated data, such as purchasing trends, seasons, and the
effects of promotions. For example, Walmart’s machine learning systems perpetually improve

the accuracy of their probability estimates, making daily re-stocking at stores globally optimally
possible. For instance, Amazon employs prescriptive analytics and simulation to check different
estimates on the effect of forecasts and determine the most appropriate methods to use when
ordering stocks hence cutting on costs related to overstocked or shortage of stocks.

Figure 4: This bar chart shows the extent to which different business analytics techniques
(predictive analytics, machine learning, prescriptive analytics, simulation) have an impact on
 13

inventory management, illustrating the effectiveness of each technique in improving inventory

levels, customer satisfaction, cost savings, and forecast accuracy.

References
Amazon Web Services. (2021, December 3). Predicting the future of demand: How Amazon is
reinventing forecasting with machine learning. Forbes.
https://www.forbes.com/sites/amazonwebservices/2021/12/03/predicting-the-future-of-
demand-how-amazon-is-reinventing-forecasting-with-machine-learning/
Decking the aisles with data: How Walmart’s AI-powered inventory system brightens the
holidays. (2024). Decking the Aisles with Data: How Walmart’s AI-Powered Inventory
System Brightens the Holidays.
https://tech.walmart.com/content/walmart-global-tech/en_us/blog/post/walmarts-ai-
powered-inventory-system-brightens-the-holidays.html
DeNittis, N. (2024, February 5). Artificial intelligence at Procter & Gamble. Emerj Artificial
Intelligence Research; Emerj. https://emerj.com/ai-sector-overviews/artificial-
intelligence-at-procter-gamble/
Dignan, L. (2017, July 28). Starbucks to step up rollout of “digital flywheel” strategy. ZDNET.
https://www.zdnet.com/article/starbucks-to-step-up-rollout-of-digital-flywheel-strategy/
Ghaffari, R. (2021). Leveraging Prescriptive Analytics for Superior Forecasting: Amazon’s
Approach. International Journal of Operations Research.
Groening, C., & Ackerman, D. (2019). Supply Chain Management at Procter & Gamble: A Best
Practices Case Study. Journal of Supply Chain Management.
 14

Harbert, T. (2019). How Walmart Uses Machine Learning and IoT to Revolutionize Retail.
Forbes.
Hickey, A. (2018, June 18). Zara stocks up on AI, big data, analytics partnerships, and
investments. CIO Dive. https://www.ciodive.com/news/zara-stocks-up-on-ai-big-data-
analytics-partnerships-and-investments/525841/
Hunt, R. (2021). The Role of Predictive Analytics in Procter & Gamble’s Demand Forecasting.
Operations Research Perspectives.
Joji, K. (2024, January 29). Decoding Walmart Luminate: Leveraging predictive analytics for
success. Insite AI. https://insite.ai/decoding-walmart-luminate-leveraging-predictive-
analytics-for-success/
Krogerus, M., & Tschäppeler, R. (2020). Data Strategy: How Walmart Uses Data to Drive
Business Success. Journal of Data Management
Loukili, M., Messaoudi, F., Ghazi, M. E., & Azirar, H. (2024). Predicting future sales: A
machine learning algorithm showdown. In Lecture Notes in Networks and Systems (pp.
26–31). Springer Nature Switzerland
Nowadly, K., & Jung, S. (2020). Using machine learning approaches to improve long-range
demand forecasting. https://dspace.mit.edu/handle/1721.1/126503?show=ful
Torkkola, K., & Madeka, D. (2021, August 9). The history of Amazon’s forecasting algorithm.
Amazon Science. https://www.amazon.science/latest-news/the-history-of-amazons-
forecasting-algorith
Vivek, R. (2023). Predictive forecasting models are used by the big box players.
Ruralhandmade.com. Retrieved June 10, 2024, from
https://ruralhandmade.com/blog/predictive-models-for-forecasting-used-by-the-big-box
15