SELECT COUNT(*) AS n,
Categories.CategoryName
FROM Orders
INNER JOIN "Order Details" AS od
ON Orders.OrderID = od.OrderID
INNER JOIN Products
ON od.ProductID = Products.ProductID
INNER JOIN Categories
ON Products.CategoryID = Categories.CategoryID
WHERE Orders.CustomerID = 'HANAR'
GROUP BY Categories.CategoryNameAnalysis Report Three - Privacy and Customer Profiling
Executive Summary
Customer data can help organizations understand purchasing patterns, improve services, and deliver more relevant marketing. However, detailed customer profiles also create privacy and ethical concerns. Customers may not fully understand what information is collected, how separate records are combined, or how a company uses the final profile. The main readings show that trust depends on more than a privacy policy or words such as ethical and responsible. Companies need to clearly explain their practices, give customers meaningful control, and provide fair value in exchange for the information they collect.
This report profiles Hanari Carnes, a Northwind customer in Rio de Janeiro, Brazil. The analysis looks at the product categories connected to the customer’s purchases and the average discounts received in each category. Beverages appear most often, while Condiments, Dairy Products, and Seafood also appear regularly. The discount analysis shows that some categories were connected to larger average discounts than others. These results demonstrate how routine transaction records can be combined into a customer profile, but they do not fully explain why the customer made each purchase.
Organizations should use purchase profiles only for clearly stated purposes. They should collect only the information they need, explain how customer profiles affect advertising and offers, provide access and deletion controls, and regularly review profiles for errors or unfair assumptions. Retailers should also use transparent and independent measures when reporting advertising performance. These steps can help organizations benefit from customer data while still respecting privacy.
Introduction
Companies collect customer data because it can improve decisions about products, services, inventory, advertising, and customer relationships. The main reading separates personal information into three broad types. Self-reported data is information that a person chooses to provide, such as an email address or age. Digital exhaust is created through the normal use of digital services, such as browsing history or location. Profiling data is created when a company combines information and uses it to predict a person’s interests or behavior (Morey, Forbath, and Schoop 2015).
The difference matters because customers do not value every type of information in the same way. Profiling data can reveal more than a customer directly chose to share. It can also be used for targeted advertising or sold to another company. The main reading explains that customers expect more value in return when data is more sensitive or when its use benefits the company more than the customer (Morey, Forbath, and Schoop 2015).
Transparency and trust are closely connected. Customers are more willing to share data with a company they trust, especially when the company explains what it is collecting, why it needs the information, and how the customer will benefit. The reading also shows that people are often unaware of their own data trail. Only 25 percent of surveyed consumers realized they were sharing location information, and only 14 percent realized they were sharing browsing history. At the same time, 97 percent expressed concern that businesses or governments might misuse their information (Morey, Forbath, and Schoop 2015).
This creates a difficult balance. Companies can use customer data to provide convenience and personalization, but the same systems can become invasive when customers do not understand them or cannot control them. Customer profiling can support useful business decisions, but companies should limit profiling to a clear purpose and make the process understandable, controllable, and fair.
Privacy vs Profiling
Retailers now use customer transaction data for more than traditional sales analysis. Many operate their own advertising platforms and use first-party purchase information to target coupons, sponsored search results, and advertisements. This can be valuable because retailers often know what customers actually purchase rather than only what they click. Amazon, Walmart, Target, Lowe’s, and other companies have built large retail-media businesses around this information (Gabel, Simester, and Timoshenko 2024).
Retail media also creates a transparency problem. The retailer may sell the advertisement, decide where it appears, and report whether it worked. Brands may not be able to see the impressions they purchased, the retailer’s margins, or the method used to assign a sale to an advertisement. The reading recommends transparent reporting systems and independent measurement so advertisers can understand what they are buying (Gabel, Simester, and Timoshenko 2024). The same principle should apply to customer profiling. People should not have to rely only on a company’s claim that its system is fair.
The article about corporate trust words supports this point. Companies that used words such as ethical, integrity, and responsibility in annual reports were associated with weaker stock reactions, higher audit fees, and a greater chance of receiving an SEC comment letter. The researchers did not prove that the words caused these outcomes. Instead, the findings suggest that companies facing more risk may be more likely to use language that tries to create trust (Ward 2024). Trust should come from visible practices instead of labels.
Personalized artificial intelligence adds another concern. Technology companies are developing AI systems that learn a user’s tone, interests, and values so the system feels more personal. This may reduce loneliness and make the product more engaging, but it can also create dependence or reinforce a narrow worldview. An AI system that tries to act like a friend may influence users without clearly showing how its answers are shaped (Higgins 2025). This is another form of profiling because the system adapts based on information gathered from the user.
Outside research shows that these concerns are widespread. A 2023 Pew Research Center survey found that 81 percent of Americans were concerned about how companies use the data they collect. Seventy-three percent said they had little or no control over company-collected data, and 67 percent said they had little or no understanding of what companies do with it (McClain et al. 2023). These findings suggest that privacy policies alone are not creating real understanding or control.
A 2024 Federal Trade Commission report examined major social media and video-streaming companies and found broad collection, retention, and sharing of personal information. The report described business models that encourage mass data collection for targeted advertising and recommended stronger data minimization, retention limits, user controls, and transparency (Federal Trade Commission 2024). Together, the readings and outside research show that organizations should treat privacy as part of product and system design rather than as a notice added at the end.
Profiling Example
Hanari Carnes was selected because the customer has enough Northwind transactions to show patterns across several product categories. The company is located in Rio de Janeiro, Brazil. The analysis uses order line items rather than only order totals because one order may contain products from several different categories.
The profile is limited to ordinary transaction data. It does not contain private health information, personal beliefs, or online behavior. Even so, joining the tables creates information that does not appear in any single table. A company could use the final profile to decide which products, coupons, or advertisements to show this customer.
Visualization One: Purchase Frequency by Category
The first analysis looks at how often each product category appeared in Hanari Carnes’s purchases. Combining the customer’s order information with product and category data shows which categories were purchased most often.
ggplot(
data = myquery1,
aes(
y = CategoryName,
x = n
)
) +
geom_col(fill = "lightblue") +
theme_minimal() +
labs(
title = "Hanari Carnes Purchases by Product Category",
subtitle = "Number of order line items in each category",
x = "Number of Line Items",
y = "Product Category",
caption = "Source: Northwind SQLite database"
)
Beverages appears most often with seven line items. Condiments, Dairy Products, and Seafood each appear six times. Confections appears three times, Grains/Cereals appears twice, and Meat/Poultry and Produce each appear once. A retailer might use these results to describe Hanari Carnes as a customer with broad interest in beverages, dairy products, seafood, and condiments.
The graph does not prove that these are the customer’s favorite categories. A line item can contain a small or large quantity, and the database does not explain why the customer made the purchase. The pattern may reflect inventory needs, prices, seasonal demand, or the needs of the customer’s own business. The profile is useful for identifying a pattern, but it should not be treated as a complete explanation of behavior.
Visualization Two: Average Discount by Category
The second analysis compares the average discount connected to each product category. This helps show whether certain categories were purchased when larger discounts were available.
SELECT AVG(od.Discount) * 100 AS AverageDiscountPercent,
Categories.CategoryName
FROM Orders
INNER JOIN "Order Details" AS od
ON Orders.OrderID = od.OrderID
INNER JOIN Products
ON od.ProductID = Products.ProductID
INNER JOIN Categories
ON Products.CategoryID = Categories.CategoryID
WHERE Orders.CustomerID = 'HANAR'
GROUP BY Categories.CategoryNameggplot(
data = myquery2,
aes(
y = CategoryName,
x = AverageDiscountPercent
)
) +
geom_col(fill = "lightblue") +
theme_minimal() +
labs(
title = "Average Discount by Product Category",
subtitle = "Discount percentages connected to Hanari Carnes purchases",
x = "Average Discount Percentage",
y = "Product Category",
caption = "Source: Northwind SQLite database"
)
Grains/Cereals has the highest average discount at 20 percent. Produce follows at 15 percent, and Condiments averages about 11.7 percent. Beverages and Dairy Products average close to 6 percent, while Seafood averages 2.5 percent. Confections and Meat/Poultry have no average discount in the recorded line items.
These averages should be interpreted carefully. Produce appears only once and Grains/Cereals appears only twice, so their percentages are based on very few observations. The analysis also does not prove that a discount caused the customer to purchase a product. It only shows that some categories were connected to larger discounts in the available transactions. A company could use this pattern to test future offers, but it should not assume that discounts are the customer’s only motivation.
Together, the two graphs create a basic purchasing profile. The first shows category frequency, while the second shows the discounts connected to those categories. This is similar to the type of information a retailer could use for targeted promotions. It also demonstrates why transparency matters. A customer may understand that a company stores individual purchases but may not realize that those purchases can be combined into a profile of category interests and possible price sensitivity.
Recommendations for Industry
Explain profiling in plain language
Organizations should clearly tell customers when purchase records are combined to create profiles. The explanation should describe what information is used, what the profile affects, and whether it changes advertisements, prices, recommendations, or access to services. A long legal privacy policy is not enough if customers cannot understand the actual process.
Collect and retain only necessary information
Companies should connect data to a specific business purpose before collecting it. A retailer may need purchase history to recommend related products, but it may not need unrelated location, contact, or behavioral information. Retention schedules should remove data that is no longer needed. This reduces security risk and limits the possibility that information will be reused for an unexpected purpose.
Give customers meaningful control
Customers should be able to view, correct, download, and delete information used in their profiles. They should also be able to opt out of targeted advertising without losing basic access to the product or service. These controls should be easy to find and use.
Separate advertising from measurement
Retailers operating advertising platforms should provide clear performance measures and allow independent review. The team selling or placing an advertisement should not be the only team deciding whether it worked. Transparent measurement can protect both customers and advertisers from misleading claims.
Avoid unsupported conclusions
A transaction profile shows what happened, but it does not always explain why. Managers should not assume that frequent purchases prove personal preference or that a discount caused a purchase. Small groups should be reviewed carefully, and important decisions should include human judgment and additional evidence.
Make privacy part of system design
Privacy, security, and fairness should be considered before a profiling system is launched. Organizations should test for errors, limit employee access, document how profiles are created, and review automated decisions. Trust is more likely when customers can see that responsible practices are built into the system instead of added only after a problem occurs.
References
Federal Trade Commission. 2024. “A Look Behind the Screens: Examining the Data Practices of Social Media and Video Streaming Services.” Federal Trade Commission.
Gabel, Sebastian, Duncan Simester, and Artem Timoshenko. 2024. “How Retailers Became Ad Platforms.” Harvard Business Review.
Higgins, Tim. 2025. “In a Lonely World, Elon Musk, Mark Zuckerberg and Even Microsoft Are Vying for Affection in the New Friend Economy.” The Wall Street Journal.
McClain, Colleen, Michelle Faverio, Monica Anderson, and Eugenie Park. 2023. “How Americans View Data Privacy.” Pew Research Center.
Morey, Timothy, Theodore Forbath, and Allison Schoop. 2015. “Customer Data: Designing for Transparency and Trust.” Harvard Business Review 93 (5): 96–105.
Ward, Lisa. 2024. “Beware When a Company Says It Is Trustworthy.” The Wall Street Journal.