Consumer Data Analytics: How do Businesses Use Consumer Data to Make Strategic Decisions and Improve Consumer Experience?

Every business keeps records. They keep records of transaction details, the date and time of purchase, and the names of customers. This is only to be expected as they wouldn't be very good businesses if they didn't do this. However, many customers do not know that businesses also collect data on their interactions with online platforms, time spent in a particular area (whether online or physical), what they thought about buying but ended up not buying, consumption patterns, demographic details and much more. The way in which the businesses use these data to profile customers, calculate business metrics, acquire and retain customers, and make business decisions is a fascinating aspect of the data-driven society that we live in.

Transactional Data vs Behavioural Data

Transactional data records the exact details of a purchase a customer makes at the time of purchase. It includes the transaction ID, customer name, date and time of the purchase, the quantity and type of product purchased, the prices, the payment methods and discounts (if any). Sales returns and subscription renewals also fall under transactional data. Transactional data is recorded very efficiently in both physical shops and online platforms using Point-of-Sale (POS) software. The software is connected to barcode scanners and card readers in physical shops and e-commerce engines in online platforms to record the transaction details including the time of purchase, payment methods, payment account, discounts and sales return. Additionally, a POS software also tracks inventory movement, order details, pricing, and more which is why it forms the foundation for data analysis in business.

Unlike transactional data, behavioural data aims to map the customer's journey from the point of entry to the point of purchase as well as all the other interactions that the customer makes with the business. In online platforms, behavioural data can be divided into first-party, second-party, and third-party data. First-party data is the data collected directly from your interactions with the platform. This includes website clicks, time spent in the website or on specific areas within the website, and scrolling behaviour. Additionally, the businesses could also monitor items added to cart (that you do not end up buying) or form submission status.

Second-party data is the first-party data of a partner business or organisation gained by the business through mutual agreement. One example of this can be found in co-branded credit cards. Partner airline companies can use the purchasing patterns and frequent travel destinations provided by the credit card companies to make personalised suggestions to the consumers. In exchange, the airline company can provide the credit card companies with booking habits, in-flight purchases and physical location. The data is exchanged between the two businesses through Data Clean Rooms (DCRs) which are digital environments where the businesses can match datasets.

Third-party data are purchased from online data-brokers that have no direct relation to the consumers. Tracking pixels, JavaScript tags and third-party cookies track your interactions with different news sites, websites, and e-commerce sites. These third-party data brokers compile data from a variety of different sources to provide businesses with a deeper insight on the customer. This can help the businesses access demographics information and purchasing data, although raising concerns of consumer privacy.

Using Data to Profile Customers

The collection of the data is only the first step in the process of data analytics. The next step is to use the data to get deeper insight on the type of customer that the business is facing.

A key profiling technique is the RFM analysis. RFM analysis uses data to determine how recently the customer made the purchase, how frequently the customer makes purchases, and the monetary value of their purchase(s). The goal of RFM analysis is to segment consumers to take targeted action. If a customer purchased recently, the consumer is more likely to make a future purchase than a customer who made the purchase a long time ago. Similarly, if a consumer makes frequent purchases, this also indicates that they will make a purchase again in the future. First-time consumers can also be transformed into frequent consumers through targeted marketing techniques. This is measured by the Consumer Retention Rate. The monetary value of the purchase also signals the likeliness of a future purchase while also allowing the business to identify high-value consumers. In RFM analysis, a score between 1–5 (with 1 being the lowest and 5 being the highest) can be given to the customer for each category. A business may choose to use a different scoring system or weight the categories differently.

RFM analysis allows the business to identify the segments of customers that they wish to retain. High-Value and Loyal Customers show high recency, high frequency, and high monetary value. New Customers show high recency, low frequency, and either high or low monetary value. At-risk Customers show low recency, historically high frequency, and high monetary value.

The business can use different strategies to target each of these groups. The first group can be retained using loyalty benefits and premium offers. The second group can be converted into High-Value and Loyal customers through follow-up emails and personalised recommendations. Sometimes, upselling (recommending progressively higher priced products) can also be done to increase the monetary value of their purchases. It is also in the business's interest to prevent the third group from leaving. The business will give updates and reminders to those customers to get them to start purchasing again. Additionally, they might also give them exclusive discounts or gift-cards as a last-ditch effort to gain those customers back.

Predicting Future Consumer Behaviour

Behavioural data such as number of visits and website interactions can be used to calculate key business metrics that help predict the future behaviour of consumers. This section will discuss three different metrics: purchasing propensity, churn rate, and customer lifetime value (CLV).

The first metric, the purchasing propensity, calculates the probability of the customer's purchase considering the contribution of various factors to the probability. The logistic regression model is a mathematical model that predicts the probability of something happening considering various parameters. The model takes the form P = 1 / (1 + e^-(β0 + β1X1 + β2X2 + … + βnXn)), where P is the probability of purchase, β0 is the log-odd when all independent variables X are 0, β1, β2, … are weighting coefficients, and X1, X2, … are the independent variables.

In a RFM-driven logistic regression model, the independent variables correspond to the different categories: Recency, Frequency, and Monetary Value. However, other parameters such as platform engagement and interactions can also be used. Parameters can be binary (1 or 0, yes or no). Logistic regression based calculation of purchasing propensity is often a comparatively heavy task, and most businesses can simply perform standard RFM analysis for most requirements.

Customer Churn measures the rate at which customers stop purchasing from a particular business. Customer Churn Rate is calculated as (Customers Lost ÷ Starting Customers) × 100%. One of the business's objectives is to ensure that the amount of customers acquired is higher than the amount of customers lost to churn. The customer churn rate is very relevant in the telecommunications industries where businesses are interested in preventing customers from switching services.

Different industries have different churn rates. According to 2025 statistics from Qualtrics, the average churn rate in the US telecommunications and wireless industry is 21% while the average rate in the retail industry is much higher at 37%. Additionally, it was found that US businesses lose a total of $168 billion to customer churn, with $35 billion of this being preventable by "keeping customers happy". This explains why businesses are very interested in using data to retain customers, especially high value customers.

Calculating the probability of customer churn can be done using a logistic-regression model (as done for purchasing propensity) with different parameters. Positive parameters include the length of time the customer has been purchasing from the business, the variety of products that the customer purchases, and the regularity of the customer's interactions. Negative parameters include usage decline, payment failures (which could lead to involuntary churn), sudden increases in complaints or support tickets, and price increases.

Customer Lifetime Value (CLV) is the total net profit that a customer is expected to bring throughout their time purchasing from a business. The formula is CLV = (Average Revenue Per Account × Gross Margin %) ÷ Revenue Churn Rate. ARPA is the average revenue the customer brings in a specific time frame (a month, a year, etc.). This is multiplied by the Gross Margin as the business is primarily interested in the net profit that the customer brings. It is then divided by the Revenue Churn Rate to determine the CLV for the whole duration of the customer's expected purchasing life. Revenue Churn Rate is the total amount of revenue lost in a particular period (either through customer churn or reduced spendings) divided by the revenue at the start of the period.

In CLV calculations, an annual revenue churn rate of 10% corresponds to an expected purchasing lifetime of 10 years for a given customer (1 / 0.10). If a customer spends $350 at a retail shop in a given year and the shop's gross margin is 40%, a 10% churn means the CLV of the customer is $350 × 0.40 ÷ 0.10 = $1,400. This means that in the 10 years the customer is expected to stay with the business, he or she is expected to bring the business a net profit of $1,400. This customer is more valuable than a customer that spends $400 on their first visit and never returns (although this statement might not hold when considering the "time value of money", which is not considered in this article).

Transactional data such as payment records, as well as behavioural data such as the customer's interactions with online and physical business platforms, help calculate these metrics which are important for the business to make effective decisions.

Applications

In the past sections, we explored the various metrics that can be calculated from transactional and behavioural data. In this section we will explore how those metrics can be used to help businesses make decisions in each area of business.

Resource Allocation

Businesses have limited financial resources, assets, and manpower. Therefore, consumer analytics plays a crucial role in deciding where to allocate the resources for maximum benefit.

The CLV is important because it can be compared to the Customer Acquisition Cost (CAC) which is the business's cost to acquire a new customer. CAC is calculated by dividing the total marketing and sales cost directly related to acquiring new customers — including the cost of ads, salaries of sales and marketing staff, the cost of assets used actively by those teams, and other related expenses — by the total number of new customers in the measured period. By dividing CLV by CAC, businesses get the LTV:CAC ratio, a key metric that helps determine the effectiveness of new customer acquisition efforts. If the ratio is greater than one, it indicates the customer will bring more net-profit to the firm than the cost it took to acquire them. If the ratio is less than one, the business will be less willing to spend resources to acquire the customer.

Additionally, CLV helps businesses allocate resources by allowing them to segment customers. To a business, not all customers are of the same value. Some customers will bring more revenue and net-profit to the business over their tenure than others. This is the quantification of the popular business proverb "80% of your sales will come from 20% of your clients". Therefore, using CLV to assign values to each customer, the business can decide which customers to allocate resources on — account managers (in banking), exclusive discounts, premium benefits, and so on.

Purchase propensity can also be used to reach the same conclusions. A business would not be willing to allocate resources on a customer that they know is not likely to make a purchase again in the future. Churn helps identify customers that are at risk of stopping their purchases from the business. If the CLV of the customer is high, the business will act on this information to prevent the customer from leaving.

In essence, these metrics are used to set retention and acquisition budgets for customers based on their projected spending and churn risks. This helps the business reach its objective of higher ROI, higher customer retention rate, and long-term revenue growth.

One example of a business using these metrics to make decisions is American Express. A student study published by the Harvard Business School highlighted how American Express uses 115 different variables about customer behaviour to power machine learning models predicting customer churn. According to the source, American Express reported that these prediction models can "identify 24% of Australian accounts that will close within the next four months". This knowledge allows American Express to allocate their resources to retaining those customers specifically identified by the machine learning model.

Demand Forecasting and Inventory Management

Customer data including previous transaction history and behavioural data predicting future purchases can also be used for macro-analysis. These data about a multitude of different customers in a particular region can be bundled together to conduct demand forecasting which allows businesses to conduct inventory management more effectively.

Demand forecasting is predicting the type and quantity of products that customers will purchase at a particular period in the future. These forecasts can allow businesses to make purchasing decisions to replenish inventory. In large regional or multinational distributors and chains, this data helps them make decisions about the regional facilities such as warehouses that they place their products in to minimise transportation costs. Additionally, these forecasts can also help optimise the flow of products from one hub to another. For example, if demand forecasting data indicates that customers in State A are very likely to want to buy Product A in the future, the business will stock the warehouse in State A with more of Product A. This reduces delays in transportation and bypasses the transportation costs required to ship Product A to State A from a warehouse in another state.

Personalised Recommendations

Personalised recommendations in online platforms help increase customer lifetime value by encouraging them to make more purchases. It also helps retain customers, because through improved search, businesses are able to improve the customer's shopping experience.

An article from Amazon News shows exactly how Amazon is able to provide customers with personalised recommendations. Amazon inputs customer data such as searches, browsing history, purchase history, shopping activity, and preferences into AI systems including LLMs. These powerful systems are then able to infer the customer's interests and the attributes of products that the customers are potentially interested in. This information can then be used to identify products that the customer is likely to be interested in buying and recommend them directly to the customers. This also optimises search by giving personalised suggestions which improves customer satisfaction and experience.

Through the use of such consumer data — including transactional and behavioural data — businesses are able to make strategic decisions that help retain customers, reduce cost, and increase net profit in the long run.

Bibliography

CS Squared and B Cubed. American Express: Machine Learning for Customer Churn Prediction and More Effective Customer Retention. Harvard Business School Digital Initiative.

Emarsys. Predictive Customer Analytics. Emarsys.

Fernando, Jason. "Behavioral Analytics." Investopedia (2024).

Fernando, Jason. "Churn Rate." Investopedia (2024).

Fullstory. Behavioral Data. Fullstory (2023).

GeeksforGeeks. What Is Predictive Analytics and How Does It Work? GeeksforGeeks (2024).

IBM. Propensity to Purchase. IBM Documentation (2023).

Levine, Ivy. How Amazon Is Using Generative AI to Improve Product Search Results and Product Descriptions. Amazon About (2024).

LP Solutions. The Power of Co-Branded Credit Cards in Loyalty. LP Solutions (2023).

Qualtrics. 30 Statistics About Customer Churn. Qualtrics (2025).

TechTarget. RFM Analysis. TechTarget (2023).

TIBCO. What Is Transactional Data? TIBCO (2023).

Wharton Executive Education. Why Customer Lifetime Value Matters. University of Pennsylvania (2024).