Customer Analytics Summary Report
This analysis was conducted to understand customer behavior, improve targeting, and support data-
driven business decisions. The dataset was analyzed using segmentation techniques, classification
models, and propensity modeling.
Part A – Customer Segmentation
Customers were segmented based on their behavioral patterns, such as visit frequency, total spend,
and average basket value. Using clustering techniques, customers were grouped into 4 distinct
segments, each representing a different type of behavior.
These segments exist because customers naturally differ in how often they visit, how much they
spend, and how engaged they are with the business. Some customers show high loyalty and
spending, while others interact only occasionally or spend less per visit.
From a business perspective, each segment should be treated differently. High-value and loyal
customers should receive loyalty rewards and personalized offers to retain them. Medium-value or
frequent low spenders can be encouraged to increase basket size through bundles or cross-selling.
Occasional or low-engagement customers should be targeted with re-engagement campaigns or
introductory offers to increase activity and reduce churn.
Part B – Logistic Regression (Repeat Visitor Analysis)
A logistic regression model was built to predict whether a customer is a one-time visitor (0) or a
repeat visitor (1). Behavioral variables such as visit frequency and spending patterns were used as
predictors.
The model showed that customers with higher visit counts and higher spending are significantly
more likely to be repeat visitors. The direction of impact was positive, meaning as visits or spend
increase, the probability of repeat visits also increases.
In business terms, this indicates that engagement drives loyalty. Customers who interact more with
the business are more likely to return. Therefore, strategies that encourage early repeat visits—such
as welcome offers, loyalty points, or reminders—can help convert one-time visitors into repeat
customers.
Part C – Discriminant Analysis
Discriminant Analysis was applied as an alternative classification approach and its results were
compared with Logistic Regression. Both models largely agreed when customer behavior patterns
were clearly separated, especially for high-value and low-value customers.
However, differences appeared when customer behavior overlapped across segments. Discriminant
Analysis relies on stronger statistical assumptions and performs best when data is well-structured,
while Logistic Regression is more flexible and robust with real-world data.
Overall, Logistic Regression is more trustworthy for business use because it makes fewer assumptions
and handles overlapping behaviors better. Discriminant Analysis is useful as a validation tool but
should not be the primary decision model.
Part D – Purchase Propensity
A simple purchase propensity model was developed to estimate the likelihood of higher spending.
Customers were scored based on their behavioral features and assigned a probability of being high
spenders.
Customers with high visit frequency and higher basket values showed the highest propensity scores
and should be targeted first. These customers are already engaged and are most likely to respond
positively to targeted offers.
The recommended business action is to prioritize these high-propensity customers with personalized
promotions, loyalty rewards, and premium bundles. Medium-propensity customers can be
encouraged using cross-selling and value-based offers, while low-propensity customers should
receive awareness or re-engagement campaigns rather than heavy discounts.
Overall Conclusion
This analysis demonstrates how customer data can be transformed into actionable insights.
Segmentation helps understand different customer groups, classification models explain repeat
behavior, and propensity modeling guides marketing prioritization. Together, these techniques
enable the business to improve targeting, optimize marketing spend, and enhance customer lifetime
value.