The question behind this project
After the first purchase, who comes back?
An order arrives, the customer leaves, and the shop waits for another visit. We follow groups of customers over time to see how many return and how their buying habits differ.
Read the analysis
E-commerce An analytical case study
Customer repeat purchases and retention over time
Does marketplace activity translate into customers who return?
The finding
Acquisition, order value, and repeat purchasing answer different questions. A cohort view makes the second purchase visible.
Analytical work
Connect marketplace reporting with customer-level cohort, survival, and segmentation analysis.
How many customers bought again?
of customers placed at least two orders in the retained observation window.
Now follow one cohort: January 2018
6,842 first-time customers. Share purchasing in each later month:
Monthly activity is not the cumulative repeat-purchase rate. A customer can appear in more than one month.
01
Start with the customer, not the order ID
Assign customers to the month of their first purchase using the persistent customer identifier. Follow that same group over time. More orders in a calendar month can reflect newly acquired customers without demonstrating stronger retention.
02
Make the denominator visible
Monthly retention divides returning customers by the original cohort size. A blank future cell means the cohort has not reached that age. Kaplan–Meier analysis accounts for customers whose observation window ends before a second purchase.
03
Distinguish customer value from basket value
Cumulative revenue per customer grows with both order count and order value. Compare these separately before interpreting repeat customers as bigger spenders. First-order characteristics can identify groups to study, but an association does not establish an intervention effect.
Decision supported
What to do with this evidence
Frame a second-purchase experiment with a fixed follow-up window. Use the marketplace view to locate operating patterns and the cohort view to evaluate customer behavior.
What this analysis cannot establish
- The marketplace and cohort analyses use different populations and aggregation paths. Do not compare headline totals without matching order status, period, customer identity, and revenue definition.
- Later cohorts have less follow-up. Geographic and first-order associations are observational, not causal estimates.
Continue the investigation