A product business compares two customer cohorts and sees an uncomfortable result: repeat purchase rate fell from 25.0% to 23.75%. The team may conclude that customers are less satisfied, reminders are failing, or product quality slipped.
That conclusion may be wrong.
A blended rate can move because customer behavior changed. It can also move because the cohort mix changed across first products, offers, or channels. A cohort-mix bridge separates those effects before the team changes the product, promotion, or retention plan.
Begin with a consistent repeat purchase definition
A practical formula is:
Repeat purchase rate = eligible first-time customers who purchased again ÷ eligible first-time customers × 100
The repeat purchase rate foundation explains the basic formula, cohort method, and measurement-window choices. This analysis starts after those decisions have been made.
For a fair comparison, lock three definitions:
- the event that makes a customer eligible, usually a completed first order;
- the event that counts as a repeat purchase; and
- the amount of time each customer receives to return.
If the January cohort has 90 days to reorder, give the February cohort 90 days too. Do not compare an older group with six months of opportunity against a newer group with four weeks. Decide how cancellations, full refunds, test orders, wholesale accounts, and subscriptions are handled, then apply that rule to both periods.
Segment the cohort by one meaningful first-order difference
Choose a dimension that could affect why or when customers return. Useful starting points include:
- first product or product family;
- starter set versus replenishable core item;
- full-price versus discounted first order;
- acquisition source or campaign;
- subscription versus one-time purchase; or
- region or fulfillment method.
Use one dimension first. Dividing a small cohort into too many cells creates noisy percentages that look precise but are driven by a handful of people.
The goal is not to find the most flattering segment. It is to understand what kind of first order entered the cohort and whether that mix changed.
Build a two-step cohort-mix bridge
A useful bridge has three rates:
- Starting blended rate: each starting segment rate weighted by the starting customer mix.
- Fixed-mix rate: each new segment rate weighted by the starting customer mix.
- Ending blended rate: each new segment rate weighted by the new customer mix.
The first step—from starting blended rate to fixed-mix rate—shows the customer-behavior effect. It asks what the total rate would have been if segment sizes stayed the same but each segment adopted its new repeat rate.
The second step—from fixed-mix rate to ending blended rate—shows the cohort-mix effect. It asks what happened when the business shifted toward or away from segments with different repeat behavior.
This is a management bridge, not proof of cause. It tells the team where to investigate next.
Worked example: every segment improves while the total falls
A skincare business compares two equal-age cohorts of 400 first-time customers. It segments them by first purchase: a replenishable core product and a discounted sampler.
| Measure | Cohort A | Cohort B |
|---|---|---|
| Core-product first customers | 300 | 100 |
| Core-product repeat rate | 30% | 35% |
| Core-product repeat customers | 90 | 35 |
| Sampler first customers | 100 | 300 |
| Sampler repeat rate | 10% | 20% |
| Sampler repeat customers | 10 | 60 |
| Total first customers | 400 | 400 |
| Total repeat customers | 100 | 95 |
| Blended repeat purchase rate | 25.0% | 23.75% |
Both segments improved. The core-product rate rose from 30% to 35%, and the sampler rate rose from 10% to 20%. Yet the total fell by 1.25 percentage points because Cohort B contained far more sampler customers.
Now hold Cohort A’s mix constant while using Cohort B’s segment rates:
Fixed-mix repeat customers = (300 × 35%) + (100 × 20%) = 125
Fixed-mix repeat purchase rate = 125 ÷ 400 × 100 = 31.25%
The customer-behavior effect is positive: 25.0% to 31.25%, or +6.25 percentage points. The change in customer mix then moves the rate from 31.25% to 23.75%, or −7.5 percentage points. Together, those effects produce the reported −1.25-point change.
The business does not have evidence that customer retention weakened within either first-product group. It has evidence that the acquisition mix shifted heavily toward a group that still repeats less often, despite improving.
Investigate the behavior effect separately
When a segment’s rate changes, review what customers in that segment experienced:
- product availability near the expected reorder date;
- formula, material, size, scent, packaging, or instruction changes;
- delivery time, damage, refunds, and customer-service themes;
- reminder timing and replenishment links; and
- the contribution left after repeat-order discounts and shipping support.
Do not assume that a higher rate automatically means healthier customer loyalty. Repeat customers who return only for a deep coupon may add less contribution than a smaller group returning at full price. Pair retention with the AOV and contribution-per-order bridge and realized customer value.
Investigate the cohort-mix effect separately
When mix changes, ask why the business acquired a different first-order population. A sampler campaign, wholesale event, marketplace listing, seasonal launch, or out-of-stock core product can reshape the cohort before any retention work begins.
Compare acquisition spending and first-order contribution by segment. The customer acquisition cost foundation helps keep the denominator and cost period consistent.
Then follow the cohort long enough to judge its real value. A sampler group may repeat less often but buy a larger second order. Another group may return quickly but require repeated discounts. The 12-month LTV reality check shows how to compare forecast customer value with completed orders, refunds, discounts, and contribution.
Turn the bridge into a monthly review
Maintain one table for equal-age cohorts. Record eligible first customers, repeat customers, repeat purchase rate, first-product or offer segment, measurement window, and contribution from repeat orders.
Calculate the blended movement, fixed-mix rate, behavior effect, and mix effect. Then assign separate actions. A negative behavior effect may call for a product, availability, fulfillment, or timing investigation. A negative mix effect may call for reviewing acquisition strategy, the first product being promoted, or the economics of a sampler.
Change one meaningful input at a time. The purpose is not to force every segment toward one benchmark. It is to understand which customers the business is acquiring, what experience they receive, and whether their next order supports both the customer and the business.
Frequently asked questions
Can repeat purchase rate fall even when every customer segment improves?
Yes. If the customer mix shifts toward a segment with a lower repeat rate, the blended rate can fall even when each segment improves. A fixed-mix bridge makes that effect visible.
What is the best window for measuring repeat purchase rate?
Use a window that matches the normal replenishment or buying cycle, then give every compared cohort the same amount of time. Label the window clearly, such as 90-day repeat purchase rate.
Which segments should a product business compare first?
Start with the first product or offer, acquisition source, and discount status. Choose the dimension most likely to change the initial customer experience and keep sample sizes large enough to interpret.
Should canceled and refunded orders count as repeat purchases?
Usually a fully canceled or fully refunded order should not represent a successful repeat purchase. Document the rule and apply it consistently to every cohort.
Is a higher repeat purchase rate always better?
No. Review margin, discounts, refunds, order value, purchase frequency, and customer value. A higher rate supported by unprofitable incentives may weaken the business even while the retention dashboard improves.
