A product business reviews a customer cohort and sees a 90-day repeat purchase rate of 30%. The number looks stable. Yet it leaves an important operational question unanswered: did most repeat customers return in the first month, or did they wait until the final weeks?
Both patterns can finish at 30%, but they imply different inventory, production, marketing, and cash needs. A 30-60-90 day view shows the pace of return.
Start with one consistent customer cohort
Before comparing timing, define who enters the cohort and what counts as a return. 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 formula, measurement windows, and basic segmentation. This diagnostic starts after those rules are set.
Use customers whose first completed order falls in the same fixed period. Exclude test and fully canceled orders. Document how refunds, subscriptions, wholesale accounts, guest checkouts, and merged identities are handled, then apply the rules consistently.
Most importantly, compare only mature cohorts. A January cohort can have a full 90 days of opportunity. An April cohort observed on May 15 cannot. Calling the younger group’s result a 90-day rate would place customers in the denominator before they had a fair chance to return.
Build cumulative 30-, 60-, and 90-day checkpoints
For each first-time customer, calculate the time to second purchase. Then count returns by day 30, day 60, and day 90.
Suppose 500 customers placed their first order in January:
| Checkpoint | Customers who returned by then | Cumulative repeat purchase rate |
|---|---|---|
| Day 30 | 60 | 12% |
| Day 60 | 105 | 21% |
| Day 90 | 150 | 30% |
The day-60 figure includes the 60 customers who had already returned by day 30. These are cumulative checkpoints, not three separate groups.
This table says more than “30% returned.” It shows that 12% returned within the first month, another 9 percentage points were added by day 60, and another 9 points were added by day 90.
Google Analytics documents standard, rolling, and cumulative cohort calculations. Label yours clearly. A cumulative view asks whether a customer returned by each checkpoint; a period-specific view asks how many returned during one interval.
Two cohorts can finish at the same rate but travel differently
Now compare two mature cohorts of 500 first-time customers each:
| Cumulative checkpoint | Cohort A | Cohort B |
|---|---|---|
| 30-day repeat purchase rate | 18% | 12% |
| 60-day repeat purchase rate | 26% | 21% |
| 90-day repeat purchase rate | 30% | 30% |
Both cohorts finish at 30% by day 90. If the report shows only that final number, customer retention appears unchanged.
The timing changed substantially. Cohort A produced 90 repeat customers by day 30, while Cohort B produced 60. Cohort B eventually caught up, but 30 second orders shifted out of the first month. That delay may matter if the business expected quick replenishment, used early repeat orders in its cash forecast, or scheduled production around a faster reorder cycle.
It does not prove customers became less loyal. A larger first order, longer-lasting product, seasonality, stockout, delayed shipping, or changed reminder timing could slow the pattern. The table identifies where to investigate.
Convert cumulative rates into interval returns
Cumulative rates are useful for comparing progress. Interval counts make the operational change easier to see.
For Cohort B:
- Days 0–30: 60 customers returned.
- Days 31–60: 105 cumulative returns minus 60 early returns = 45.
- Days 61–90: 150 cumulative returns minus 105 returns by day 60 = 45.
For Cohort A, the same calculation gives 90 returns in days 0–30, 40 in days 31–60, and 20 in days 61–90.
That is the contrast: Cohort A is front-loaded, while Cohort B is spread later. Examine the interval that changed instead of launching a broad coupon because one dashboard number felt weak.
Match the checkpoints to how the product is used
Thirty, 60, and 90 days are useful examples, not universal rules. Choose checkpoints around the expected buying cycle.
A frequently consumed food item may need 14-, 28-, and 42-day checkpoints. A skincare product might need 45-, 75-, and 105-day checkpoints. Seasonal or durable products may require longer windows.
Use the first product or product family to set a reorder timing expectation. For multi-product first orders, group by the item most likely to determine the next need or analyze common baskets separately. Do not force different product lifespans into one “normal” reorder date.
Investigate the interval that moved
When early returns slow but the final rate holds, review what changed before and around the expected reorder date:
- Was the frequently reordered product in stock?
- Did the first order contain more units, making it last longer?
- Did a size, formula, scent, flavor, package, or usage instruction change?
- Did shipping delays move the customer’s actual start date?
- Did replenishment messages arrive earlier or later?
- Did customers return for a substitute or related item instead?
Compare customer data with stockouts, production dates, formula versions, fulfillment delays, refunds, reviews, and support messages.
Also inspect economics. Faster repeat customers are not automatically better if the second order requires a deep discount, free shipping, or expensive incentive. The AOV distribution test helps show whether a few large orders distort the average, while the customer lifetime value cohort window extends the view beyond one second purchase.
If the overall repeat rate moved as well, use the repeat purchase cohort-mix bridge to separate changed customer behavior from a different mix of first products, offers, or channels.
Run a five-step monthly timing review
- Select one fully mature first-customer cohort.
- Calculate cumulative return rates at three checkpoints matched to product use.
- Convert cumulative totals into interval counts.
- Compare with three prior equal-age cohorts and flag the interval that changed.
- Investigate one operational or customer-experience cause before choosing one measured response.
Keep the table small enough to repeat monthly. The purpose is not to manufacture a perfect retention score. It is to find whether customers return when the business expects them to—and whether inventory, production, and communication are ready when they do.
Frequently asked questions
Is a 90-day repeat purchase rate enough on its own?
No. It shows how many eligible customers returned by day 90, but not whether those returns happened early, late, or evenly across the window.
Should every product use 30-, 60-, and 90-day checkpoints?
No. Set checkpoints around the product’s typical use, replenishment, or seasonal cycle. Keep them consistent when comparing cohorts.
Can one customer count at more than one checkpoint?
Yes, in a cumulative table. A customer who returned by day 30 remains included at days 60 and 90. Count only the first repeat order for this timing analysis.
What if customers buy several product types in their first order?
Group common baskets or choose the product most likely to drive the next need. Avoid blending products with very different expected lifespans without labeling the limitation.
Does a faster second purchase always mean better customer loyalty?
No. Check discounts, refunds, order contribution, satisfaction, and later behavior. Faster timing is useful evidence, not proof of loyalty or profitability.
Practical takeaway
Take one mature cohort and add two earlier checkpoints to the final repeat purchase rate. If the 90-day result looks steady but early returns shifted later, investigate availability, order size, product use, fulfillment, and reminder timing before assuming demand disappeared.
A final percentage tells you how many customers came back. A timing curve tells you when the business had to be ready for them.
