The reorder gap

In May I wrote a six-month plan for a pharmacy and health ecommerce brand, the one from the repurchase case. Four goals, three gated phases, baselines from the store's own orders with business bulk buying, about 12% of revenue, taken out. One of the inputs I inherited was the day a customer comes back: the post-purchase plan put the second order at day 73, and the reminder meant to catch it fired near day 89 for almost everyone.

Then I measured it. Across 2,477 customers who had placed a second order, the median gap was 111 days. Only 23% had reordered by day 45, when the new-buyer series stopped. And the predicted-date reminder matched an observed reorder within 14 days 3% of the time: 58 of 2,165 repeat buyers. For the rest it arrived while the last order was still on the shelf.

Where the 73 came from

I went looking, because a number that cannot be recomputed is not a number. The 73 was an average of averages over about 234 rows, from a segment that favoured fast repeat buyers. That segment now held nobody, so the figure could not be checked by anyone, including the person who first produced it. It had become a memory, and the flow had been firing on a memory for months.

The failure is ordinary and worth naming. An average of averages weights every row equally whether it holds five customers or five hundred, and a segment built to find quick reorders will find them. The number was true of the segment and false of the base.

Why a fixed day misses everyone

Even the right median misses most people, because a median is the middle of a spread, not a point everyone lands on. Reorder timing at this brand depended on what was bought: a 30-day supply and a 90-day supply do not run out together, and a person on two products runs out of one first. A single reminder day for the whole base is right for the few customers who sit at the median and early or late for everyone else.

So the new flow does not use a day. It starts from the purchase and times reminders by product category, from 18,412 orders, and it stops the moment someone reorders. The agency's email team wrote the copy; I built the timing from the orders. It went live at the end of September 2026, so this piece is the diagnosis and the build, not a result from the flow.

The reminder is a service or an insult

At the right time a replenishment reminder is a service: the product is running out and the message says so. Weeks early it is a small insult repeated monthly, and the person learns to ignore the brand's emails before they ever need one. The difference is one query against the store's own orders, and most brands have never run it.

The same query found something else. In a 7,045-profile suppression segment at this brand, 75% of profiles had the same last-activity date, from one bulk account event. Ignoring that event, the median last activity was 461 days ago, and 112 profiles held 75% of the segment's past customer value. Those went on a protected list before anything else was touched. A shared date in a segment is a sign that a system wrote it, not a person, and it is worth checking before a suppression or a winback fires on it.

Do this Monday

1. Export every customer's first and second order dates from the store. Two columns, one row per repeat buyer.
2. Take the median gap in days. Not the mean, which long gaps pull up, and not an average of segment averages.
3. Count how many repeat buyers had reordered by the day your post-purchase series ends. At the brand it was 23% by day 45.
4. Put your reminder's send day on the same line. If it lands weeks before the median, it is reminding people to buy what they still have.
5. Split the gap by product category if your products run out at different rates. The flow should start from the purchase and use the category's gap, and stop when the person reorders.

In one line: measure the gap from first to second order in the store's own dates, by category, and time the reminder from the purchase, not from a plan.