When should the reorder prompt land?

Two lines of arithmetic and one judgement call.

Days of supply = pack size ÷ daily consumption. A 12 lb case of meal-replacement powder, at two 4 oz servings a day, is 192 oz going out at 8 oz a day: empty on day 24. Nothing about that is an estimate — it is division, and every product a customer finishes and rebuys has the number sitting inside it already.

The judgement call is where the prompt lands relative to day 24. We put the window between seven and three days before depletion, and recommend the midpoint.

Why three days is the floor

Three days is a physical constraint, not a psychological one. Prompt on the day the pack empties and the customer is already out. They order, your 3PL picks and packs, a carrier moves it — and for four days they use something else. Whatever they buy to cover that gap is what your reorder prompt was supposed to prevent. If your own door-to-door fulfilment is slower than three days, move the entire window earlier by the difference; the calculator does not know your 3PL, and you do.

Why seven days is the ceiling

Further out than a week and the prompt arrives to someone with a full cupboard. They do not need it, so they ignore it — and you have spent a send teaching them that your reorder messages are not urgent. Do that four cycles running and the flow is dead. Not because the copy was wrong. Because the timing trained them.

Why 30 days is the wrong default for nearly everyone

Thirty days is an accounting convention. It is the billing cycle, so it became the subscription cadence, so it became the reorder prompt. Consumption does not care. A 12 lb case and a 4 lb box from the same brand, sold into the same household, empty on completely different days — and a 60-capsule bottle taken twice daily empties on exactly the same day as a 30-capsule bottle taken once. Pack size on its own tells you nothing. The ratio is the whole answer.

Which is why the two most common failures look like opposites and share a cause: the brand that prompts every 30 days regardless of SKU, and the brand that never prompts at all because it “has a subscription.”

The output is only as good as the consumption rate, and the consumption rate is not on the packaging. Three sources, best first.

1. Your own reorder data. For every customer who bought the same SKU twice, take the median gap in days between order one and order two. That is a measured consumption rate, per SKU, in the exact unit you need. Median, not mean — a handful of stockpilers will drag an average somewhere useless.
2. Cadence changes and skips. Every subscriber who moved from 30 days to 45 told you your default was too fast. Every skip said it again, louder. Most brands log both events and read neither.
3. Asked at signup. Serving size, how often, how many people in the household actually use it. Only worth collecting if something downstream consumes it — a field nobody reads is worse than no field, because it costs you a conversion on the way in.

Where this model breaks

• Households vary. Two people drinking from one bag halves the cycle. If you cannot capture household size, you are modelling an average that fits nobody exactly — which is still far closer than 30 days for everyone.
• People stockpile. A customer who took a two-for-one promo is permanently a full cycle behind your model, until suddenly they are not.
• Consumption is not flat. Supplements are often taken twice a day for a loading month and once after; a skincare routine runs morning and night in week one and once a day by week six. Model a loading period as its own segment if you have one.
• The first cycle is different. A trial size, a sample sachet in the box, the half-finished pack of the old brand still in the cupboard. The first reorder window is usually longer than every window after it.

None of these are arguments for staying on day 30. They are arguments for segmenting first, then running the same division inside each segment.