What is a replenishment email?

A replenishment email reminds a customer to buy a product again when they are likely to be running low. The timing runs from their last order to the point the product is expected to run out.

The setup is the easy part. Two things are harder: choosing when it goes out, and telling whether it creates reorders or only takes credit for reorders that were coming anyway.

A replenishment email, in short. It goes to someone who bought a product that gets used up, a little before it runs out. Where the amount bought decides how long it lasts, time it from that supply. Where customers use it at different speeds, start from the median gap between orders of that product among buyers who have had time to reorder, then test the delay. Stop it the moment they reorder. Because it lands near reorders that were likely anyway, its attributed revenue can include orders it did not cause. A group held out at random is the clearest way to measure what it adds.

What is a replenishment email?

An automated email that reminds a past buyer to reorder something they use up, such as a supplement, coffee beans or a pack of seltzer.

Klaviyo’s help page on replenishment flows describes it as a flow for “products that your customers purchase repeatedly within a certain timeframe”, timed from the buying cycles you find in your own purchase data.

How is it different from a win-back email?

By timing, and by what it assumes about the customer.

A replenishment email assumes the customer is still buying and is about to need more. A win-back email assumes they have been away longer than expected. The same customer can move from one to the other: a missed reorder window can become a win-back trigger later.

A replenishment email can sit inside a wider post-purchase email flow. It is the email in that flow whose timing depends on knowing how long the product lasts.

What starts and stops the flow?

A qualifying order starts it. A qualifying new order stops it.

In one email platform’s setup guide, the flow is triggered by a placed order and can be limited to buyers of one product. A filter “checks before every email sends to ensure that customers have not purchased a product since entering the flow.” If the flow covers one product, the guide says to limit that check to the same product.

Without the check, someone who reordered yesterday is told they are running low.

When should the first reminder go out?

Before the product runs out, with enough time left to order and receive more. The day counts vendors suggest are rules of thumb, not tested results.

One email platform’s help page gives the example of a supplement with a 30-day supply: send the first reminder “about 25 days after customers enter the flow.” The same platform’s blog guide puts the delay for a 30-day supply at 20 to 25 days, “to give customers enough time to place a new order and receive it before they run out of their supply.”

Both describe one pack. At the same consumption rate, two 30-day packs represent twice the supply of one, so time the reminder from the whole order. The gap before run-out exists for delivery, so a store that ships slowly needs the earlier end.

The guide also names the cost of each mistake. Too soon, and you could end up “encouraging customers to repurchase your product well before they need it.” Too late, and “your customer may have already found an alternative.”

What if customers use the product at different speeds?

Then time it from that product’s reorder history: the median gap between its reorders, from buyers who have had time to reorder, tested against real results.

The guide’s example: “a 24-pack of seltzer could take one person four weeks to drink, while another person might finish it over the span of four months.” Its answer is to use how long the average customer takes. If that is 10 weeks, set the delay at nine weeks, and keep checking the estimate as behaviour changes.

Work it out per product, not per store. A cosmetics brand’s foundation, the guide notes, may last longer than its mascara. And treat the result as an estimate: your orders show when people bought again, not when they ran out.

Should you use the average time between orders?

Use the median of the product’s own reorder gaps, not the mean, which a few slow buyers pull later. Take the gaps from buyers who have had time to reorder.

Eightx’s reorder-interval benchmark defines its store-wide interval as “the median gap, in days, between consecutive orders from the same customer.” Its reason for the median: order gaps are right-skewed, “so a handful of once-a-year buyers drag the average up and out of usefulness.”

For a replenishment email, narrow that to one product. For each customer who bought it more than once, count the days between one order containing the product and their next order containing it. Sort those gaps from shortest to longest. If there is an odd number, take the middle gap; if there is an even number, average the two middle gaps. A store-wide gap mixes in orders for unrelated products. Keep orders of very different sizes apart where you have the volume: a two-pack order may support a later reminder than a one-pack order, but that remains an estimate to test.

The calculation only sees customers who have already come back. Anyone who has not reordered yet has no gap to count. Statisticians call this right-censored data: the times that run past the end of the observation window are missing. In a recent group of buyers, the quick repeaters are the ones counted, so the median looks earlier than it really is.

So choose the group of buyers and the date range before you look at the result, and use buyers from long enough ago that most of those who will reorder have had time to. Note how many gaps the median rests on, and treat it as a starting point to test.

Do not borrow the number. The same page says “No single public report publishes a clean vertical-by-vertical table of average days between orders”, and describes its own category figures as best estimates rather than single-source figures. Your own median gap is the number to build on.

Can a tool predict each customer’s next order?

One email platform’s predicted next-order date can help a store with enough order history, but its maker says it is not expected to be exact for any one person, and it does not account for which product was ordered.

Klaviyo’s predictive analytics estimate an expected date of next order. The feature needs order data sent from an ecommerce integration or the API, at least 500 customers with orders that were not cancelled, refunded or zero-value, at least 180 days of order history with orders in the last 30 days, and some customers with 3 or more orders.

Where a customer’s orders show a pattern, the prediction follows it. Where they do not, it falls back on how the store’s other customers behave. A one-time buyer’s date is calculated from data across all customers, so it reflects the store’s customers in general, not that buyer’s habits.

The date is not tied to a product. The same page says the prediction “doesn’t consider what products the customer ordered”. For products with distinct cycles, it recommends separate order-triggered flows instead, each limited to products that share a cycle and each with its own delay. It also warns against counting down to the date, because repeat customers “will simply get the same sequence of emails leading up to every order which may result in unsubscribes.”

So the predicted date does not replace a replenishment flow. Where products have known cycles, or the store is below those thresholds, the pack size and the product’s median reorder gap are the working method.

How many reminders should you send?

Two reminders, then one follow-up with an incentive after the expected reorder date, in one vendor’s rule of thumb. It is not a tested number.

The help page suggests “two reminder emails, and then one follow-up after the projected buying cycle has passed that includes an extra incentive like a discount or a coupon”, and says to “remember not to badger your customers.”

The next section covers the discount. To choose the count, split buyers at random between versions with different numbers of reminders, and compare reorders first and unsubscribes second.

Should the last email carry a discount?

Only if a separate test shows it pays for itself.

The follow-up goes out after the expected reorder date to people who have not reordered yet. Because order gaps are right-skewed, some of those people are simply slower than typical and may have reordered anyway. If they use the discount on an order they would have placed anyway, the discount gives away margin, and the flow’s report can still count the order.

Testing the whole flow cannot separate the discount from the reminders. To test the discount, take the customers who reach the last email and, at random, send half of them a version without it. Compare how many reorder and the margin left after the discount. Keep it only if the extra orders are worth more than the margin given up on orders that would have come anyway.

What should the email say?

Mostly: here is the thing you bought, and here is how to get more.

One email platform’s help page suggests content “that reminds the customer to purchase this same product again.” It also suggests related products. For coffee beans, that could be “a new flavor or a mug.”

Keep the reorder as the main action. A suggestion works as a smaller second one.

Do reminders actually change what people do?

They can: in one randomized test, weekly email reminders raised gym members’ weekly exercise frequency by 13%. That test was not a store.

In a field experiment with 2,463 gym members, Habla and Muller sent weekly email reminders to a randomly chosen half. In the three months after the last reminder, visits were still 12% higher, though the authors note that result is less precise and slightly smaller in some robustness checks. The reminders did not change how long people kept their memberships or whether they renewed.

The authors say “Limited attention and habit formation can best explain these results.” They also note that an email deleted unopened may still have worked, by bringing the gym “to the top of the receiver’s mind.”

That is a plausible reason a replenishment email could work: it reaches someone who has not got round to reordering. It does not show that your recipients meant to reorder, or how much an email would add in a store.

When can a reminder fall short?

When the effect fades, or attention is not the only barrier. In a large randomized trial, refill reminders sent once a refill was at least 7 days overdue left adherence higher at 3 months in an exploratory analysis, but not significantly higher at 12 months.

The trial, published in JAMA, enrolled 9,501 patients at 3 US health care systems who were late refilling a cardiovascular medication. It tested three kinds of refill reminder against usual care. In the adjusted analysis at 12 months, the average share of days covered by filled medication was 2.2, 2.0 and 2.3 percentage points higher in the three reminder groups than with usual care, and none of those differences was statistically significant after correcting for multiple comparisons.

The reminders were texts, or automated phone calls for patients without cell phones. Unlike refill reminders sent before the due date, these went out only once the refill was late.

A separate analysis in the same paper, which the authors call post hoc and exploratory, looked at the first 3 months. There, the average share of days covered was 5.6, 4.8 and 5.2 percentage points higher in the three reminder groups than with usual care, and the median initial refill gap was about 5 days shorter. Most patients refilled within 30 days of the reminder. The authors say the lack of benefit beyond 30 days “may be related to the typical intervention decay observed for other interventions”, and that the 12-month measure “may not have been a sensitive enough measure to assess the impact of the intervention.”

For the 12-month outcome, the authors concluded the reminders “did not improve medication adherence”, and that “Poor medication adherence may be due to multiple factors.”

Neither study is a store, and neither tested a reminder sent before a product runs out. Read together, they suggest a reminder can change what people do in the near term, with no guarantee the effect lasts, and that it may help a customer who has not got round to reordering, but it cannot be assumed to win back one who switched brands or no longer needs the product. Test your own flow rather than borrowing either result.

Why can a flow’s attributed revenue overstate what it adds?

Because Klaviyo can credit an order to an email when the order occurs during that message’s configured attribution window and the recipient has taken an attributable action such as opening or clicking. The window starts when the profile first receives the message. That sequence does not show that the email caused the order.

Klaviyo’s attribution page sets new accounts to last-touch attribution with a five-day window for email opens and clicks. The event-attribution window starts when the profile first receives the message; it is not restarted by each open or click. In the page’s example, the subscriber receives and opens the email on day 1, clicks on day 2 and buys on day 4, so the email gets credit; a purchase on day 12 does not. A later eligible message, or changed settings, can move the credit elsewhere.

A replenishment email is timed to land shortly before the reorder the store already expects, so some of the orders credited to it may have been placed without it. The window can also miss an effect that shows up later. Either way, attributed revenue is not a measure of what the flow adds.

How do you test whether the flow is adding orders?

Hold some buyers out at random.

Before anything is sent, split the buyers entering the flow at random. Most get the emails, and a held-out group gets none of them. Decide the reorder window in advance. When it closes, use the difference in the share who reordered as the primary estimate of what the flow adds. Also compare unsubscribes and margin after any discount. With few buyers the difference may not be clear; treat an unclear result as inconclusive, not as proof the flow works.

If timing matters, compare the cumulative share that has reordered by each day while keeping non-reorderers in both groups through the cutoff. Do not compare days to reorder only among the people who came back: the flow can change who is in that group.

Where does this advice stop applying?

It does not apply to products that do not run out, such as a sofa or a one-off gift, or to customers on a subscription that refills the product automatically. One email platform’s guide separates the two: a replenishment email is for a customer who “isn’t automatically repurchasing the product.” Leave active subscribers out of the flow. The timing is also coarse when one product sells in very different order sizes or bundles: estimate those separately where you have the volume, or treat one shared delay as a first test.

The reminder research comes from a gym and from patients refilling medication. It shows what reminders can and cannot change, not what an email adds in a store. The day counts here (25 of 30 days, 20 to 25 days, nine of 10 weeks) are vendor rules of thumb, not tested best timings. Published reorder intervals by category are estimates, and even your own product’s gaps show when people bought again, not when they ran out. Five days is one tool’s default attribution window for new accounts, and a store that changed its settings has a different window.