Which Customer Segments Predict the Next Order?

Customer segments timed to each store's own buying cycle predicted what customers did next far better than store averages. One-time buyers still inside their store's usual window to a second order were 3.5 times as likely to reorder within 90 days. That held in 98% of the 319 Shopify stores we tested.

The same study showed why a fixed rule, such as calling a customer lapsed after 365 days, fits most stores badly. This post covers how we tested it, which segments held up, the one that failed, and how to apply the findings with your own order data.

How we ran the backtest

We took 319 active Shopify stores that use By the Numbers, each with an established customer base. For two past dates, October 1, 2025 and April 1, 2026, we rebuilt each customer's history as it stood that day. Every segment's settings came only from orders placed before that date. Then we checked what the same customers did over the following months.

We scored each segment by lift: the segment's rate divided by the rate of a comparison group in the same store. A lift of 3 means members did the thing three times as often. The figures below are the median store's lift, first for October and then for April where both are given.

Before looking at results, we set a bar. A segment passed if its median lift was at least 1.2 and it beat the comparison group in at least 70% of stores, on both dates. All results are aggregates across stores; no figure here describes a single merchant.

Start with your store's own buying cycle

Most of the segments rest on one number per store: the median days from a customer's first order to their second. We counted customers who ordered again within a year and skipped second orders placed the same day, which are usually split orders. We measured it on customers whose first order was one to three years old, so slower reorders had time to happen.

That is about a three-fold spread between a fast store and a slow one. A follow-up timed for a 28-day store reaches a 92-day store's customers months early. A lapse rule that fits one store is wrong for the other.

Five segments that held up

Due for a second order. These are one-time buyers whose first order falls between the 25th and 75th percentile of the store's first-to-second gap. They were 3.5 times as likely to reorder within 90 days as one-time buyers overall (3.4 in April). The segment beat that baseline in 98% of stores in October and 99% in April. This is the audience for a second purchase campaign aimed at first-time buyers.

Rising stars. These are customers whose first order was within the past year and whose first-year spend ranked in the store's top 10%. Over the next 180 days they spent 4.8 times as much as new customers overall (4.7 in April), and reordered 2.7 times as often. They are worth recognizing with early access or a personal note before their second year.

Lapsed high-value customers. These are customers in the top 20% by total spend whose last order was more than three times the store's typical gap ago. Over the next 180 days they spent 3.1 times as much as lapsed buyers overall (3.2 in April). Part of that is expected, since big spenders tend to keep spending. The practical point is that a win-back budget goes further on this group than on the whole lapsed list.

Lapsing repeat buyers. These are repeat customers whose last order was 1.5 to 3 times the typical gap ago. They reordered at 0.58 times the rate of repeat buyers who had ordered more recently, but 4.2 times the rate of those further gone. That window is where a win-back sequence for at-risk customers still has the most people to win.

Full-price loyalists. In stores where at least 30% of orders used a discount code, this is repeat buyers who had never used one. Over the next 180 days, their orders used a code about half as often as repeat buyers' orders overall (0.52 and 0.49). They also reordered less than other repeat buyers (0.75 times). That makes the segment useful for leaving people out of discount sends, and a poor choice for targeting.

Several narrower segments passed too. Buyers whose first order included a store's strongest product type or collection came back about 1.5 times as often as other one-time buyers, comparing only customers who joined in the past year.

The segment that failed: replenishment due

The idea was simple: in stores with a tight buying cycle, remind repeat customers when their last order nears their usual gap. It failed for two reasons.

First, tight cycles were rare. We defined one as the slower reorders (75th percentile) taking at most twice as long as the faster ones (25th percentile). That held in only 7 to 9 of about 300 stores, mostly with gaps of exactly 30 or 60 days. Those look like subscriptions, which renew without a reminder. In the median store, the slower reorders took 6.8 times as long as the faster ones.

Second, it lost to a simpler rule. Repeat buyers who had ordered more recently than the "due" group were more likely to reorder in the next 30 days, in 85% of stores. "Ordered recently" beat "due now."

There is no "due now" moment

The chance of a reorder falls steadily after each order. It does not peak at the typical gap, for one-time buyers or for repeat buyers. These are the median store's figures for October 2025; April gave the same curve.

So a segment window means "still likely," "fading," or "gone," not "due." The earliest window is the strongest, so a second-order email does not need to wait for the typical gap. In this data, "gone" starts at about five times the typical gap. For the 28, 50, and 92-day stores above, that is about 140, 250, and 460 days.

A 365-day lapse rule is 13 cycles late for the 28-day store. For the 92-day store it comes at four cycles, when those customers still reorder above the repeat-buyer average.

Young stores: your cycle looks shorter than it is

A young store's own data understates its buying cycle. When we rebuilt stores from only their last 90, 180, or 365 days of orders, their gaps came out 3.0, 2.1, and 1.5 times too short, in 90% of stores. Only quick reorders had had time to happen.

The fix is to measure only customers whose first order is at least a year old. Until you have about 25 of them who ordered again, a typical figure for stores in your category is a useful guide, and blending the two works better than either alone. With 10 such repeat buyers, a store's own median gap was typically off by 40%. With 50 it was off by 15%, and with 200 by 7%. A category figure alone was off by 34%.

How to apply this to your store

1. Find your days to second order. Take customers whose first order was one to three years ago and who ordered again within a year. Find the median days between their first and second orders, skipping same-day orders. Note the 25th and 75th percentiles too.
2. Write lapse rules as multiples of that gap. Treat 1.5 to 3 times the gap as lapsing, and more than about five times as gone. Drop the fixed 365-day rule.
3. Contact one-time buyers early. The best window opens right after the first order. Shape the follow-up around the 25th to 75th percentile window rather than a calendar date.
4. Spend win-back effort where it pays. Put the lapsed high-value group first, and keep full-price loyalists out of discount sends.
5. Skip replenishment reminders unless your reorders cluster tightly. If they do, check whether they come from subscriptions first.
6. On a young store, borrow a category figure and switch to your own once about 25 customers with year-old first orders have come back.

In By the Numbers, the Time Between Orders report shows the median days between consecutive orders, by purchase sequence. You can then build each group above as a customer segment from rules over order history and spend. Because a segment is defined by rules, its membership follows customers as they move between groups. For catching churn signals before customers reach the lapsed window, see predicting churn with segments.

What this study does not show

These results are about prediction, not persuasion. A segment that finds the people likely to reorder does not prove a campaign would change what they do. To measure that, hold a random slice of each segment out of the campaign and compare the two groups.

The sample is active stores with established customer bases, so newer or smaller stores may see different numbers. Spend figures compare averages, and a few large customers can move an average. Each figure is the median store's result, and your store may sit well away from the median. Measure your own cycle before you change a rule.

Keep reading

If you want the full picture on customer segmentation, start with the customer segmentation playbook.

Related reading: Build Segments That Self Update in Shopify, Segmentation Mistakes Shopify Stores Make and Dynamic vs Static Customer Segments in Shopify.

By the Numbers builds this into the dashboard. See customer segments.