Purchase Frequency: The Average That Hides Why First-Time Buyers Do Not Come Back

You run a store selling something people use up: coffee, supplements, pet food, skincare. The dashboard says purchase frequency last year was 2.3 orders per customer, which sounds fine. Then you pull up the customers acquired in October and November and a large share of them never placed a second order at all. One store-wide number cannot show you that gap. This article covers what purchase frequency is, how to calculate it properly from your own order export, and how to use the split numbers to time reorders and win-backs.

Key takeaways

• Purchase frequency is total orders divided by unique customers over a period. The store-wide average blends one-time buyers, repeat customers and subscribers into one figure that hides where second orders fail.

• To measure it properly, split the order export by customer, by first product bought and by acquisition month, and use the median days between consecutive orders rather than the mean.

• The shorter the gap between first and second order, the more likely a customer stays and the more they are worth over time.

• Five actions move the number, each tied to a data signal: reminders timed to the observed gap, leading acquisition with the fast-reorder product, subscriptions after order two, bundles from real repeat pairs, and win-backs before the gap doubles.

• Agencies should report purchase frequency by cohort and by subscriber status, never as one blended number.

What purchase frequency is, and the standard formula

Purchase frequency measures how often customers buy from your store in a defined period. Twelve months is the usual window. It gives you a baseline for whether customers are returning more or less often over time.

Purchase frequency = total number of orders ÷ number of unique customers

Illustration, with round numbers: last year your store processed 18,000 orders from 7,200 unique customers. Purchase frequency is 18,000 ÷ 7,200 = 2.5 orders per customer per year.

That number is useful for year-over-year trend tracking. It does not tell you when customers reorder, which product starts strong retention, or which cohort is slipping.

One distinction worth keeping straight: repeat purchase rate is the share of customers who bought again at least once. Purchase frequency is the average number of orders per customer. A store can have a purchase frequency of 2.5 and a repeat purchase rate of 22%, which means a small group of frequent buyers carries the average while most new customers never return.

Why the store-wide average misleads you

Calculating purchase frequency across all customers flattens very different behaviours into one number.

• One-time buyers drag it down. In many consumable stores the majority of customers place exactly one order in a year. Their weight pulls the average far below what the loyal base actually does.

• Subscribers inflate it. A subscription customer on a monthly refill orders eight to twelve times a year. When those customers sit in the same average as everyone else, non-subscriber behaviour becomes invisible.

• Mixed product cadences blur it. Coffee bought every three weeks and a serum bought every ninety days should not share one target. The store-wide metric merges them, so no goal derived from it is right for either product.

• Cohort shifts hide early problems. Customers acquired from a paid campaign in Q4 may behave differently from organic buyers in Q1. The blended average absorbs that signal before you can act on it.

To fix the second-order gap you need purchase frequency per customer, per first product bought, and per acquisition month.

How to calculate purchase frequency correctly from your order export

This part is hands-on. You need an order CSV and a spreadsheet.

Minimum columns: order id, customer id or email, order date, product id or SKU. Revenue is optional but useful.

Filter to a clean window, for example the last twelve full months, and keep that window fixed whenever you repeat the calculation. Longer windows produce higher frequencies, so a moving definition makes trends meaningless.

Per-customer order counts. Group by customer id within the window and count distinct orders per customer. Compute the average and the median. The median is usually the more honest figure because a handful of high-frequency subscribers will not distort it.

Time between purchases. For each customer with two or more orders, sort by date, calculate the days between each consecutive order, then take the median gap. Again, the median resists the one customer who came back after 400 days.

Split by first product bought. Identify each customer's first order in the window and tag it with the main product or category. Compute median days to second order by that first product. This is where you see which products open the door to a repeat purchase and which do not.

Cohort view. Group customers by acquisition month and compute orders per customer and median days to second order within each cohort. For a coffee brand, a median of 28 to 35 days among returning customers is healthy. A median over 70 days, with many cohorts never reaching a second order at all, is a retention problem, not a marketing problem.

Affinsy runs these same splits from an uploaded order history, including the reorder cadence per product, and turns any of them into an exportable customer list. The logic is what you just did in the spreadsheet.

Purchase frequency, the second-order window and lifetime value

The more often a customer buys and the shorter the gap between purchases, the more they are worth. Acquisition cost is paid up front, so every additional order compounds the return on it.

• The second-order window is the typical number of days between the first and second order, by cohort and by first product. Most retention work should concentrate here, because a customer who reaches order two is far more likely to reach order three than a first-time buyer is to reach order two.

• A leading indicator. Stable or shrinking gaps in recent cohorts point to steady future revenue. Widening gaps predict a slowdown months before it shows in sales.

• Quality over raw speed. A moderate frequency among satisfied, long-lived customers beats a high frequency with fast churn. Frequency only compounds when customers stay.

• Model subscribers separately. Subscription orders follow a different pattern from one-off reorders, and blending them into a single lifetime value forecast distorts both.

Knowing when each cohort reorders also helps you plan stock and spend, because reorder volume by week stops being a guess.

A worked example from raw export to decision

Round numbers, as an illustration.

A pet food brand exports last year's orders: 12,000 orders from 6,000 unique customers. Store-wide purchase frequency is 2.0. Not bad, not great.

Flag first orders. For each customer, mark the earliest order in the window as order one, the next as order two, and so on. Calculate days between order one and order two for everyone with at least two orders.

Compare cohorts. Customers who first bought in January show a median of 35 days to second order. March customers show 50 days. June customers rarely come back at all within 90 days. That widening gap says something changed in acquisition, product or experience around spring.

Split by first product. Customers whose first order included the 12 kg dry food bag reorder in a median of 40 days. Those who first bought the 1 kg trial size reorder in a median of 22 days. The trial pack is the stronger opener even though its order value is lower.

What to watch for. Flattening curves in some cohorts, longer gaps after particular campaigns, and product combinations that shorten the time to the second order. This is the same raw logic that sits behind market basket analysis and cohort analysis when a tool runs them for you.

Five actions that improve purchase frequency, each tied to a data signal

1. Time reorder reminders to the observed gap. Use the median days between first and second order per key product. Send the nudge a few days before it. If coffee customers reorder at day 27, remind them at day 24. A flow timed to the store-wide average is early for some products and late for the rest.

2. Lead acquisition with the fast-reorder product. Put the budget behind the product with the shortest healthy cadence. If the trial kit produces earlier second orders than the big bundle, the trial kit is the better front door even at a lower first-order value.

3. Offer subscriptions after order two. Wait until you know a customer's natural cadence, then set the interval to match it. Pushing a subscription on the first visit, at an interval the customer has not yet lived, is how subscriptions end up paused within two months.

4. Build bundles from real repeat pairs. Look for products bought together across consecutive orders, not just in the same basket, and bundle those. Avoid pairing products that substitute for each other. The product bundling entry covers the difference.

5. Trigger win-back before the gap doubles. When a customer's time since last order approaches twice their usual gap, message them. A 30-day cycle means day 55 to 60, with a product-specific reminder rather than a blanket sale.

For agencies, each of these is a testable change: a timing shift, an offer, a lead product. Report the effect on median days to second order for the cohort that saw it.

The limits of the number on its own

Purchase frequency tells you when customers did not come back. It cannot tell you why.

• A drop in frequency for one cohort could be product quality, a shipping change, or a campaign that set the wrong expectation. The number alone will not say which.

• Pair it with post-purchase surveys, review text and support tickets for the cohorts where the gap widened.

• Check the acquisition channel. A cohort from one paid source or one influencer often shows a longer gap because the promise in the ad did not match the product. Segmentation by source next to the cohort table usually makes this visible in an afternoon.

How agencies should report purchase frequency

One blended purchase frequency on a slide does more harm than good.

Minimum reporting set:

• Purchase frequency for the last twelve months, split by acquisition month.

• Median days to second order for the last six to twelve cohorts.

• Separate lines for subscribers and non-subscribers, and for first-time and existing customers.

Then one slide that connects a movement in those numbers to a specific piece of work: the retimed reorder flow, the new bundle, the win-back sequence. Clients want to see which action changed which cohort, not a chart with a trend line.

Affinsy turns these splits, including which products follow which across orders, into segments and customer lists sized for a campaign, so the report starts from the client's data rather than from a template.

Next steps

The store-wide average gives you one number. The per-product and per-cohort splits tell you where to act.

This week:

1. Export the last twelve months of orders.

2. Calculate first-to-second order gaps by first product and by acquisition month.

3. Pick one product or cohort and run a timing test on its reorder reminder.

Repeat the export monthly so shifts show up while they are still small. If you would rather have the splits done for you, the 48-hour analysis returns the second-order window, the reorder cadence per product and the customers due this month.

FAQ

Is there a good purchase frequency number to aim for?

Not a universal one. Coffee or pet food brands might see eight to twelve orders a year from their most loyal customers, while a furniture store sees one or two over several years. Compare your own cohorts and products over time instead: is this March's cohort buying more often than last March's, and do customers who start with product A reorder sooner than those who start with product B?

How often should I recalculate it?

Quarterly for strategic decisions, and monthly for stores with short replenishment cycles such as skincare or supplements. Recalculate after a product launch, a major promotion, or a change in shipping or pricing.

Gross orders or net of returns?

Net of cancelled orders and clear returns, because a returned order is not a completed purchase. What matters most is consistency: pick a definition and apply it the same way every period.

How do multi-channel customers affect the calculation?

A customer who buys on the web store and on a marketplace can appear as two people if the identifiers differ, which lowers frequency artificially. Consolidate customer ids by email, phone or loyalty id where you can.

Can I improve purchase frequency without subscriptions or a loyalty program?

Yes. Reminders timed to observed gaps, bundles built from actual repeat pairs, pack sizes that match consumption, and win-backs sent before customers drift all move the number without a program behind them. Aligning communication with the real time between purchases is usually the fastest win.