Discounting a second order to someone who was already going to place it is the most expensive habit in retention work. The order still arrives. The winback flow still claims the revenue in its report. Margin is lower than it would have been if nobody had intervened, and the dashboard records a success.
Lifting a repeat purchase rate is mostly a matter of removing obstacles from the second order and approaching people while they are still receptive, rather than buying the order back with a coupon. Most useful ecommerce customer retention work is unglamorous: making it quick to reorder what someone already bought, reaching them inside the window when the decision is still open, and measuring by cohort so a real effect becomes visible at all. Incentives have a place. They rarely deserve to be the first lever pulled, because they are the only one that reduces the value of orders the store would have received anyway.
What Repeat Purchase Rate Actually Measures
Klaviyo defines repeat purchase rate as “the percentage of customers who have bought from you more than once within a specific period of time”, calculated by dividing returning customers by total customers and multiplying by 100. The same reference suggests that a good repeat purchase rate is typically around 20% to 30%, while noting the figure runs higher for affordable or perishable goods and lower for high-value categories such as technology or luxury.
Treat that range as orientation rather than a target. A store selling filters or supplements operates on a natural reorder cycle a furniture retailer does not have, and identical programme quality will produce very different numbers in each. A benchmark can suggest whether your figure is unusual for ecommerce as a whole. It cannot say whether it is unusual for your catalogue, which is the only comparison that should move a budget.
Definitions also differ between the systems on the desk. Shopify treats a returning customer as “a customer who placed an order, and whose order history already includes at least one order”. Its documentation adds a detail that quietly breaks a lot of monthly reporting: “The data in customer reports is based on the entire order history of the new customers in the report, not only the orders that were placed during the selected timeframe.” A repeat rate pulled for a single month is therefore not a month’s performance. It is the eventual behaviour of that month’s customers, much of which has not happened yet.
Measuring Ecommerce Customer Retention Without Fooling Yourself
Cohort reporting removes most of that distortion. Shopify’s customer cohort analysis groups buyers by first order date and follows each group forward, which lets January’s customers be compared with April’s at the same age rather than on the same calendar day. Any serious conversation about ecommerce customer retention needs that alignment, because almost every intervention takes weeks to surface.
Analytics platforms offer a parallel mechanism with an important limitation. GA4’s cohort exploration lets you set an inclusion criterion that adds a user to a cohort and a return criterion they must meet afterwards, using acquisition date, events, transactions or conversions. Google’s documentation then states that “Cohorts are based on the user’s device data only. User-ID is not considered when creating a cohort.”
That sentence should decide where the number comes from. Someone who buys on a phone and reorders on a laptop is two users to a device-based cohort and one customer to the order table. GA4 cohorts remain useful for understanding return visiting behaviour and the paths people take when they come back, but the commercial repeat purchase rate belongs to the ecommerce platform, where orders attach to customer records rather than browsers.
Even there the record has to be trustworthy. Guest checkout, a second email address, a work address on one order and a personal one on the next, and a marketplace channel that never writes back all split one buyer into several. Every split inflates the first-time customer count and depresses the repeat rate. Before judging any programme, check how often the same person appears twice, because that check often explains more of the gap than the marketing does.
Four measures usually earn their place in reporting, each answering a different question about the same population.
- Repeat purchase rate by cohort, not by month. A shift here often points at acquisition quality rather than at the retention programme, which is a very different problem to fund.
- Median time to second order. The median beats the average because a handful of very fast reorders can hide a long tail. It is also the number that says when outreach should happen, and it is frequently shorter than the schedule the email programme runs on.
- Conversion from second order to third. The step from one order to two is largely about whether the first experience was acceptable. The second transition often holds up far better, which argues for concentrating effort earlier.
- Revenue per cohort over a fixed horizon. Rate alone can be improved with cheap low-value reorders. Tracking what a cohort is worth at 90 or 180 days keeps discounting honest and connects the work to the customer lifetime value arithmetic that justified the budget.
The On-Site Levers That Decide the Second Order
Lifecycle messaging takes most of the ecommerce customer retention budget, but it can only invite someone back to a store that may or may not be ready for them. What happens after the click is often where the second order is lost, and those failures rarely appear in a campaign report.
Reordering Should Be Faster Than Searching
For consumables, replacement parts or anything bought in a variant, the fastest route to a repeat order is a path from the account area straight back to the exact item, in the exact size, with address and payment method already held. Where that path does not exist, a returning buyer has to repeat the original research, and some proportion will not bother or will land on a competitor’s product page during the attempt.
The trade-off sits at first purchase rather than second. Forcing account creation before checkout suppresses first-order conversion, which is why guest checkout remains the sensible default for most consumer stores. The workable compromise is to let people buy as guests and offer account creation once the order is confirmed, when the value of saving details is obvious and the risk of losing the sale has passed. B2B stores can reasonably take the opposite position, since account-specific pricing and reorder history are frequently why the buyer chose the store at all.
The Post-Purchase Experience Is a Lever, Not a Service Cost
Delivery accuracy, the clarity of tracking information and the difficulty of a return often determine the second order more decisively than any message sent afterwards. A late parcel with no explanation, or a return that took three emails, tends to produce a customer who never complains and simply does not come back. Those losses are invisible in retention reporting because nothing negative was ever recorded.
The order confirmation page is the one moment when attention is guaranteed and the buying decision has already been made. Treating it as a receipt wastes it. Used well, the post-purchase page can carry account creation, realistic delivery expectations, usage information that reduces returns, and a single relevant next item. Overloading it with offers tends to backfire, because the page also sets the tone for how the store will behave from here.
The Storefront Should Recognise Someone Who Has Bought Before
Returning customers arrive carrying context the store usually discards: a size, a preferred variant, a delivery address, a history of what did not work. Shopify describes customer segments as “dynamic, rule-based customer lists that let you group together customers who have similar characteristics”, built from purchase history and other attributes. Segments defined once can drive storefront treatment, messaging and offer eligibility from a single definition, which matters because the most common cause of contradictory retention campaigns is three systems each holding their own idea of who counts as lapsed.
Lifecycle Levers and the Order in Which They Usually Pay
Triggered messaging remains the efficient part of most lifecycle programmes, in the narrow sense that it produces disproportionate revenue for the volume sent. Omnisend reports that email automations accounted for 30% of revenue from 2% of sends in 2025, earning roughly 16x more per send than scheduled campaigns. That compares automated with broadcast email rather than proving any particular flow caused a purchase, and it is best read as evidence about relevance and timing.
Timing is where the largest avoidable losses sit. Klaviyo observes that for most brands the window for a second purchase is 30 to 60 days, yet many wait until 90 days to act with a winback flow, by which point momentum has faded. A store whose median time to second order is six weeks and whose first reminder goes out at twelve is not running a weak programme. It is running a well-built programme aimed at a moment that has already passed.
A sensible sequence therefore starts with messages that cost nothing per order and asks for margin only later: onboarding and usage content, a review request timed to actual delivery, a replenishment reminder set from the real consumption cycle, then reactivation, and only then a discount for people who have genuinely stopped responding. Reversing that order teaches the customer base to wait for the coupon, which is hard to undo. How these lifecycle email programmes are built matters less than whether each message has a reason to exist beyond filling a slot in a calendar.
Loyalty Programmes Are a Platform Decision First
Points schemes are often proposed as the answer to a weak repeat rate, and they are among the harder levers to reverse. What they cost depends heavily on the platform. Adobe Commerce includes reward points natively, and its documentation is explicit that “This is an exclusive feature that is available only in Adobe Commerce and is not available in Magento Open Source”. Points can be earned for purchases, registration, newsletter signup, referrals and approved product reviews.
On Shopify or WooCommerce the same capability arrives as a third-party application, bringing a subscription fee, another store of customer balances, and a dependency at checkout. None of that is disqualifying, but it changes the calculation. A points programme also creates a liability: unredeemed balances are a promise margin has to absorb whenever customers choose to use them. That is a reasonable cost when the scheme genuinely changes behaviour and an avoidable one when it mostly rewards purchases that were already going to happen.
Some Churn Is a Billing Failure Rather Than a Decision
Stores running subscription or auto-replenishment models lose a share of customers to expired and temporarily declined cards rather than to dissatisfaction. WooCommerce documents a retry system for exactly this, describing it as something that “can help recover revenue otherwise lost due to a customer’s payment method being temporarily declined”. Its default rules apply five retry attempts across roughly seven days, with customer emails on some attempts, after which the renewal order is marked Failed.
Confirming that this recovery path is configured, and that its emails actually reach people, is among the cheapest retention work available. It needs no creative, competes with nothing for attention, and the customers it saves had not chosen to leave. Teams that have never separated involuntary churn from deliberate cancellation are usually applying expensive tools to the wrong problem.
It helps to know how much of total demand this population represents. Salesforce reported that the share of orders from repeat buyers grew 5% year over year, and expected more than one in three online peak-season orders to come from repeat buyers. That indicates where seasonal revenue originates. It does not establish that any specific loyalty tactic produced the shift, and it should not be used to argue for one.
Comparing the Main Retention Levers
The levers differ less in effectiveness than in what they cost to run, what they risk, and how long they take to produce a readable signal. Those three properties usually decide which one a particular business should start with.
| Lever | Best suited to | Ongoing cost | Main risk | Time to a readable signal |
| Faster reorder and account experience | Consumables, parts, variant-heavy and B2B catalogues | Development once, then low | Pushing account creation too early and losing first orders | One full reorder cycle |
| Post-purchase and onboarding messages | Almost any store, particularly where returns or misuse are common | Low per order | Adding volume without adding usefulness | Four to eight weeks |
| Replenishment and timed reminders | Products with a predictable consumption cycle | Low | Timing set by convention rather than by measured cycle | One to two cycles |
| Discounted winback | Genuinely lapsed customers who no longer respond | Direct margin on every order it touches | Subsidising purchases that would have happened anyway | Fast, but easy to misread without a holdout |
| Loyalty or points programme | Frequent, lower-value purchases in competitive categories | Platform or app fee plus accrued point liability | Rewarding existing behaviour; hard to withdraw later | Two quarters or more |
| Subscription and billing recovery | Any store with recurring orders | Minimal once configured | Assuming it works without testing the emails | One billing cycle |
The risk column is what separates them in practice. Cheaper levers mostly put effort at risk, while the expensive ones put margin and customer expectations at risk, and expectations are far harder to reset than a schedule. That asymmetry argues for exhausting the structural work before committing to anything funded out of every future order.
Key takeaway: a retention programme should be judged on the orders it created, not the orders it touched. Any lever that attaches itself to purchases already in motion will look successful in its own reporting, which is why a holdout group and a cohort view are worth more than another tactic.
A 90-Day Sequence That Produces Evidence Rather Than Activity
Retention work stalls because everything appears urgent and nothing is measurable for weeks. Ordering the work by what the next step depends on resolves most of that.
- Repair the customer record. Establish how many buyers are duplicated across guest orders, multiple addresses and channels that do not write back. Until this is known, every later figure carries an unquantified error.
- Build the cohort baseline. Pull repeat rate and revenue by first-order cohort for at least twelve months, so recent cohorts are read against older ones at the same age. This usually exposes at least one assumption the team held without evidence.
- Measure the real second-order window. Calculate the median interval between first and second order by product category. Programme timing should follow that number rather than a template, and categories with different cycles should not share a schedule.
- Fix the undiscounted paths first. Reorder routes, account creation after checkout, delivery and returns communication, and the timing of existing flows. These cost development time rather than margin, so a modest effect still pays.
- Introduce one incentive with a holdout. Hold back a random share of the eligible audience and compare cohort revenue, not response rate. Without a control, an incentive that subsidises existing demand looks identical to one that creates new demand.
- Review by cohort age, then decide what to keep. Be willing to retire a flow that only reallocated orders. A programme that never removes anything accumulates messages nobody can justify.
The sequence assumes someone owns it. In practice the work sits between marketing, development and operations, and it stalls when each assumes another has it. WD Market’s published work with RIPO International covered checkout flow optimisation, B2C and B2B journey separation and catalogue restructuring within one programme, which is the kind of consolidation that lets storefront and lifecycle changes be judged together rather than in separate reports. Internal expert input required: add a verified WD Market example of a reorder or account-area change and the operational effect it produced, with no invented figures.
Where Retention Programmes Commonly Go Wrong
Most ecommerce customer retention failures are not failures of effort. They are failures of attribution, timing or ownership, and each carries a predictable consequence.
- Crediting every touched order to the flow that touched it. Lifecycle platforms attribute generously by design, and the totals often exceed what the store actually gained. Budget then moves toward whichever tool reports most confidently rather than whichever intervention changed behaviour.
- Running retention work while first-order conversion is broken. A store losing buyers at checkout is paying to send people back into the same obstacle. A conversion audit usually returns more here than any lifecycle investment, because the second order cannot be won on a journey that fails the first.
- Setting reminder timing by convention. A cadence copied from a template will miss most categories, and the error is silent: the message still sends, the report still shows a modest response, and nobody sees the orders that went elsewhere in week seven.
- Treating the loyalty programme as the strategy. Points rarely compensate for a slow reorder path or an unclear returns policy. Launched over unresolved friction, a scheme adds a recurring cost to a problem it was never able to address.
- No single owner for the second order. Marketing is often measured on acquisition and campaign performance, development on delivery of requests, and customer service on ticket resolution. Responsibility for what happens between the first and second purchase may then belong to nobody in particular, and work with no owner rarely survives a busy quarter.
Choosing the Lever That Fits Your Catalogue
Choose replenishment timing and reorder speed when the catalogue is consumable and the consumption cycle can be estimated from order history. The lever is cheap, the signal arrives within a cycle or two, and being right about timing usually outperforms being generous about price. Choose account and reorder work when a large share of revenue comes from B2B buyers or variant-heavy products, where the cost of repeating a decision is highest.
Choose post-purchase and product education when returns are frequent or the product needs correct use to satisfy. In those categories the second order is often decided by whether the first one worked, so content that prevents disappointment may do more than any offer. For considered, infrequent purchases such as furniture or high-value equipment, accept that the natural cycle is long and shift attention to accessories, consumables, service and referral. Pushing a repeat purchase against a category’s own rhythm tends to produce discounting and little else.
Choose billing recovery first if any part of the business runs on recurring orders, since it is the only lever that recovers customers who never decided to leave. Approach a points programme last, once reorder paths, timing and post-purchase communication have been addressed, because it is the hardest commitment to withdraw after customers have accrued a balance.
There are also situations where ecommerce customer retention should not be the priority. A store with a very small customer base has too little data for cohort analysis to say anything reliable, and one whose acquisition is dominated by a single promotion may be watching a cohort effect rather than a programme effect. In both cases, more customer insight work is usually a better investment than more messaging.
Deciding Where Your Second Order Is Being Lost
Very few established stores are short of retention tactics. What they more often lack is a repeat rate measured in a way that could actually register an improvement, and a reorder path clear enough that a satisfied customer never has to think about it. Both of those come before any decision that costs margin.
Cohorts rather than calendar months, a customer record that does not split one buyer into three, timing taken from the observed second-order window, and a holdout whenever an incentive is introduced: those four conditions turn ecommerce customer retention from an assumption into something a business can act on. The levers themselves are well understood. The order they are applied in, and the evidence used to judge them, is where the money is made or lost.
From a Retention Guess to a Measurable Repeat-Purchase Plan
If your repeat purchase rate has been flat and nobody can say confidently why, the useful first step is a diagnosis rather than another flow. WD Market’s CRO and growth support includes a cohort baseline built from your platform’s own order data, a review of how duplicated customer records distort the figure, an assessment of the reorder and post-purchase paths on your storefront, and a prioritised plan separating structural fixes from anything that costs margin.
Tell us what your current repeat rate is and how it is calculated, and we will tell you what it would take to make it move. You can reach the team through our contact page, and we publish ongoing ecommerce conversion and retention notes on WD Market’s LinkedIn page.
Frequently Asked Questions
How do I calculate repeat purchase rate correctly?
Divide the number of customers with more than one order by the total number of customers, then multiply by 100. The difficulty is choosing the population. Applying the formula to everyone who ordered last month mixes people who have had a year to return with those who have had a fortnight. Group buyers by their first order date instead, then read each group at the same age, and take the figures from the ecommerce platform rather than from a web analytics report.
Is a 20% repeat purchase rate good?
It depends almost entirely on what you sell. Published guidance places a typical range at roughly 20% to 30%, but that blends categories with very different buying rhythms. A supplement brand and a furniture retailer cannot be held to the same figure. The more informative comparison is against your own earlier cohorts at the same age, and against the natural reorder cycle of your products. A stable figure in a long-cycle category may be healthier than a higher one that is drifting down.
Can GA4 tell me my repeat purchase rate?
Not reliably, because GA4 cohorts are built from device data and do not consider User-ID. One person shopping on a phone and later on a laptop can appear as two users, which understates repeat behaviour. GA4 suits questions about how returning visitors navigate and which channels bring them back. For the commercial figure, use the ecommerce platform, where orders are attached to customer records. Reconciling the two is useful; treating them as interchangeable is not.
When should a winback email be sent?
Base it on your measured median interval between first and second order rather than a standard schedule. Klaviyo notes that the second-purchase window for most brands falls between 30 and 60 days, while many programmes wait until 90 days, which is often too late to matter. Calculate the interval separately for categories with different consumption cycles. A single store-wide cadence will usually be early for one part of the catalogue and late for another.
Do loyalty programmes actually increase retention?
They can, though the effect is easy to overstate because points usually go to customers who were already buying. Before launching one, check that the reorder path, delivery communication and returns process are working, since a scheme built over that friction adds cost without removing the cause. Consider the liability as well: unredeemed points are a commitment margin has to absorb later, and withdrawing a programme once balances exist is considerably harder than starting one.
Who should own ecommerce customer retention internally?
It works best when one person owns the outcome and can commission work across functions, because the levers sit in marketing, development and operations simultaneously. Splitting ownership by channel tends to produce a well-run email programme pointing at a storefront nobody has improved. Where no internal role can hold that scope, an external team working across analytics, UX and development is one way to keep storefront and lifecycle changes accountable to the same measure.