Stop Paying Twice for the Same Order: Rethinking Ecommerce Customer Retention

Real ecommerce customer retention is a margin outcome, not an email channel. A customer who repurchases only because you sent a 20% code has not been retained, they have been rented. The brands that compound do three things: they measure retention at cohort level, they separate incremental repeat orders from spontaneous ones, and they treat discount depth as a cost of goods.

• Retention rate measures repeat behavior and says nothing about whether the repeat was profitable.
• Cohort analysis exposes decay that a blended customer retention rate hides.
• Discount driven repeat purchase rate inflates order volume and deflates contribution margin.
• Customer lifetime value is only usable when it is net of discounts, returns and shipping.
• Holdout tests are the only way to prove that a retention program produced orders you would not have received anyway.

Ecommerce Customer Retention Is a Margin Metric, Not an Engagement Metric

Judge ecommerce customer retention on contribution margin per cohort, never on open rates or list growth. The arithmetic settles the argument faster than any framework. Take a brand with an $80 average order value and a 60% gross margin: every order carries $48 of gross profit before marketing costs.

Now apply a 20% sitewide code to lift repeat orders. The discount removes $16, so each repeat order now carries $32. Run the numbers on a base of 100 existing customers:

• Before the campaign: 24 repeat orders at $48 = $1,152 of gross profit.
• After the campaign: 30 repeat orders at $32 = $960 of gross profit.

The campaign raised the repeat purchase rate by 6 points and destroyed $192 of gross profit per 100 customers. Every retention dashboard in the market would have logged that as a win. This is the structural failure of most retention marketing: the discount is paid to all 30 buyers, while only 6 of them were incremental. The other 24 were going to buy regardless, and you funded their loyalty out of your own margin.

The correct threshold is simple. A promotional retention play only creates value when the margin on the incremental orders exceeds the margin surrendered on the orders you would have won anyway. At a 20% discount and a 60% gross margin, you need roughly a 50% lift in repeat orders just to break even. Very few campaigns clear that bar, and almost none are measured against it. Churn rate, meanwhile, keeps falling on paper while the P&L quietly gets worse.

Your Blended Customer Retention Rate Is Hiding a Cohort Problem

A single blended customer retention rate is an average of populations that behave nothing alike, and averaging them destroys the signal you need. Two cohorts acquired in the same month through different channels can post the same 12 month retention number while following completely opposite curves.

Illustrative model, not observed client data. The numbers below are chosen to show the shape of the problem, and every figure in this article can be recalculated from the assumptions stated.

The discount led cohort looks better for a quarter, then collapses. The search led cohort starts slower and is worth almost two and a half times more after a year. A blended number reports both as roughly 12% and tells you nothing actionable. Cohort analysis is what turns retention from a vanity line into a budgeting input, because it lets you price acquisition by the profit a channel actually delivers over 12 months rather than by first order revenue.

Running this properly requires two things most ecommerce stacks lack: clean first-party data that survives consent loss and platform gaps, and a measurement setup built around incrementality rather than last click attribution. If you are not sure your customer database can support cohort work at all, that is what a CRM audit establishes before you model anything. Without both, cohort curves are reconstructions rather than observations, and every conclusion you draw from them inherits the error.

Most Ecommerce Customer Retention Programs Pay Customers Who Would Have Come Back Anyway

The single largest waste in ecommerce customer retention is subsidising behavior that was already going to happen. A win back flow sent to lapsed buyers will always report revenue, because a share of those buyers were returning on their own schedule and the email simply captured the last click before checkout.

The fix is a holdout. Withhold the program from a randomly selected 10% of the eligible audience, run it for a full purchase cycle, then compare revenue per eligible customer between the two groups. The difference is the program’s real contribution. Everything above that line is attribution collecting credit for demand that already existed.

Three programs deserve this test before anything else, because they are the ones that reliably overstate themselves:

• Post purchase email marketing flows. High reported revenue, high natural repurchase overlap. Typically the most inflated line in the stack.
• The points based loyalty program. Rewards frequency you already had. Its incremental effect usually sits in the tail of low frequency buyers, not in the top decile it pays most.
• Blanket reactivation discounts. Sent to everyone inactive for 90 days, including buyers with a natural 120 day cycle who were never lost.

Run three holdouts over one quarter and you will usually find that reported retention revenue overstates incremental retention revenue by a wide margin. That gap is not a measurement inconvenience. It is budget you can redeploy.

The Second Order Is the Only Retention Milestone That Compounds

Concentrate on the move from first to second order, because that is where the probability curve breaks. Across ecommerce categories, the likelihood of a third purchase given a second is far higher than the likelihood of a second given a first. The second order is a qualification event: it converts a buyer into a customer with a demonstrated cycle you can model.

Two numbers matter more than the aggregate retention figure:

• Second order rate at 90 days. The share of a cohort placing a second order within 90 days of the first. This is your earliest reliable predictor of cohort value.
• Median time to second order. Sets the cadence of your lifecycle program. A brand with a 45 day median and a 30 day email cadence is talking to people who are not ready, and training them to ignore you.

Most brands get this backwards by pushing average order value on the first purchase through bundles and thresholds. Raising first order average order value at the cost of second order rate is a bad trade in almost every category, because you capture margin once instead of building a cycle. Where the category allows it, a subscription model solves the problem structurally by removing the repurchase decision entirely, which is why replenishment categories post retention curves the rest of ecommerce cannot match.

Ecommerce Customer Retention Breaks When Acquisition Chooses the Wrong Customers

Retention is largely decided before the first order, which is why ecommerce customer retention cannot be fixed downstream by a lifecycle team working alone. If acquisition is optimised on first purchase volume at the lowest possible customer acquisition cost, it will systematically buy the cheapest customers available, and the cheapest customers are cheap because their expected value is low.

The correction is to move the acquisition target from cost per first order to payback on 12 month contribution margin. Carrying over the two cohorts from the illustrative model above, that single change reprices every channel in the mix:

• A channel at $28 customer acquisition cost delivering $61 of 12 month gross profit returns 2.2x.
• A channel at $52 delivering $149 returns 2.9x, despite an 86% higher acquisition cost.

Judged on cost per acquisition, the first channel wins and gets scaled. Judged on customer lifetime value, the second wins and the first is quietly draining the business. This is the decision Primelis handles daily across 300+ worldwide brands and more than $1 billion of media managed each year: the retention problem is usually a media allocation problem wearing a CRM costume.

Where to Take Your Ecommerce Customer Retention Next

Ecommerce customer retention is a margin question answered with cohorts and holdouts, not an engagement question answered with sends. Discounted repeat orders can raise every reported metric while shrinking gross profit, blended retention rates conceal cohorts that behave in opposite directions, and acquisition targets set on cost per order will keep buying customers who were never going to stay.

Start with the least expensive diagnostic available. Pull your last four acquisition cohorts, plot second order rate at 90 days and gross profit per customer at month 12, then hold out 10% of one lifecycle flow for a full purchase cycle. Two weeks of work will tell you whether your retention program is creating orders or paying for them. Bring the output to a 30 minute cohort review with our team of 200+ specialists, built on 15 years of running this analysis, and leave with a ranked list of what to cut and what to fund.