Customer retention strategy for ecommerce: count it in margin, not in opens

You know the claim: a small gain in retention produces a very large gain in profit. It appears in almost every retention article. One vendor's roundup of retention statistics traces it to a 1990 Harvard Business Review article by Reichheld and Sasser¹. Industry blogs, including one from an analyst firm, repeat it².

Notice who is doing the repeating. The versions in circulation are second-hand. The same roundup shows that the figures quoted today are not the ones in the 1990 article: the range changed as the claim was retold¹.

We do not use its numbers, for three reasons:

• A statistic is only as good as its original source. We quote retention figures only when we can point to that source.
• A decades-old figure about other businesses cannot tell you what a retention gain is worth to your margin.
• Your own cohorts can answer the question directly. That is the better evidence.

So here is the replacement: do not ask what retention is worth on average. Ask what it is worth to you.

Three measures do most of the work.

Cohort analysis. Group customers by a shared trait, usually the month of their first purchase, and follow their behaviour over time. GA4 defines a cohort in these terms³, and its retention overview report is built around cohorts of new users⁴. GA4 tracks users on your site, though. For repeat orders, your shop's customer and order data is the better base.

Repeat purchase rate. The share of customers who bought more than once in a period: customers with two or more orders divided by all customers. Beware the label. An analytics vendor points out that "repeat purchase rate" is defined in different ways, by customers or by orders, and the same data gives different answers⁵. Pick one definition and write it down. The detail is in repeat purchase rate formula.

Churn. The rate at which you lose customers over a period. Without subscriptions there is no cancellation date, so you first have to decide when a customer counts as gone. Shopify, a platform vendor, suggests a window without purchases that matches the buying cycle of the product⁶. A shop selling coffee beans and a shop selling sofas need different windows.

The same goes for the retention rate formula. An analytics vendor argues that the version taken from subscription businesses fails in shops without contracts, and that ecommerce needs cohort-based measurement⁷. Our view is the same, and it is why we start with cohorts.

Retention work is a set of tools. None is the strategy.

Email automation flows are sequences sent automatically when a customer does something or meets a condition. Typical ones are welcome, abandoned cart, post-purchase, win-back and replenishment emails. For the flows that matter and how to tell if they pay, see email marketing for ecommerce.

SMS reaches people differently and comes with its own consent rules. Email vs SMS marketing for ecommerce covers which message goes where, and SMS marketing platforms for ecommerce covers the tool category.

Replenishment. For products that run out, a reminder timed to the usage cycle is the obvious message. It works only if you know the cycle, which your cohorts show.

Loyalty programme. It rewards repeat purchases with points, tiers or benefits. Customers who already buy often are the likeliest to sign up, so its effect has to be measured against a comparable group of non-members. An academic study of one loyalty programme and repeat-purchase patterns reports only a modest change in behaviour after customers joined⁸. So do not assume that members buy more because of the programme. Read loyalty programme vs discount repeat customers margin and see loyalty and subscription platforms for ecommerce.

Win-back. Messages to lapsed customers. The question is when to try again and when to stop; see win-back email campaign ecommerce.

Before you write a subject line, two things set the limits: whether you may message someone, and whether the message reaches them.

On consent, we do not give legal advice, and a blog post is the wrong place to summarise the law. Before any flow goes live, ask your lawyer which messages you may send to whom in each market you sell to. Then read SMS marketing for ecommerce, which puts consent first.

On deliverability, meaning whether your email reaches the inbox rather than simply being accepted by the receiving server, Google's sender guidelines require every sender to authenticate mail with SPF or DKIM and to keep the spam complaint rate low. Senders of 5,000 or more messages a day to personal Gmail accounts also need DMARC and one-click unsubscribe in marketing messages⁹. In plain terms: prove the mail is yours, make leaving easy, and do not annoy people. Gmail Yahoo bulk sender requirements has the checklist, and Google Postmaster Tools covers monitoring.

Opens are a separate trap. Apple's description says Apple Mail Privacy Protection (MPP) downloads remote email content in the background by default, even if the email is never opened, and routes it through two relays so that no single party sees both the recipient's IP address and the content¹⁰. For senders, an open signal no longer proves a person read the email. If your reporting leans on open rate, a drop may mean nothing or a rise may mean nothing. See Apple Mail Privacy Protection open rate for what to measure instead.

Here is the central problem. Some of the customers who get your flow would have bought again anyway. Their order lands in the flow's revenue report and gets the credit.

The fix is a holdout test: keep a random group of customers out of the flow, and compare repeat orders and margin between the two groups. The difference is the part the flow actually added.

Example (illustrative). A win-back flow is planned for 10,000 lapsed customers. A random 1,000 are held back and the other 9,000 get the emails. The flow's dashboard credits it with 300 orders. The held-back group placed 25 orders per 1,000 customers with no email, so about 225 of those 300 orders would likely have happened anyway. The flow added closer to 75, and margin on those comes after discounts.

A holdout needs enough customers to see a difference, and some flows cannot afford it. Then name the weaker method you use instead and its bias. Our view: the more a flow costs in discount, the more it needs a holdout.

Retention is not a vanity metric. It lands in the P&L in three places: revenue from repeat orders, discounts you gave to win them, and the cost of the messages and programme. What is left is contribution margin from repeat customers.

LTV (customer lifetime value) is what a customer is worth over time, and there is more than one way to calculate it. Our convention is cohort-based LTV, counted after discounts and returns, using contribution margin. It connects retention to the marketing decisions the CFO cares about: payback on acquisition and the cash you need while you wait.

For ZFD: we report retention in repeat purchases and margin, never in opens. We treat flows, loyalty and win-back as tools to test, and we start with the question "who would have come back anyway?".