A customer who stops buying rarely tells anyone. Usually they stop clicking first, then vanish from your reports. Customer retention measures the share of customers you keep from one period to the next. Email is one of the first places where this drop-off shows. But the emails have to arrive: a list that degrades silently skews the calculation before it even starts.
Customer retention definition and rate formula
The retention rate is calculated over a set period (a month or a year) with this formula: (customers at end of period − new customers acquired during the period) / customers at start of period × 100. New customers are subtracted from the numerator, otherwise strong acquisition would hide the departures. The period matters: 88% per month and 88% per year describe very different situations.
Worked example, with invented figures for demonstration: a store has 1,000 customers on January 1 and 1,080 on December 31, of which 200 were acquired during the year. The retention rate is (1,080 − 200) / 1,000 × 100 = 88%. Churn (the attrition rate) is its complement: lost customers / customers at start of period × 100, so 120 / 1,000 = 12%.
Retention or loyalty?
Retention is a count: it answers “how many stayed?”. Customer loyalty covers the actions that give people a reason to stay, from points programs to an ongoing relationship, which is what retention marketing covers. One is a metric, the other a set of levers. The levers have their own article, so this one measures first, then moves on to the first scenarios to set up. A satisfaction survey (NPS) and a points program belong to those actions, as does an abandoned cart reminder. The dedicated article covers them.
The customer life cycle gives a useful reference point: an overall rate blends customers of 6 days with customers of 6 years.
Why retention weighs on profits
In a 2006 publication, Bain & Company puts a figure on the stakes: increasing retention by as little as 5% can raise profits by up to 95%. Bain presents this result as a theory dating from 2006: it gives an order of magnitude, with no value as an average or as a promise.
Two indicators connect this finding to the budget. Customer lifetime value (CLV) estimates what a customer brings in for as long as they stay; customer acquisition cost (CAC) adds up what it took to win them. If a segment’s CLV no longer covers its CAC, two options open up: lower the cost of winning new leads, or lengthen the time that already-acquired customers stay active.
Why industry benchmarks mislead
Roundups of “average retention rate by industry” rarely rely on a shared method. One company counts a customer as lost after 30 days without a purchase, another after 12 months. A SaaS subscription signals departure through a cancellation, while an online store has no contract to mark the break. Comparing 88% at one with 60% at the other means nothing as long as the definition of “active customer” differs.
The only reliable point of comparison is your own history, always calculated the same way, by cohort: all the customers acquired in March, followed month after month. A cohort that loses 40% of its members within the first two months can point to an onboarding flaw that the annual average buries. Fictional example: 500 customers acquired in March, 410 still active in April, 330 in May, 300 in June. The 3-month retention rate comes out at 60%. The useful reading lies elsewhere: 90 customers left in the first month, 80 in the second, 30 in the third. The drop-off concentrates in the first 8 weeks, which probably points to onboarding and the first order rather than the established relationship. An annual average of 88% would show none of this, since 2-year customers, far more stable, dilute the recent departures. For the rate to be useful, fix the definition of an active customer once and for all (a purchase in the last 12 months, for example) and write it into the dashboard. Changing the definition mid-year makes the curve unreadable.
Reading retention in an email program
The retention rate arrives late: it records the departure. Email provides earlier signals, provided you know which ones to trust.
The open rate is the first to discard as proof of engagement. With Mail Privacy Protection, Apple prevents the sender from knowing whether a message was opened, so messages nobody read can show up as opened (Apple’s documentation). On a list where Apple Mail weighs heavily (iPhone, iPad, Mac), a rise in opens can sit alongside a drop in real interest.
Rely instead on:
- the click rate relative to emails sent (CTOR, which is based on opens, has the same bias);
- the share of contacts with no click or purchase in the last 90 or 180 days, set according to your purchase rhythm;
- unsubscribes and spam complaints, which are declared departures;
- revenue per email sent, calculated by segment and by month, set against the other marketing indicators on the dashboard.
A segment whose clicks erode for 3 months in a row deserves an alert before the retention rate drops. Set the alert threshold yourself, for example 2 consecutive months without a click for a customer who used to buy every month, then measure how many of these customers still buy anyway: that false-positive rate tells you whether the signal is worth tracking. Segmentation by last purchase date and by frequency lets you see it without waiting for the annual review.
Inactive contacts cost more than their silence
An inactive contact stays in your sends, and this is where retention meets deliverability. Messages sent to addresses that no longer exist cause hard bounces and push the bounce rate up, which tells the mailbox provider that the list is poorly maintained. Those sent to merely inactive contacts inflate volume with no response. Sender reputation can suffer on the messages that follow.
On the complaints side, Google asks bulk senders (about 5,000 messages or more per day to personal Gmail accounts) to keep their reported spam rate below 0.1% and avoid reaching 0.3%. Above 0.3%, the sender is no longer eligible for Google’s mitigation, and becomes eligible again only after seven consecutive days under that threshold. Chasing inactive contacts endlessly eats into this margin as soon as some of them click “report spam”. The details of the rules are in our article on Gmail’s sender requirements.
A list decays even when nobody touches it
Of the 126 million addresses CaptainVerify verified in 2025, only 57% were confirmed valid. The company also estimates that a list loses about 23% of its addresses every year, as detailed in its analysis of email address value. Job changes and closed domains leave mailboxes abandoned. Some of your “lost” customers may never have left the brand: they simply no longer receive anything. The calculation is short: 20,000 addresses × 23% annual loss = 4,600 contacts who may no longer be able to reply, buy or unsubscribe within 12 months. They still sit in your denominator.
A customer who left for a competitor calls for a commercial response; an address that has become invalid calls for cleaning. A simple signal separates the two: a departed customer usually clicked or bought before leaving, while a dead address bounces or never reacted. Cross-checking the list of “lost” customers against the last 12 months of bounces is often enough to split the two groups before writing to anyone. To sort them out, a verification test on a sample of the list, for example 2,000 contacts drawn at random, shows the share of invalid addresses before you commit to a win-back campaign. CaptainVerify’s list verification is built for this check.
Three flows to launch first
There is no need to launch 10 scenarios at once. The first is the welcome email, where the onboarding of new subscribers is decided; the mistakes in this message are costly because it lands early, when the subscriber still remembers signing up. The second follows the purchase: confirmation and usage tips, then a review request. The third is aimed at those who are fading, with a re-engagement email that comes before a removal rule: with no reaction after 2 or 3 reminders, the contact leaves the regular sends.
These flows fall under automation. Applying the removal rule to a list full of dead addresses fixes nothing: cleaning comes before automation.
Testing and pacing your sends
An A/B test is only worth running on one variable at a time and on a sample large enough to separate the versions: test the subject line or the call-to-action button without mixing the two. The winning criterion must be the click or the purchase, never the open, for the reason given above.
Frequency is set the same way: raise it in steps on one segment, watching unsubscribes and complaints. Each segment has its own rhythm: a monthly buyer tolerates a weekly send, a dormant customer far less. Personalization by segment comes next.
Mistakes that distort the reading
Six mistakes come up most often:
- confusing retention with repurchase: a customer who stays subscribed without buying is “retained” by the formula yet generates no revenue;
- calculating a single average: an overall 88% can hide, for example, 95% among customers of more than 2 years and 70% among new ones;
- measuring engagement on opens, which Mail Privacy Protection makes unreliable;
- counting as lost customers whose address has simply become unreachable;
- chasing inactive contacts without cleaning the list first, which adds bounces to an interest problem;
- restarting all sends at once after a pause, which mixes active and dormant contacts.
Restart in steps instead, beginning with the segment that clicked recently.
