Repeat rate slips a few points, or subscription churn ticks up, and within a week you have a plan on your desk.
More email flows. Add SMS. A loyalty program gifting schedule. A win-back campaign. Maybe a refresh of the subscription experience. The plan is competent, the team is motivated, and the whole thing will take many months and a meaningful budget.
I’d pause before you green light that ecommerce retention plan. Not because lifecycle work is bad — it’s genuinely valuable when the problem lives in the marketing touchpoints. But in a meaningful share of the cases I see, it doesn’t, and the retention plan on your desk is a treatment written before anyone did a proper diagnosis.
Retention is a downstream metric. A change in it often originates somewhere upstream.
A retention number is an average of stories that have nothing to do with each other
A 35% repeat rate, or 22% annual churn, is one number summarizing a lot of very different behaviors.
It contains customers who never intended to buy twice. Customers who tried the product and didn’t care for it. Customers who liked it but couldn’t justify the price. Customers who liked it and left for a competitor anyway. Customers who liked it so much they stocked up and don’t need more yet. Customers whose lives changed in ways that have nothing to do with you.
It can also contain customers who never made a decision to leave.
Recurly’s July 2026 network data puts median annual churn for ecommerce subscription businesses at 4.25%, of which 1.38 percentage points is involuntary: expired cards, bank declines, billing limits and other payment failures. Roughly a third of measured churn in that category comes from customers who did not actively choose to cancel.
That distinction matters. A meaningful part of the retention number your team is preparing to address through marketing may actually be a payments-operations problem. A better email flow and better payment recovery are not substitutes for one another.
This is the broader issue with aggregate retention metrics. The number identifies a business outcome. It doesn’t tell you which of the underlying customer behaviors produced it.
The dashboard tells you where the change occurred, not why
The founder’s starting question is usually straightforward: Retention fell. Why?
That’s the right place to begin. The CEO doesn’t need to open the cohort table and diagnose the answer personally. But the team’s first move should be to decompose the headline number before prescribing a retention program.
Did retention deteriorate across the customer base, or only among newer cohorts? Did the change begin with customers acquired from one channel, one promotion or one hero product? Is first-to-second purchase deteriorating while later repeat remains stable? Did subscription churn rise because more customers cancelled deliberately, or because payment failures increased?
Those cuts tell you where the problem entered the system.
If older cohorts are behaving normally and the deterioration begins only with customers acquired after a new promotional offer, the investigation should start with acquisition. If the decline is concentrated between the first and second purchase across virtually every source, product experience and value become more plausible explanations. If renewal behavior appears stable except for a spike in failed payments, there is little reason to begin with lifecycle strategy.
Marketing dashboards are organized by function — acquisition, retention, brand health, lifecycle — not by cause. So a number below target lands in the retention section and becomes the retention team’s problem. Your reporting structure ends up writing the diagnosis, even when the cause sits somewhere else entirely.
So build the habit the other way round: use the dashboard to locate the change, then work upstream until you know what produced it. The dashboard tells you where. It will never tell you why.
Five things that masquerade as retention problems
Five patterns account for most of the misdiagnoses I’ve seen.
Payment failure. A customer whose card expired or whose bank declined a renewal never made a retention decision. Before you redesign lifecycle, separate voluntary cancellation from involuntary churn and find out whether payment recovery is doing its job.
Your acquisition mix changed. The customers you’re bringing in now aren’t the customers who used to repeat. It shows up constantly in brands that scale paid into broader audiences or launch a more aggressive first-purchase offer. The new customers convert, so acquisition looks healthy, but they relate to the product differently than your original base. Retention fell because the input changed. The lifecycle program is working exactly as it was.
The product doesn’t hold its promise after the first purchase. The first transaction is discovery. The second requires confirmation that the value was real. Optimize the acquisition proposition hard for conversion while the product experience doesn’t sustain the promise, and retention is simply where that gap finally shows up.
The cadence doesn’t match how people use the product. The customer likes it and fully intends to buy again — you’re just asking them to buy too much, too soon. A 30-day subscription on something used every 60 days will generate cancellations. So will a pack size that pushes replenishment past the repeat window you’re measuring. That’s consumption and assortment architecture, not communications.
The category moved. A new entrant changed the value equation. Your customer hasn’t decided you’re worse; they’ve decided the alternative is enough better at the price difference. Look here first in brands that haven’t revisited their competitive set or their brand strategy in two or three years.
All five push the same headline number down. All five need completely different fixes. That’s the entire reason to diagnose the cause before you hand the problem to a function.
Nine months of good work on the wrong variable
Consider a composite based on a pattern I’ve seen more than once.
An $18M home-care subscription business sold a starter system followed by regular shipments of concentrated cleaning refills. Monthly churn sat at 13%, with the sharpest drop between the second and fourth shipments. Acquisition was bringing customers in cheaply on a heavily discounted starter offer, and the retention team had spent most of a year on onboarding, post-purchase education, subscription communications and loyalty benefits. They proposed a stock up and save offer and more loyalty program benefits to win back lost subscribers.
Nine months of genuinely good work moved churn from 13% to 11%. They were proud of it, and they were right to be — the customer experience really was better. The business result was not.
Then we spoke with roughly thirty recent cancellers and compared what they told us with the cohort data.
Two issues emerged quickly. First, many households were using the concentrates at roughly half the rate assumed by the subscription model. By the time the next shipment arrived, they still had unopened product from the prior one. Cancellation was a rational response to accumulating inventory.
Second, a substantial group had viewed the discounted starter offer as a way to try the system rather than as the beginning of an ongoing subscription. The metric classified them as churned subscribers; from the customer’s perspective, they had completed the trial transaction the acquisition offer encouraged.
Those are two different upstream problems sitting inside the same retention number: a usage-rate problem and an acquisition-intent problem.
The fixes were correspondingly different, and neither was a lifecycle fix. The business moved the default replenishment cadence out and built flexible shipment intervals for lighter-use households. Then it rebuilt the acquisition offer so the economics no longer depended on converting trial-minded customers into subscribers they never intended to become.
The cadence change alone took monthly churn from 11% to 7%, and cohorts acquired under the revised offer retained better than every promotional cohort before them. Thirty conversations did what nine months of program work couldn’t — not because the program work was bad, but because it was aimed at the wrong variable.
Three questions before you fund anything
Your dashboard has a specific job and it does it well: which cohort changed, when the deterioration started, whether it’s concentrated in one channel, offer, product or stage. What it can’t tell you is why the people inside that cohort behaved differently. No amount of cleaner ecommerce metrics will get you there.
So once you’ve located the break, ask three questions. None of them takes more than two weeks to answer.
Are the customers I’m acquiring today the same customers I was acquiring eighteen months ago? Acquisition mix is the most common hidden driver of retention movement and the easiest of the three to check. If the answer is no, go find out what changed at acquisition before you touch the lifecycle program.
What is the customer actually doing with the product between purchases? Usage rate, expected cadence, how the second and third interactions land. You cannot infer this from transaction data. Go and have the conversations with actual customers — thirty interviews is usually enough, and it’s the fastest customer insight you’ll ever buy.
What changed in the competitive set over the last 24 months? Customers don’t leave in a vacuum. They leave for something. Knowing what tells you whether this is your problem or the category’s.
Payment behavior sits alongside those three rather than inside them, so separate voluntary from involuntary churn first. Do all of that, and if the evidence still points to lifecycle — build the lifecycle program. It’ll pay for itself. My objection was never to the work. It’s to committing real money before anyone can say what problem the work is meant to solve.
Diagnosis before intervention
Acquisition mix, product-value match, usage rate, competitive change and payment operations all sit upstream of the retention number. Working downstream while the cause sits elsewhere is how good teams spend two quarters improving a program and move the business by nothing. Once you figure out what is actually going on, the intervention is usually the easy part.
There’s a hiring lesson buried in this too. Your lifecycle team should be excellent at lifecycle. Your acquisition team should be excellent at acquisition. But somebody has to work backward from the business result — across acquisition, customer behavior, product, operations and marketing — and decide where the problem actually lives before handing it to one of those teams. That’s not a functional skill. It’s the strategy job above them. As your company grows, it’s common to discover it overindexes on marketing execution over marketing strategy.
If your churn isn’t behaving the way it should
Begin by getting curious. Get into the cohort data, find where the change happened, then go talk to customers from the affected group about what the data can’t explain. A little curiosity only needs weeks, not quarters. I’ve seen curiosity kill a six-figure lifecycle build and replace it with a change to product pack size.
A curious dig into retention behaviors is a common starting point in my fractional CMO work, partly because it’s a low-stakes way for us to find out whether we work well together, and partly because it tests the assumption underneath the plan before you commit real budget to executing it. Sometimes the answer is a stronger lifecycle program. Sometimes it’s an acquisition offer, a pack size, a subscription cadence, a product problem or a competitor you hadn’t clocked.
If your retention numbers are doing something the team can’t clearly explain, diagnose that before you add another program to the roadmap. And if you’d rather not do it alone, let’s talk →
