Strategy
7 ecommerce retention metrics US online retailers should stop tracking
Seven ecommerce retention metrics to stop tracking, with the threshold tests for acting on each one and the limits the numbers cannot cover.
What to take away
- Repeat purchase rate, email open rate and average order value sit at the top of most Shopify dashboards and move least when retention work changes.
- Replace them with a 90-day second-order rate by cohort and a click-to-delivered email rate.
- Act when a monthly cohort falls below 20 percent second-order rate, or five points under its own trailing average.
- No metric here explains why a customer left, and none survives a weak attribution model.
Repeat purchase rate is an average, not a signal
Repeat purchase rate is the share of first-time buyers who place a second order inside a chosen window. Most default reports leave that window open forever, so a 2019 customer who ordered again in 2021 still counts.
The figure then parks near 30 percent and says little about last month's buyers. The ecommerce retention metrics to stop tracking are usually the ones copied from a template dashboard.
The average hides timing. A buyer who orders twice in one week differs from one who returns after 14 months, yet both lift the same number.
Cohort tracking replaces this view. Group buyers by first-purchase month and follow each group for 90 days. The mechanics are in ecommerce analytics, which shows how to rebuild the view without breaking year-over-year comparisons.
Threshold to act: a monthly cohort below 20 percent second-order rate at day 90, or five points under its own trailing six-month average. When that happens, inspect the welcome sequence and the post-purchase sequence before touching the loyalty program.
Metrics that flatter a GA4 report
| Metric | Where it misleads | What to use instead |
|---|---|---|
| Repeat purchase rate (all time) | Averages years of cohorts into one figure | 90-day second-order rate by cohort |
| Email open rate | Pixel pre-fetching inflates opens on Apple Mail | Click-to-delivered rate |
| Average order value | Free-shipping thresholds push baskets up and margin down | Contribution margin per order |
| Session duration | Background tabs count as engaged time | Product view to add-to-cart rate |
| Loyalty points issued | Points measure a liability, not a behavior | Redemption within 60 days |
| Follower growth | Paid reach rarely maps to buying | New customer share from owned lists |
| Unsubscribe rate | A large denominator keeps the rate flat | Complaint rate per send |
Apple's Mail Privacy Protection arrived with iOS 15 in 2021. It pre-fetches images, so an open can register before anyone reads the message.
The FTC's CAN-SPAM compliance guide covers rules that matter more than opens for lawful commercial email.
What to watch instead is the subject of signal versus noise. That guide sorts email numbers that predict repeat orders from the ones that only look busy.
Example: reading a 90-day cohort table
An illustrative skincare store with 3,000 first-time buyers in a quarter showed a 34 percent repeat purchase rate, which looked healthy. Split by month, the 90-day second-order rate read 26 percent, 19 percent and 21 percent. The middle cohort had passed through a checkout change that broke the post-purchase email trigger for 11 days.
That one column told the team more than the headline rate had in two years. The 19 percent cohort sat under the 20 percent action line, and the fix was a trigger, not a campaign.
What these numbers cannot tell you
A 90-day second-order rate says nothing about margin. A second order can be a replacement, a returned item bought again, or a gift buyer who never returns. None of that shows up in the rate.
Small stores face a second problem. With fewer than 500 first-time buyers a month, one strong week swings the cohort rate by several points, turning the 20 percent line into noise. Read two quarters before you act on it.
If the drop came from the site rather than the list, ecommerce conversion optimization sets out how to separate the two.
Attribution and its limits
Google Analytics 4 spreads a returning purchase across several touchpoints, and consent mode drops some sessions before they are counted. Email platforms log a click when a scanner follows a link. Both habits inflate whichever channel you are watching, usually email.
Under Section 5 of the FTC Act, a retention claim you publish needs evidence behind it. A dashboard figure you cannot defend is a business risk, not just a weak number.
When to stop measuring and decide
Review the cohort table monthly, decide quarterly. If a metric has produced no change in two consecutive quarters, retire it. The test is blunt: name the decision the number would change. When nothing comes to mind, the metric is decoration.
A metric that has never changed a decision is not a metric. It is a habit.
Programs that survive this filter are described in ecommerce retention marketing, which separates spending that returns a profit from spending that returns a report.
