Customer Retention Metrics: The Small Scorecard That Changes Decisions
A small scorecard beats fifty lagging metrics when every movement leads to a defined diagnosis.
A useful ecommerce retention scorecard needs five decision numbers, not fifty: cohort retention, repeat purchase rate, time to second order, contribution-aware lifetime value, and a customer-friction signal.
Retention is not one “loyalty” number. It is a decision system: whether customers return, how quickly they return, whether those orders create profit, and what experience problems may push them away. Purchase frequency and average order value provide diagnostic context.
Table of Contents
- Which customer retention metrics belong on the scorecard?
- Why is ecommerce retention harder than subscription retention?
- What should you do when each metric moves?
- How should the scorecard be segmented?
- Which tools deserve a place in the stack?
- Sources
- FAQ
Which customer retention metrics belong on the scorecard?
A practical scorecard balances behavior, timing, economics, and friction. Keep definitions stable so monthly comparisons remain meaningful.
| Metric | Formula or definition | What it tells you |
|---|---|---|
| Cohort retention rate | ((ending customers − new customers) / starting customers) × 100 | Active customers remaining after a defined period |
| Repeat purchase rate | customers with 2+ orders / customers × 100 | Customers who have purchased at least twice |
| Purchase frequency | orders / unique customers | Orders per customer in a period |
| Average order value | revenue / orders | Average transaction value |
| Time to second purchase | Median days for an eligible cohort to place order two | Speed from first to repeat order |
| Contribution-aware LTV | Customer revenue adjusted for margin, fulfillment, returns, support, discounts, and acquisition costs | Whether retention creates durable value |
| Customer-friction signal | Defined combination of returns, support contacts, complaints, refunds, delivery issues, or negative feedback | Where experience may weaken retention |
Cohort retention deserves priority because it preserves timing. Group customers by first-order month, acquisition source, or first product, then compare repeat activity at the same cohort age. January and June cohorts should not be compared by total revenue when their observation periods differ.
Repeat purchase rate answers whether customers crossed the second-order threshold. Pair it with time to second purchase: the same rate after 45 days is healthier than after 300 days. Purchase frequency and AOV explain revenue changes; larger baskets can mask fewer buyers, while higher AOV may reflect discounts, bundles, or a low-margin mix.
Contribution-aware LTV is the economic guardrail. Revenue-only LTV can overstate value when returns, paid acquisition, expedited shipping, support labor, or discounts consume contribution. Use the average order value calculator as a starting point, then extend the model to contribution.
The friction signal should be defined consistently and reviewed alongside retention. It can combine return rate, support contacts, complaints, refunds, delivery issues, and negative feedback, helping the team find experience problems before another campaign is launched.
Why is ecommerce retention harder than subscription retention?
Subscription businesses have a built-in billing rhythm; missed payment or cancellation is clear. Ecommerce timing varies with category cadence, seasonality, replenishment, budget, and preference.
A skincare buyer may return in 30–60 days; a luggage buyer may wait two years. A universal 90-day inactivity rule would label the latter churned too early. Set the window from observed category cadence: start with median or upper-quartile time between orders, then test 60, 90, 180, or 365 days. Use the point at which a customer is meaningfully overdue, not a convenient dashboard refresh.
Cohort maturity matters. Do not judge last month’s cohort on 180-day retention. Mark a cohort eligible only when it has enough observation time. For time to second purchase, calculate median days among customers whose window has matured or who already reordered; customers without enough time are not failures.
Seasonality requires like-for-like comparisons. Compare holiday cohorts with holiday cohorts, and separate promotional periods from ordinary trading. A lift during a deep discount may not persist at normal prices.
What should you do when each metric moves?
| Movement | Questions | Response |
|---|---|---|
| Cohort retention falls | Lower-intent acquisition? Delivery, quality, or mix change? | Compare source/first product; review returns, complaints, incidents |
| Repeat purchase rate falls | Did customers reach a second-order opportunity? Was first order satisfying? | Improve education, replenishment reminders, and second-order offers without automatic over-discounting |
| Time to second purchase rises | Reminder late, or category cadence longer? | Trigger around expected use/replenishment; test earlier, relevant messages |
| Frequency falls while AOV rises | Fewer customers buying larger baskets? Promotions changing mix? | Inspect customer-level behavior; protect repeat demand and margin |
| Revenue rises but contribution-aware LTV falls | Discounts, returns, paid media, or service costs absorbing gain? | Reallocate spend, tighten promotion rules, investigate product/fulfillment costs |
| Support contacts and returns rise | Promise unclear? Sizing, setup, delivery, or quality issue? | Fix the root experience problem before adding retention messaging |
Movement prompts diagnosis, not proof of causation. Retention changes can reflect product quality, acquisition mix, pricing, delivery, promotion, competition, or tracking. Use comparisons and customer evidence before acting.
A falling repeat rate with rising returns is stronger evidence than either signal alone. A longer second-order interval with stable satisfaction may simply reflect a longer category cycle. Review contact reasons, resolution time, refunds, and post-resolution behavior; the customer service operating model may matter more than another campaign.
Lifecycle messages should follow timing and behavior, not send every customer the same discount sequence. Use email marketing and ecommerce marketing automation workflows, judging programs by incremental repeat behavior and contribution rather than sends or clicks.
How should the scorecard be segmented?
A blended number can hide the decision you need. Segment by acquisition source or campaign, first product or category, first-order cohort, geography or delivery region, gross-margin band, return or support history, and new versus returning status.
Source indicates intent; first product shows repeat path; margin and service history reveal economics and friction. Behavioral segments—returned, contacted support, used a discount, or bought a replenishable product—often outperform demographic cuts. Keep segments large enough, use the same definitions and windows, and annotate pricing, catalog, fulfillment, and tracking changes.
Which tools deserve a place in the stack?
Choose tools for data quality and actionability. A platform that cannot connect orders, customers, refunds, returns, support, and contribution data will make fragile dashboards.
Make it easy to define customer and order consistently, build mature cohorts, reconcile revenue with refunds, discounts, returns, and costs, trigger an action, and compare results with a baseline or control. The commerce platform is the source of truth for orders and customers; analytics maps journeys; service explains friction; email and automation intervene; reporting joins data without changing definitions.
Before adding software, ask: Which January customers made a second purchase within 90 days by first product and source? What was contribution after returns and discounts? Did delivery-complaint customers return less? A compact scorecard can live in the SellerTrove retention report, while planning can use the stack builder. The goal is a shorter path from trustworthy signal to tested action.
Sources
FAQ
What is a good customer retention rate for ecommerce?
There is no universal benchmark. A useful rate depends on category cadence, product price, customer need, margin, and observation window. Compare cohorts at the same age, then look for sustained improvement within your own business. A 90-day rate may help replenishment products but be meaningless for infrequent purchases.
What calculation window should ecommerce retention use?
Use a window matching expected repurchase. Start with observed order intervals by category, then test a practical inactivity period. Report 30-day, 90-day, and 180-day retention when useful, but keep definitions consistent.
What is the difference between repeat purchase rate and retention rate?
Repeat purchase rate measures customers with two or more orders. Retention rate measures whether customers from a starting group remain active after a defined period using `((ending customers − new customers) / starting customers) × 100`. Repeat purchase is threshold-based; retention is time- and cohort-based.
Should LTV include returns, support, and acquisition costs?
Yes, when LTV guides investment decisions. Revenue-only LTV can overstate value. Include margin and variable costs such as discounts, fulfillment, returns, support, and acquisition. Start with a simple model, then improve it as reliable cost data improves.
We track pricing and new tools across the whole catalog. Get an email when prices move or a better tool launches.