Why a product-usage model needs rewriting for a shop
In a software product, the stages are separated by behaviour: someone signs up, reaches first value, comes back the next week and eventually pays. A shop compresses most of that into a single moment. The visitor who checks out has been acquired, activated and has generated revenue in one session, and there is no usage to observe between orders. What happens next is silence until the next purchase, or for ever. The interesting part of an ecommerce dashboard is therefore not the long chain before payment but the short gap between the first and second order, which the generic dashboard barely shows.
Payment order matters too. In a subscription product, revenue grows with each month of retention; in a shop, each order is separate, and the first one is often bought at a loss through paid media. The ecommerce conversion hub covers how the first order is won; AARRR's job here is to show what that order is worth and whether a second follows.
Decisions that change for sales
| Stage | Generic AARRR | AARRR for ecommerce sales | Decision it forces |
|---|---|---|---|
| Acquisition | New users from any channel | New visitors and new buyers, split by own store and marketplace | Which buyers the shop can actually retain |
| Activation | First value inside the product | The first order, with its margin after acquisition cost | Whether a loss-making first order is acceptable, and for how long |
| Retention | Returning to use the product | A second order within the category's natural cycle | The repeat window for each category |
| Referral | Invitations sent and accepted | Reviews, gift orders shipped to new people, user photos | Which signals count and how to attribute them |
| Revenue | Payments per period | Revenue and margin per customer over a fixed horizon | The horizon and whether margin or revenue is reported |
Make the first order the activation event, and price it
Since there is no product to adopt, the first order is the point of first value. The decision AARRR adds is to record what that order earned after acquisition cost and discount, not only that it happened. A shop that wins first orders at a loss is betting on repeat purchaseShare of customers who placed more than one order in the period.; the dashboard should show the size of the bet. Report first-order conversion rate (CVRShare of visits or clicks that convert.), first-order value and first-order contribution side by side. A first order bought with a deep welcome discount is still activation, but it is a weaker one, and some shops record it as a separate row.
Set the retention window from the buying cycle
Retention in a shop is a second order, and the question is within how long. Pick the window from the category's natural cycle: how long existing repeat buyers took to come back. A single window across very different categories produces a blended rate that rises or falls with the product mix alone. The repeat purchase and loyalty guide covers the replenishment and cross-category routes that move the stage; the lifecycle mechanics live in the ecommerce email and SMS flows guide, where post-purchase timing should follow the same window.
Count revenue per customer, over time
Order value (AOVAverage revenue per order.) describes a basket. The revenue stage in AARRR should describe a customer: total revenue, and ideally contribution, per customer acquired in a given month, read at fixed ages such as three, six and twelve months. This is the bridge to lifetime value (LTVTotal value a customer brings over time.), and the LTV and CAC guide explains why margin rather than revenue should carry the decision. The choice to make is the horizon: long enough to include the typical second and third order, short enough to be read within a planning year.
Find referral in reviews, gifts and photos
Most shops have no invitation mechanic, so referral has to be found elsewhere. Reviews written by buyers persuade later visitors; gift orders put the product in a new household; photos posted by customers reach their followers. Count the ones the shop can observe: review submission rate per order, gift orders shipped to a new address, and orders arriving through a referral code where one exists. Attribution here is loose, so treat the stage as a health measure rather than a driver to be credited with sales.
Separate marketplace buyers from own-store buyers
A shop selling on marketplaces as well as its own site has two populations. Marketplace buyers often cannot receive the shop's emails and may repeat on the marketplace where the shop cannot connect the orders. Put them in a parallel row. Paid acquisition through shopping and catalogue ads and Meta catalogue ads mostly feeds the own-store row, where the shop can retain what it buys; spending to win marketplace buyers should be judged on first-order contribution alone.
Decision points in a worked AARRR for ecommerce sales
An invented homeware shop, Ashgrove Home, builds an AARRR dashboard and finds its constraint between first and second order
Ashgrove sells candles, textiles and small furniture. The dashboard reads the cohort of new customers from March, at 120 days, for own-store buyers; marketplace buyers sit in a separate row.
| Stage | Definition used | March cohort | Rate |
|---|---|---|---|
| Acquisition | New visitors to own store | 60,000 | – |
| Activation | First order on own store | 1,200 | 2.0% of visitors |
| First-order economics | Average first order, contribution after ad cost and discount | £58 / –£4 | – |
| Retention | Second order within 120 days | 216 | 18% of first-time buyers |
| Referral | Review submitted / gift order to a new address | 168 / 72 | 14% / 6% of first orders |
| Revenue | Revenue per customer at 120 days | £71 | – |
| Marketplace row | First orders on marketplace | 1,500 | Repeat not observable |
The weakest point is clear once the stages are priced. Each first order loses £4, so the shop only makes money on customers who come back, and 82% of them do not within 120 days. Splitting retention by first purchase category shows the gap: 31% of customers whose first order was a candle returned within the window, against 6% whose first order was furniture. Furniture buyers are not unhappy; their buying cycle is simply far longer than 120 days.
Decisions: the window becomes 90 days for candles and textiles and 365 days for furniture, reported separately; prospecting campaigns lead with candles and throws as entry products; a post-purchase flow sends a refill reminder at 45 days for candle buyers; and the revenue stage reports contribution per customer at 12 months so that the loss on the first order is set against what follows.
Without priced activation, the dashboard would have shown a healthy conversion rate and hidden that every first order lost money. A single retention window would have blamed furniture buyers for not repeating on a consumable's schedule. And revenue per order would never have revealed that the business case rested entirely on the second order.
Measuring AARRR for ecommerce: KPIs and cadence
Weekly: visitors, first-order conversion rate, first-order value and contribution by channel, since paid media decisions turn on them. Monthly: second-order rate for each monthly cohort at fixed ages, by first purchase category, and review and gift rates. Quarterly: revenue and contribution per customer at the chosen horizon, compared across cohorts, with return on ad spend checked against contribution rather than revenue. Ecommerce events in the analytics tool give order data per session; customer-level reading needs the order system or a customer data platform, because analytics identifiers do not survive across devices and months. The conversion-rate and ROAS calculators help with the per-stage arithmetic.
Where AARRR misleads for a shop
Seasonality distorts cohorts: customers acquired in a gifting peak repeat less because many bought for someone else. Returns sit outside the stages unless subtracted, so a first order later refunded still counts as activation. Guest checkoutLetting a shopper complete an order without creating an account, with the account offered afterwards when the details are already captured. breaks the link between orders and customers, so retention is understated for shops that allow it. Promotions can pull a second order forward without adding a third. And the framework's straight FunnelModel of the customer journey stages. shape hides that a returning customer is reached, persuaded and converted again for each order, which is closer to a loop than a line.
Common mistakes
- Activation without margin. First orders bought at a loss look like success.
- One repeat window for every category. The product mix moves the retention rate on its own.
- Revenue per order as the revenue stage. It describes baskets, not the customers who fill them.
- Marketplace buyers blended in. Retention falls for customers the shop could never contact.
- Counting refunds as activation. A returned first order was never a customer gained.
- Reading every cohort at the same calendar date. Older cohorts have simply had longer to repeat.
