Attribution for Repeat Purchases: Your Email Flows Are Being Underpaid

Last click reporting hands repeat purchases to whatever ad the customer touched last. How replenishment and post purchase flows lose credit, and a saner way to split it.

A customer who has bought four times gets a replenishment email, clicks it, and buys the fifth. Somewhere upstream they also saw a prospecting ad that week, and in a last click report the ad platform takes the order, the email platform takes a view through assist at best, and the flow that actually caused the purchase gets budgeted like it does nothing.

Repeat purchase attribution is where platform reporting diverges furthest from what happened, because the platforms bidding for budget only see the new customer game clearly. Your existing customers are being re-sold to you in your own reports, and the flows doing the retention work look weaker every quarter.

Separate the questions before the models

Attribution arguments go in circles because three different questions share one report. Who found this customer. Who prompted this order. Who should get more budget. For repeat purchases the first question is settled history; the order prompt is the live one, and it is usually owned media: an email, an SMS, a subscription reminder, a post purchase flow with good timing. Reporting that gives the order to the last paid touch answers none of the three questions well.

Split the reporting by order number. First orders get the acquisition view, platform claims and all, argued about separately. Second orders and beyond get a retention view where owned channels are credited on click and on timed sends, and paid touches are shown as exposure, not ownership. The fights shrink immediately, because most of them were two teams citing the same blended number for different jobs.

The replenishment timing test

For consumables and repeatable products, the cleanest incrementality read you can run is a holdout on timing. Hold a slice of due customers out of the replenishment flow for one cycle and compare reorder rates and reorder timing against the group that got the email. Expect two effects: some held customers buy anyway, later, and some defect or forget. The flow's true value is the gap, sooner plus saved, and it is almost never the full revenue of every order the flow touched, which is what the email platform reports.

Run it per product cycle, not once for the whole file. A thirty day replenishment product and a ninety day one behave differently, and a flow timed a week early can pull orders forward without creating any, which looks like performance until the following cycle dips. Pull forward is the repeat purchase version of a vanity metric: real revenue, wrong conclusion.

Budget follows the split, or the split was theatre

The point of separating first and repeat orders is a budget consequence. If retention flows are carrying repeat revenue that paid platforms were claiming, the prospecting budget was being justified partly by other people's orders, and it should shrink or prove itself on first orders alone. That is an uncomfortable meeting. It is also the only version of attribution work that pays for itself, because everything before it is chart rearrangement.

Related reading

If this topic is useful, these earlier guides go deeper on connected questions: Reading a Lift Test When the Conversion Counts Are Small, Coupon Extensions and Last-Click Credit: When the Final Click Did Nothing, and View Through Conversions: A Sceptical Read Before You Pay for Them.

Frequently asked questions

Why does last click undervalue email flows for repeat orders?

Because an existing customer who clicked a replenishment email often also touched a paid ad recently, and last click hands the whole order to the ad. The flow that prompted the purchase gets little or no credit.

How should repeat purchase attribution be reported?

Separately from first orders. Credit owned channels for repeat orders on click and timed sends, and treat paid touches as exposure rather than ownership.

How do you test whether a replenishment flow is incremental?

Hold out a slice of due customers for one cycle and compare reorder rate and timing. The gap, orders sooner plus orders saved, is the flow's value, not every order it touched.

What is pull forward in repeat purchases?

Orders the flow merely moved earlier without creating them. The current cycle looks strong and the next dips. Test per product cycle to see it.