Guide · Shopify and Meta

Why does Meta overrate refund‑heavy customers on Shopify?

Meta learns from what a customer spends at checkout. Refunds and repeat orders come later, and the bidder never sees them.

With purchase-value optimisation, Meta learns from the value you send when the order is placed. Refunds and repeat orders happen later, so a big basket that mostly comes back counts as a high-value customer, and a modest first order from someone who rebuys every two months counts as a low one. Meta then finds more of the first kind.

Updated November 2026. Meta facts checked against Meta’s developer docs on 18 November 2026.

What does Meta learn from at checkout?

When a campaign optimises for purchase value, Meta bids on the value you send with the Purchase event. Meta’s own guide says its system values conversions in proportion to the value passed back (Meta, value optimization). On Shopify that value is the order total at checkout, sent by the Meta pixel and the Conversions API.

Meta sees the purchase it can attribute to an ad. It does not see what happens next: the parcel that comes back, the second order three months later, the subscription that renews.

Which Shopify customers look best on day one but aren’t?

  • Bracket buyers order the same item in two or three sizes and keep one. The first basket looks big, and most of it comes back.
  • Wardrobers buy for one occasion, wear it once and return it.
  • Buyer’s remorse turns a confident first order into a refund within days.
  • One-and-done buyers buy once on a discount and never come back.

To Meta each of these is a high-value purchase. The bidder learns that people like them are worth paying for, and goes looking for more.

Which customers look ordinary on day one but are your best?

The opposite mistake. A modest first order from a customer who comes back every two months looks like any other modest first order. Their value arrives in later orders that Meta never ties back to the ad that found them, so the bidder pays too little for people like them. Your repurchase rate tells you how big this group is.

Why doesn’t sending refunds to Meta later fix it?

Meta’s list of standard events has no refund event (Meta Pixel reference), and its value optimization guide describes no way to take value back from a purchase it has already received. A refund that posts two weeks later does not change the value the bidder learned from on the day of the order. Your refund rate is how much of that learning was wrong.

Google Ads does let you retract or restate a conversion after the fact. Even there, the correction arrives after the auction has bid on that customer. The fix is to send a better value at the start.

On your own store

How do you measure this?

Compare what each customer spent at their first purchase with what they were worth later, net of refunds. Two numbers tell you most of it:

  • Share of one-year value visible at first purchase. If it is high, the first purchase is a good signal. If it is low, your ads bid on a fraction of what customers are worth.
  • Best customers kept. Of the 10 customers who look best at first purchase, how many are still among your best a year later. Few means the first purchase ranks customers poorly.

Best Customer Check measures both on your own orders, at 120, 180 and 365 days as far as your history allows. It also shows refunded revenue, repeat revenue, and repeat rate by first-order size. It is a free, read-only Shopify app, and takes 1 to 10 minutes.

The fix

What should you send Meta instead?

A value that already accounts for what happens next: each new customer’s predicted value, net of expected refunds and with expected repeat orders, sent as the conversion value. Meta supports this directly with pLTV value optimization, where the predicted value goes with the purchase or follows up to 7 days later. See Meta pLTV value optimization and pLTV on Meta.

That is what Churney does for brands spending $300,000 or more a month on paid acquisition. It predicts each new customer’s value from your own data within days of the first purchase, sends it to Meta, Google and TikTok, and proves the result in a split test against your current campaigns. Book a call.

FAQ

Questions ecommerce teams ask

Does Meta subtract refunds from purchase value optimisation?

Not by itself. Meta’s standard events include no refund event, and Meta documents no way to remove value from a purchase it has already received. The bidder learns from the value sent at checkout.

Should I optimise Meta for first-time purchases only on Shopify?

It stops you paying again for customers you already have, but it does not fix the value problem. A new customer who sends most of the basket back still looks valuable on the day of the order.

How do bracket buyers affect Meta ROAS?

They inflate it. The ad gets credit for the full basket, the return comes later, and reported ROAS stays above what you kept. Meta also learns to find more bracket buyers.

How much order history do I need to measure this?

For Best Customer Check, at least 200 customers whose first order is 120 days old or more. The one-year view needs 200 customers whose first order is at least a year old.

Is there a free Shopify app that shows how refunds distort my ad signal?

Yes. Best Customer Check is free and read-only on the Shopify App Store. It shows how much revenue your ads counted at the first purchase that later came back, and how much of a customer’s one-year value they saw at all.

Sources: Meta, value optimization; Meta Pixel standard events reference; Meta, pLTV integration guidance.

See it on your own orders.

Free, read-only, about ten minutes from install to report.