How a sale gets attributed to an ad
When someone clicks your ad, the platform adds a click id to your landing URL: fbclid for Meta, ttclid for TikTok, gclid for Google, ScCid for Snapchat, epik for Pinterest. If that click id, or a cookie based on it, is still around when the buyer pays, the platform can say: this sale came from that ad.
Platforms also match on the buyer's identity, using hashed email and phone, and some count view-through conversions, where someone saw an ad without clicking and bought later. Each platform does this on its own, with its own rules.
Attribution windows decide what counts
Every platform credits a sale only if it happens within a set window after the click or view. Meta's default is 7 days after a click and 1 day after a view. Other platforms use their own defaults, and you can often change them.
That means the same ad can look profitable or unprofitable depending on the window you read. Products with a longer buying cycle, like furniture or gifts planned ahead, need a longer window to be judged fairly.
There's a second, quieter window: how long the click itself survives on your side. A platform can only credit a sale inside its window if it still receives the click or can match the buyer. If Safari has already cleared the cookie after a few days, a longer platform window doesn't help. Keeping the click on the server, tied to the visitor, is what makes the platform's window actually usable.
Why platforms both under-claim and over-claim
Under-claiming happens when the platform never sees the purchase or can't connect it to the click: blocked pixels, cookies cleared by Safari, in-app browsers, buyers switching device. The ad did its job; the platform just doesn't know.
Over-claiming happens because every platform attributes independently. A buyer who clicked a TikTok ad on Monday and a Meta ad on Thursday may be counted by both. Add up what all platforms claim and it can be more than you sold.
Shopify's own reports look at it from the store's side, based on UTMs and referrers of the session that placed the order. Useful, but they only see the last visit, and miss buyers who came back by typing your URL.
What you can fix, and what you can't
- Fixable: purchases that never reach the platform. Send every paid order server-side.
- Fixable: clicks that are forgotten before the buyer pays. Save the click id at landing and keep it.
- Fixable: weak matching. Send hashed email and phone with each purchase.
- Partly fixable: platforms each claiming the same sale. Compare against your real revenue.
- Not fixable by any tool: buyers who refused consent, or who never clicked anything traceable.
A simple Shopify ad attribution routine
You don't need a data team to keep Shopify ad attribution honest. A short weekly routine, fifteen minutes on a Monday morning with your store and ad dashboards side by side, catches most problems:
- Compare each platform's reported purchases with last week's paid orders. A sudden drop on one platform usually means a tracking problem, not a creative problem.
- Add up what all platforms claim and compare it with your real revenue. Far below means missing data; far above means overlapping claims.
- Read each platform on the same attribution window every week, so you compare like with like.
- Look at new versus returning buyers. Returning buyers often come back without a fresh ad click.
- Check the per-order status in your tracking tool for failed sends before blaming the ads.
How Settl handles attribution
Settl saves every click id and UTM tag the moment a visitor lands, first-party, and keeps it with that visitor for 90 days. Returning buyers are recognized, even on a new device once they've bought before. When the order is paid, it goes to each connected platform with the click it needs, and the platform does its own matching.
In the dashboard you see spend, sales and ROAS per ad, from your connected ad accounts and your real orders, plus advice on which ads to scale or switch off. Settl doesn't build multi-touch models; it gives each platform complete data and shows you the result per ad.