Meta attribution for a store: 7-day click only, or keep 1-day view in the main ROAS?

Say a fashion store wants Meta's 7-day click plus 1-day view numbers in the weekly report. I'd make 7-day click the headline and put view-through in a separate column that nobody bids on. On a short purchase cycle, view conversions tend to inflate ROAS, and then the number stops matching Shopify revenue.

Some people here keep view-through in the main figure and argue the lift is real. That holds up better for prospecting with a long consideration cycle, where a view can genuinely start the journey. But a store that mostly sells to people who already saw a retargeting ad is different. There the view window often just takes credit for orders that were coming anyway.

How do you reconcile Meta's numbers with the backend when they drift apart?

Agreed on click as the headline. To check the gap, pull Meta's purchases for the same date range, split by click-through and view-through, and put the total next to Shopify's order count for that period. Shopify will usually come in lower, since Meta also counts some cross-device orders and some that would have happened anyway. The view-through slice is the part that matters. If it makes up a big share of Meta's purchases while Shopify's totals barely move, the view window is taking credit for sales that were already coming.

To test it directly, pause retargeting for a week or two, or split the audience with a holdout if Meta's split test tool is available to you, and watch whether Shopify revenue actually drops. That tells you more than any attribution setting.

The pause test is a good idea, but I'd track total store revenue instead of just Meta's reported purchases. When retargeting goes dark, some of those buyers come back through search, email or direct visits, so Meta's line can fall hard while Shopify barely moves. A store with a strong email list might see only a small dip and wrongly conclude the view window was pure padding.

Compare total revenue against the same weeks last year, or against a region you left running if you have one. Does the drop usually land near the view-through share, or well below it?

I'd expect the drop to land well below the view-through share. Some view conversions are real, just not incremental the way Meta reports them. The same-weeks-last-year comparison worries me because of seasonality. A fashion store's sales swing with the calendar more than most promos do, so a year-over-year gap can look like a retargeting effect when it's really just November. A region holdout is cleaner since both regions face the same calendar, as long as their order volumes are similar and the spend split stays steady for the whole test.

How similar do the two regions need to be before you trust the result?

Order volume matters less than how closely the two regions tracked each other before the test starts. I'd pull several weeks of the daily revenue ratio between them, because if that ratio bounces around, any gap you see during the test is just noise. Do you have a way to check that before you spend anything?

It depends on whether the two regions have a stable ratio before the test, so that's the first thing to check. Iris's daily ratio idea works, but I'd go a step further and regress one region on the other over the pre-period, then look at how widely the residuals spread. If that spread is tight compared to the lift you expect, the holdout will be able to pick it up.

Regressing one region on the other is the right move, and the residual spread tells you more than the raw ratio does. What it comes down to is the size of the lift you're hoping to detect. If you expect retargeting to be worth 5% of revenue, you need a very tight pre-period fit, but if it's 20%, a rougher match can still work.

Lift size matters most, agreed. The pre-period also needs to span a full cycle of your purchase pattern, not just a few weeks. For a fashion store that usually means at least one weekend-heavy stretch plus any paydays or pay-cycle dips, because a short window can look tight by chance and then drift once the test starts.

Full cycle makes sense, but budget movement can quietly break the test. If you switch off retargeting in one region and the account's total budget stays the same, Meta often pushes that money into prospecting there. Those prospecting sales land in the treated region and fill part of the gap you're trying to measure, so the region looks like it lost less than it really did.

You can handle it two ways: cut retargeting without reallocating so regional spend stays flat, or log prospecting spend per region every week next to revenue. If the prospecting numbers move during the test, you're measuring the shift, not retargeting.

Do you cut the retargeting budget outright for the holdout, or let Meta redistribute it?

Cut it outright and put the budget cap at the ad set level, so Meta can't quietly move the money somewhere else. That way the setup answers the spend question and you don't have to babysit it with weekly checks. Also keep prospecting spend per region flat for the whole test, since even a 10% shift there can swamp the retargeting effect you're trying to measure.