Measurement & attribution
The numbers in Ads Manager are not gospel - they're a model. Understanding what they actually mean is what stops you and your clients from making confident decisions on misleading data. One of the most misunderstood areas in the field - and a serious credibility marker.
1What attribution actually is
Attribution = the rules for deciding which ad gets CREDIT for a conversion. A customer might see your Reel Monday, a Story Wednesday, click a Feed ad Friday, buy Saturday. Which ad caused it? Attribution settings answer that - and different rules produce very different "results" from identical reality.
Ads Manager conversions are an estimate built from three layers of assumption: privacy-era modeling, attribution rules, and Meta grading its own homework. The attribution window — default 7-day click + 1-day engage-through + 1-day view (as of mid-2026, check current options in your account) — is the dial: widen it and reported conversions rise; narrow it and they fall. Same sales, different reported numbers.
2Why the numbers are a model, not truth
- Privacy/iOS (Day 10) - with less individual tracking, Meta MODELS conversions it can't directly observe (statistical estimation, not literal counting).
- Attribution choices - the window and last-touch logic are assumptions, not facts.
- Platform self-attribution - Meta marks its own homework; it tends to claim credit for conversions that might have happened anyway, or that other channels also claim.
This is why Meta's reported conversions often don't match your Shopify / GA numbers. Both can be "right" by their own logic. Expecting them to tie out exactly is a beginner error.
What changed in 2026 (know this boundary): Meta retired the old 7-day-view and 28-day-view options, and "click-through" now counts link clicks only. Video and social interactions that used to count as clicks are now reported separately as engage-through. The practical effect: reported results from before vs after these changes are not directly comparable — don't misjudge a "winner" across that boundary.
| Same campaign, same week | Purchases | ROAS |
|---|---|---|
| 7-day click + 1-day engage-through + 1-day view | 95 | 3.6 |
| 1-day click only | 58 | 2.2 |
| Shopify orders tagged to Meta | 71 | — |
Illustrative figures, not a benchmark — your results will vary. Nothing about the business changed between rows — only the counting rules did.
The professional move: track the typical ratio between Meta-reported and backend conversions for each account (e.g. Meta reports ~1.3× Shopify). A stable ratio is normal; a sudden shift in the ratio is the real signal.
3The concepts that keep you honest
- View-through vs click-through - view-through (saw, didn't click, later bought) is the softest, most over-credited. Click-through is stronger.
- Last-click bias - over-credits the final touch, under-credits upper-funnel ads that created demand.
- The correlation trap - retargeting "shows" amazing ROAS because it targets people already about to buy. High reported ROAS ≠ high incremental value (tomorrow).
- Blended metrics - watch blended ROAS and CPA as a sanity check against any single platform's self-flattery. Blended ROAS = total revenue ÷ total ad spend (all channels, same period); blended CPA = total ad spend ÷ total new orders.
- Cost per NEW customer - total ad spend ÷ new customers acquired. Counting only NEW customers strips out the people who would have bought anyway — the same correlation trap retargeting exploits.
Three reps each told "you get commission if a client buys within 7 days of your call." A client who was always going to buy talks to all three - all three claim the commission. Sum the claims and you've "paid" for the deal three times; the numbers overstate everyone's impact. Meta's reported conversions are like one rep's self-reported commission claims: useful, but inflated by overlap and generous rules. The honest manager checks total revenue (blended).
- Treating Ads Manager ROAS as literal truth — and over-investing in channels that just claim credit.
- Panicking that platform numbers don't match Shopify — they never will exactly.
- Loving retargeting's inflated ROAS without asking if it's incremental.
- Ignoring upper-funnel because last-click under-credits it.
Your edge: explain attribution honestly, set expectations that reported ≠ incremental, report blended numbers.
Recap - 30 seconds
- Attribution = the rules deciding which ad gets credit; the window changes reported numbers without changing reality.
- Ads Manager conversions are a MODEL (privacy modeling + attribution assumptions + self-attribution), not a literal count.
- They won't match Shopify/GA - expected; learn your account's ratio.
- Beware view-through, last-click bias, and retargeting's inflated ROAS (reported ≠ incremental).
- Watch blended CPA/ROAS; honest measurement is a serious client-trust edge.