Pulsar · Metrics playbook · No. 11 · Meta-attributed vs pixel-recorded. Real numbers from the test account, 30-day window (EGP).
Meta Says 6, You Sold 1
Why Meta’s purchase count and your real sales don’t match — and which one to trust for money decisions.
Companion to No. 01–10 — leans on No. 04 · The Invisible Sales (is the pixel number complete?), No. 05 · Who Really Made the Sale? (attribution by real click path) and No. 01 · Efficiency & Fatigue (the CPA you should actually be reading).
Contents
01 Two numbers, one sale · 02 Why they differ · 03 The real gap · 04 The CPA twist · 05 It cuts both ways · 06 Which to trust · 07 The fix
Two numbers, one sale
Meta’s reported “purchases” and your pixel’s recorded purchases are two different measurements of the same funnel — and they rarely match. They aren’t two views of one counter; they answer two different questions. Meta answers “how many sales can my ads take credit for?” Your pixel answers “how many Purchase events actually fired?” On this account, one creative reads Meta 6 vs pixel 1 — same ad, same month, same store.
Meta reports the sales it wants credit for. Your pixel reports the sales that actually happened.
Neither number is a lie. They’re produced by different rules — and the rest of this playbook walks the rules (section 02), the real gap they open on this account (section 03), what that gap does to the CPA you make decisions with (section 04), and the honest answer to “so which one do I believe?” (sections 05–06).
Why they differ — the mechanics
Meta counts generously, by design. Three mechanisms inflate its number relative to what your pixel records — none of them is a bug, all of them are attribution policy:
One buyer, one sale — who gets to count it?
A typical path: saw the ad → clicked → bought days later. Toggle the windows to see which rules let Meta claim the sale — the pixel records 1 Purchase event either way.
Interactive in the HTML version. The default state prints above: with both default windows on, Meta claims the sale twice over — the Day-2 click sits inside the 7-day click window and the Day-4 impression sits inside the 1-day view window. The pixel records exactly 1 Purchase event, attributed to the Day-2 click. In the view-through scenario (click removed), Meta still claims the sale on the 1-day view alone — while the pixel’s click-path attribution reports the buyer as not ad-attributed.
Net result: Meta almost always reads higher. Windows, view credit and modeling each add sales the pixel never saw an ad-click for — and they stack. That’s the whole gap in section 03.
The real gap — this account, 30 days
Same store, same month, both counters running. Whole account: Meta attributes 18 purchases across 48 creatives; the pixel recorded 10 real purchases (11 distinct buyers by click path — one bought without a tracked ad click). One creative: Meta 6, pixel 1 — a 6× gap on 1,217.72 EGP of spend.
Meta-attributed vs pixel-recorded
Two levels of the same gap. The account-level ratio looks tame; per creative it gets extreme — and per creative is where you make scaling decisions.
Interactive in the HTML version — a toggle switches between the two levels. This creative: Meta-attributed 6 vs pixel-recorded 1 (6×), on 1,217.72 EGP of spend and 1 real order. Whole account, 30 days: Meta 18 across 48 creatives vs pixel 10 real purchases, 11 distinct buyers by click path (~1.8×).
Why the gap costs you money — the CPA twist
Here’s where the gap stops being academic: Meta divides your spend by its own inflated count, so the CPA it shows you is flattering. On this creative: Meta CPA = 1,217.72 ÷ 6 = 202.95 EGP. Real cost per confirmed sale = 1,217.72 ÷ 1 = 1,217.72 EGP. The ad looks 6× cheaper than it really is.
The real-CPA calculator
Defaults are this creative’s real numbers. Type your own — both CPAs update live.
Interactive in the HTML version — a live calculator over spend / Meta conversions / pixel purchases. The printed values are the real defaults: spend 1,217.72 EGP, Meta 6, pixel 1 → Meta CPA 202.95 EGP vs real cost per confirmed sale 1,217.72 EGP — a 6× overstatement of how cheap the ad is.
The consequence chain: you scale a creative Meta flatters → real orders don’t follow the budget → your ROAS is overstated, so nothing looks wrong → you keep funding a loser. The whole failure runs on one division by the wrong denominator. The CPA you should actually watch is the one built in No. 01 · Efficiency & Fatigue — computed on confirmed sales.
But the gap cuts both ways
Careful: a low pixel number is not automatically the truth. The same gap appears when the pixel is under-tracking — a blocked browser pixel, weak match quality, no server-side events. Then the pixel number is too low, and Meta’s count may be the one closer to reality.
Meta too high, or pixel too low? Check both before you decide.
| Symptom | What’s broken | How to confirm | Verdict on the gap |
|---|---|---|---|
| Coverage high, EMQ strong, gap large | Nothing — Meta is over-attributing | Coverage & match quality per event type (No. 04) | META TOO HIGH |
| Coverage patchy, EMQ weak, no server-side | Your tracking — the pixel misses real sales | Same check — low EMQ / missing CAPI shows up immediately | PIXEL TOO LOW |
This account: coverage is 100% and match quality is solid (No. 04 · The Invisible Sales) — so here the pixel count is trustworthy and Meta is genuinely over-attributing. That verification step is what makes the 6-vs-1 verdict safe to act on. Skip it, and you can’t tell an inflated claim from a broken counter.
Which number to trust — and when
The honest answer is both — for different jobs. Each number is the right tool for exactly one decision, and the expensive mistakes come from crossing them.
Optimize with Meta’s data. Decide with your own.
The fix — four steps, then read both numbers
You can’t make the two counts equal — they follow different rules by design. What you can do is make both as honest as possible, then use each for its job.
Rule out under-tracking first
Send server-side (CAPI) events with strong match quality so both numbers are as complete as possible (No. 04 · The Invisible Sales). Until this is done, a low pixel count proves nothing.
Know your attribution window
7-day click vs 1-day view — know which setting your reports use, so when you compare Meta to the pixel you’re comparing like for like, not a window to an event count.
Reconcile monthly
Meta ↔ pixel ↔ store, once a month. Treat a big gap as a flag, not noise — it’s either over-attribution (section 03) or under-tracking (section 05), and the table above tells you which.
Judge creatives on confirmed sales
Scale and kill on pixel-recorded purchases and real revenue, never on Meta’s attributed number — that’s the 202.95-vs-1,217.72 lesson from section 04, and the real CPA from No. 01.
Where Pulsar fits, in one honest line: Pulsar shows both numbers on every creative — “Purchases · Meta” next to “Purchases · Pixel” — counts real people rather than duplicate events, and measures match quality so you know whether the pixel number is trustworthy before you act on it. The deep dives: No. 04 (is the pixel number complete?), No. 05 (attribution by real click path), No. 01 (the real CPA).