How should we reconcile Meta-reported conversions with Google Analytics, GA4 and store totals, and which number do we trust for what?
Our answer
Use store orders and revenue as the number of record. Use Meta's conversions only to compare campaigns inside Meta, after an incrementality adjustment (How much Meta ROAS overstates return), and lift tests for what Meta caused. Never add platform-reported conversions together.
- Confidence
- Medium
- Strongest proof
- 8 Meta rules
- 3 datasets, 10 practitioners, 6 opinions or claims
- Sources
- 16 sources
- 27 statements
- M
New here? Take the 1-minute tour
Why this is the answer
3 reasons and our judgmentMeta and tools such as Google Analytics count purchases differently. Windows, time zones, view-through and cross-device credit and lost referrers all differ, and Ads Manager shows only purchases attributed to ads (Meta).
Store revenue outranks any platform figure. Even a calibrated incremental ROAS stays below revenue and contribution margin, and when they move apart you recalibrate the measurement (Common Thread Collective, an ecommerce ad agency).
Platform-reported results overcount. They include buyers who would have bought anyway and halo credited to the wrong channel, argues WorkMagic, an incrementality testing firm. Added together, platforms double-count orders each one claims (WeltPixel).
Our judgment
We put store totals first and keep Meta's numbers for comparisons inside Meta. That joins Common Thread Collective's ranking, Wicked Reports' advice to decide on the store's own revenue, Grüns' use of platform data for day-to-day moves, WeltPixel's view of Shopify as the system of record and Meta's note that Ads Manager shows only attributed events. We add a monthly check that planning numbers add up to the store's top line, following Michael Kaminsky (Recast) and Common Thread Collective.
Our judgment, built from Common Thread Collective, Wicked Reports, Grüns, Meta, Michael Kaminsky (Recast) and WeltPixel: our reading of these sources, not a finding from any one of them.
What the sources say
27 statements from 16 sourcesMeta's help page says on-device conversion measurement is on by default, is not active in the EU and UK, works alongside the Pixel and Conversions API, and does not change attribution windows or models.
“Measuring conversions on customer devices affects which conversions can be measured, not how they're credited to your ads.”
Across 62 Meta and TikTok geo lift tests it ran for 19 brands, WorkMagic found 58% of the lift landed off the brand's own site (14% Walmart, 16% Amazon). Counting Walmart raised iROAS 16%.
“When Walmart sales were added to the outcome, iROAS across these tests was 16% higher than without them.”
ECORN, a Shopify agency, says pixels record only the initial order, missing post-purchase upsells. Upsell.com data from more than 11,000 stores puts about 54% of accepted upsells on the first one-click offer.
“Shopify's pixel integrations record the initial order and its value, not the add-on a shopper accepts a few seconds later”
PPC Land says that from November 12, Meta's on-device conversion counting will make campaigns look like they sell a little more, while attribution windows and models stay the same.
“Nothing about the campaigns will have changed. What changes is the counting.”
CTC's hosts say to audit tracking by comparing total purchases in Shopify with those in Meta's Events Manager; they need not match exactly but shouldn't be far apart.
“Look at the number of purchases that you see in Shopify. Look at the number of purchases that you see in meta.”
Connor Roland says Meta reports a conversion on the day it happens, which inflates reported ROAS on peak days by ignoring lag from earlier days' spend.
“Meta will attribute that conversion on Friday when the conversion happened. And what it will end up doing is it will inflate the row as reported on the day of Friday.”
For one client, Marin Istvanic says a 2.8 ROAS Facebook day came alongside a promo email that drove about 40% of revenue; he estimates the real Facebook ROAS was closer to 1.7.
“Your real Facebook-driven ROAS yesterday was closer to 1.7.”
Meta's revised ad terms, effective October 30, 2026, drop the line calling automated features optional and make advertisers solely responsible for ads and targeting, even when Meta's automation produced them.
“You are solely responsible for the Order, Ad Creative, targeting decisions, and ad placements, including for reviewing, approving or removing them”
CTC reports that Meta began testing AI-built landing pages for ad clicks on October 14, 2026, on fewer than 5% of impressions in eligible Advantage+ campaigns, with an opt-out in Ad Account Settings.
“On October 14, 2026, Meta began rolling out a test where its AI generates a custom landing page for ad clicks”
To reconcile Meta and third-party conversion counts, Meta says to match attribution windows, time zones and conversion date settings, then check raw pixel fires for duplicates or tag setup problems.
“use the Meta Ads Data Advisor, a Chrome extension, to check for duplicate pixel fires, which can sometimes be caused by redirects.”
Meta says third-party tools can under-count Meta conversions because they rely on cookies and referrer URLs and miss view-through and cross-device sales. Meta's explanation partly favors itself.
“may under-report conversions on Meta technologies”
Meta says Ads Manager results are usually lower than Events Manager counts because Events Manager shows most events received, while Ads Manager shows only attributed events.
“Events Manager shows most events that have been received, while Ads Manager and Ads Reporting show attributed events”
Meta told advertisers it would add in-app product browsing to existing ads from October 28, 2026 unless they opt out, claiming 3% more conversions in a test of about 117,700 advertisers. PPC Land could not verify it.
“Starting October 28, 2026, we'll add product browsing to your existing ads.”
Both guests on this DTC Podcast episode say platform-reported revenue is the number they stopped trusting; Drew Fallon says it always runs higher than the P&L. No size is given for the gap.
“It's all higher than what you get on your P&L.”
Phil Kiel, whose agency buys media, says to run a separate Black Friday campaign without manual bids, load budgets at midnight, watch Shopify hourly since recent Ads Manager hours are partial, and check your credit line.
“Check your Meta credit line today. If it can't carry the weekend, delivery stops mid-sale.”
Gruns' growth head matches the tool to the decision: incrementality tests, a mix model and marginal return for big budget calls, and attribution and platform data for daily moves.
“I'm also not letting attribution decide whether Meta deserves another $500K.”
WeltPixel, which sells Shopify tracking, says Shopify owns the revenue number and GA4 always undercounts email against Klaviyo; adding Klaviyo's revenue to paid channels' double counts.
“Report the platform numbers side by side, label the model and window on each, and keep Shopify's total as the denominator.”
Stackmatix, a paid media agency, says treat Reddit's reported conversions as an upper bound and test with a 10 to 20% audience holdout or a two-week regional pause. Under about 30 events a week, optimize higher up.
“Treat Reddit's attributed conversions as an upper bound, not a ledger.”
WorkMagic, which sells mix modeling, says David Protein found its Meta attribution inflated by 36%, and after reallocating on incrementality revenue grew 34% and profit 37%. No method or period is given.
“David Protein found their Meta attribution inflated by 36%”
Michael Kaminsky of Recast says that if forecasts built on last-touch numbers keep missing, those numbers are wrong; your view of marketing performance should match month-end top-line results.
“Your beliefs in terms of marketing performance should be consistent with the top-line results that you see at the end of the month.”
Michael Kaminsky, whose firm Recast sells mix models, says most businesses do not need one and it can mislead them; he cites studies where last-touch was about 85 percent as good as experiments in simpler mixes.
“most businesses do not need an MMM. In fact, an MMM will be net negative for them because it will actively mislead them.”
A DTC Podcast guest says attribution tools let her fact-check media buyers, for example catching exchange orders counted as new sales.
“it allowed me to fact check anyone who's running our media buying for me”
Andrew Faris relays Haus and CTC analyses finding the true incremental value of ads ran about 10 to 20% higher than 7-day click attribution showed, measured before Meta's March 2026 click change.
“true value contribution of your ads... was about 10 to 20% higher”
CTC says even a well-calibrated incremental ROAS should answer to revenue and contribution margin; if iROAS improves while margin is flat or falling, recheck the measurement.
“Even a well-calibrated incremental return on ad spend should always be subordinate to the realities of revenue and contribution margin”
CTC, which sells geo tests, says across hundreds of its tests Facebook acquisition's median incrementality factor is 1.14, with single tests ranging from 0.5x to 2.4x, so one test is not enough.
“single test results are insufficient”
Wicked Reports, an attribution vendor, says Conversions API purchases should match Shopify orders within a few percentage points, with matching event IDs, and to decide on your own revenue data.
“within a few percentage points”
Meta changed click-through attribution in March 2026 to count only link clicks, to match tools like Google Analytics. Other clicks move to engage-through, and a video engaged view drops from 10 seconds to 5.
“allow your Meta reporting to align better with third-party tools like Google Analytics”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it2 figures from cited sources
- The strongest disagreementThe best case against the answer, from a cited source
- How sure we areWhat its confidence rests on, and what would change the answer