Meta showed 5.4x ROAS. After matching to orders, it was 1.2x.

dusting

Pulled a week of data for an apparel store to compare Meta's reported purchase value with orders I could deterministically tie back. the delta was eye-opening. Meta's dashboard: $41k purchase value, $7.7k spend = 5.4x ROAS. After matching to actual orders: $9.5k revenue, same spend = 1.2x ROAS. Factor in COGS at 60% margin, and you're looking at a loss of roughly $2k a week. breakeven ROAS would need at least 1.7x.

this isn't a "Meta is useless" take, but a gap that wide completely shifts optimisation and budget allocation. A customer clicks a Meta ad on Monday, returns via email Thursday, buys Saturday. Shopify or GA4 may credit email or direct, while Meta still claims the conversion. Comparing dashboards solves nothing.

The method: stitch click IDs, UTMs, session history, customer identifiers, order data, then track the customer across sessions until purchase. without that, you're comparing one attribution model against another with black-box data.

interestingly, Google Ads was underreporting revenue by about 40% against order-level data. So the issue cuts both ways.

Some colleagues suggest splitting Meta conversions by click-through and view-through, noting that default 24-hour view-through attribution inflates numbers. Others argue the backend data should be the source of truth, looking at incremental lifts rather than trying to match every Meta-influenced purchase. That's a more macroeconomic take - don't reconcile, hypothesise from directionality.

Anyone else doing attribution comparisons at order level? What deltas are you seeing between reported and real ROAS?

justauser

i've seen exactly this pattern. Splitting out click-through and view-through conversions makes a massive difference-Meta will happily claim a sale if someone just glanced at an ad without clicking, as long as they bought within the default 24-hour window.

plus, tracking and UTM parameters have this nasty habit of falling off or breaking, so you end up with even less reliable data. Every time I've done a deep dive into what Meta reports versus what actually happened, it's been inflated. Worth checking your own numbers if you haven't already 👀

growthgeek

Anyone who takes Meta's reported ROAS at face value is either lying to themselves or their boss. That 5.4x number? Pure vanity. i treat Meta's attribution as noise - a directional signal at best, just to see if the platform is even bothering to register any conversions.

Source of truth has to be your own backend data, no debate there. But don't waste time trying to stitch every single order back to a Meta click. That's a fool's errand. Look at incremental lift over a control group. Form a hypothesis, run the experiment, measure the real delta. anything else is just performance theatre.

bigcoffee

That gap is brutal. A couple of things I'd dig into first:

How many purchase events are you actually sending? Low volume inflates Meta's reported ROAS because the algorithm has fewer conversion signals to work with. Seen plenty of accounts where the pixel fires on <50 purchases per week and the reported number is complete fantasy.
What's the event quality score on those purchases? If it dips below 8.0, Meta likely isn't using them for optimisation properly. You can check that in Events Manager under diagnostics.

Also worth confirming-are you running a server-side feed alongside the pixel, or just relying on the browser event? That mismatch alone can cause the gap you're seeing.

metricsmadness

it's the old story of vanity metrics versus ground truth. have you sorted out event deduplication and got CAPI running in tandem with the pixel? Without that, the surface-level ROAS is almost always inflated

neonlights

mate, if your attribution model isn't counting that first touch properly, you're just leaving 4x+ ROAS on the floor.

Had a client last quarter: Meta showed 6x ROAS, actual matched orders came in at 1.8x. drove me nuts until we fixed the tracking to credit the initial channel, not just last click.

That "1.2x" number is telling you your attribution is broken, not that Meta is underperforming. Direct and email don't exist in a vacuum-they're just closing what Meta opened. Fix the model or kiss your scale goodbye.

ctrjunkie

Honestly, this is the kind of reality check every marketer needs to run at least once. that $41K vs $9K gap? Classic view-through attribution trap. Meta's default window basically credits itself for any conversion where someone saw an ad within 24 hours, didn't click a thing, then bought later through email, Google, or just typing the URL directly.

Before writing the platform off entirely, there's a couple of things worth checking. Pull your click-through ROAS separately from view-through in Ads Manager - go to Columns > Compare Attribution Windows. that click-through number usually feels a lot closer to what you're actually seeing in your bank account. Also, double-check if you're sending both server-side and pixel events. shopify's native reporting sometimes double-counts when CAPI and the pixel both fire for the same purchase. Messy.

The real lesson here isn't "Meta is lying." It's that blending view-through and click-through into one ROAS number hides everything happening under the hood. Meta's actually pretty decent at driving awareness - it just doesn't get last-click credit. Meanwhile, Google underreports because people search branded terms after seeing a Meta ad. Separate the two, weight them differently, and suddenly your allocation picture makes a lot more sense.

out of curiosity - what attribution window are you using? Default 7-day click + 1-day view, or did you narrow it? happy to take a look at your account setup if you want a second pair of eyes. Free screenshare, no strings attached.

maya-b2b

That's not quite the same metric though - ROAS versus POAS (profit on ad spend) tells a very different story.

But here's the thing: Meta might still have been the reason that prospect made the purchase. They see your advert on Meta, then search for your brand on Google. google picks up the conversion in Shopify, but Meta was the first touchpoint - so Meta also gets credited with a conversion.

If you're serious about accuracy - and it sounds like you are - look into tools like Reaktion or Triple Whale. The cost is minimal compared to the quality of data you unlock

leadgen

Oh, I feel this one in my bones. After chasing those shiny Meta dashboard numbers for months, I finally got fed up and built my own first-party pixel on Azure. The difference was night and day - suddenly I was looking at a consistent 9+ EMQ instead of the garbage 5.4x they kept showing me. It's absurd we have to hack together our own tracking just to see the truth, but here we are.

24fire

Meta can still serve a purpose, but matched order data and contribution profit are the only true north

ethan-pr

we run an ecommerce brand and have seen this exact pattern play out. Platform ROAS is never the final number. my team always sanity-checks against contribution profit - net sales minus COGS, shipping, fees, and ad spend. The real scare is when a campaign looks strong inside Meta but quietly erodes cash margin

coopplayer

honestly, your deterministic match is just as flawed - cross-device journeys, iOS blind spots, cleared cookies.. that gorgeous 5.4x Meta painted is a fantasy, but scraping the bottom with 1.2x ignores the invisible browsing path. real number sits somewhere in that hazy middle, not at either extreme