Every platform claims the same conversion: why ad platforms overreport

You add up Meta's conversions, Google's, and TikTok's, and the total lands higher than the number of customers you actually have. Nothing is broken. Ad platforms overreport because each one attributes conversions independently, inside its own window, by its own rules, view-through included. A customer touched by three channels is counted three times, once per dashboard, and nothing deduplicates across platforms. Summed dashboards exceed reality by design: platform numbers are optimization signals, and the real count lives in your store, CRM, or warehouse.

This is written for the moment someone asks you which of those numbers is real. The short answer is that each platform is grading its own homework, and once you can see that, you stop trying to reconcile the dashboards and start giving each number a job.

#The arithmetic that breaks trust

Nobody experiences this as "working as designed." Here is how it actually arrives: a founder adds up Meta's 60, Google's 55, and Klaviyo's 40, gets 155 "conversions" in a month with 90 orders, and asks the only reasonable question (which of these numbers is real?), usually of whoever is spending the money. It is a credibility break, and it lands on marketing. The honest answer ("none of them, and all of them, depending on the question") sounds like evasion unless you can show the mechanics. So here they are, traced through a single conversion.

#One conversion, three claims

Here's a single purchase as the systems record it. One customer: clicked a Google Search ad Tuesday, saw a Meta retargeting ad Friday afternoon (no click), opened a promo email Friday evening, bought Friday night. One order in Shopify. Now the claims: Google books a click conversion on Tuesday's click date (7-day click window), Meta books a view-through conversion from Friday's impression (1-day view window), and the email tool books an open-attributed conversion Friday. Three dashboards, three conversions, one order.

[DATA (vault article): real worked-example table rows replace this hypothetical.]

#Why this happens: three mechanisms, no conspiracy

Last-click-within-my-window, per platform. Each platform answers "did I touch this conversion within my window?", never "was I the reason?" and never "did someone else also touch it?" Platforms don't see each other's touchpoints and wouldn't yield credit if they did. Attribution here is self-graded per player.

View-through conversions. Meta and display channels count users who merely saw an ad and later converted. View-through isn't worthless (ads work without clicks), but it's the most inflationary counting rule in the stack, because impressions are abundant and cheap to associate with purchases that were happening anyway.

Window asymmetries. A 7-day-click/1-day-view platform and a 30-day-click platform are playing different games on the same field. The longer the window, the more conversions "belong" to a channel, which is also why comparing ROAS across platforms with different windows is numerically meaningless.

#It's an incentive, not a lie

Here's the position: overreporting is the incentive structure of self-graded homework, and calling it fraud misses what the numbers are for. A platform's conversion measurement exists to feed its own delivery optimizer: its bidding systems need dense, fast, self-referential signal about which impressions lead to conversions it can see. Grading generously is what makes the optimizer work. The number was never designed to be your bookkeeping; you promoted it to bookkeeping when you put it in a board slide.

So we refuse both of the usual framings. "Platforms lie" is too cynical to be useful: the numbers are honest answers to a question you didn't ask. "Just trust the platform" is innumerate: independently graded claims always sum past the real total. The workable stance is in between, and it assigns each number a job.

#What each number is actually for

Platform-reported conversions are optimization signals. They tell the platform's bidder (and you, at the campaign level) what's working within that channel. Feeding them well is precisely the game described in Smart Bidding is a signals problem: what the algorithm actually runs on; starving them breaks delivery. Read them for direction, per channel, never summed.
Your store / CRM / warehouse is the count. Orders, pipeline, revenue, recorded once and deduplicated by construction. Every reconciliation in this cluster ultimately reconciles to this layer.
Cross-channel credit, the question "which channels actually earned the budget," belongs to attribution and triangulation on your own data, where every channel's touches sit in one place and no player grades their own homework.

That hierarchy is the whole fix. The teams that stay sane pin dashboards to jobs: platform numbers steer platform knobs; the warehouse settles counts; attribution splits credit.

#When overreporting IS a bug

Everything above assumes clean tracking: structural overcount on top of correct events. Two patterns mean you have an actual defect instead: a platform counting more events than exist (pixel + CAPI dedup miss, duplicate tags, triaged in Why Facebook Ads conversions don't match Shopify and Why your GA4 and Google Ads conversions don't match), or a platform's claims jumping without a matching change in orders (a ratio break, not a level: tracking changed, customers didn't). The tell in both cases: structural overcount is stable; bugs move. The recurring checks that catch them early live in The conversion tracking QA checklist: test it like you'd test code.

#So how do you split credit?

Not here, and not by any platform's dashboard. Splitting credit is an attribution problem: multi-touch models on warehouse data, incrementality tests where the stakes justify them, and triangulation between what platforms claim and what the count shows. That's its own discipline with its own pages. Start with marketing attribution.

What Buron ships today is the ground that discipline stands on: the attribution datasets put platform-claimed conversions and warehouse-credited conversions side by side, per channel, continuously. It is the one-conversion-three-claims picture above turned into a standing view instead of a quarterly forensic exercise. When the claimed-to-credited ratio moves, that's a finding in your inbox, and when the board asks which number is real, the answer is one screen instead of a week of reconciliation.

Frequently asked questions

Why is my ROAS different in every platform?

Each platform computes ROAS from its own attributed conversions (counted within its own window, by its own model, including view-through) divided by its own spend. Since every platform claims any conversion it touched, the same order inflates several ROAS numbers at once. They're optimization signals, not comparable accounting.

Why do my ad platforms report more conversions than I have orders?

Because attribution is claimed independently: a customer who saw a Meta ad, clicked a Google ad, and opened an email gets counted by all three. Nothing deduplicates across platforms, so the dashboard sum exceeds reality by design. Count orders in your store, CRM, or warehouse, never by summing dashboards.

Which platform's conversion numbers should I trust?

Trust each platform about itself, for optimization: its numbers steer its bidding well. For the actual count, trust the system that records transactions: your store, CRM, or warehouse. For splitting credit between channels, no platform is the referee; that's an attribution and triangulation job on your own data.