Your Meta dashboard shows 50 conversions. Google Ads claims 32. Your CRM logged 42 actual sales. Same campaign, same week, same audience. Three numbers, zero agreement.
This is not a bug. It is not platform fraud. It is the predictable outcome of three measurement systems, each built on different assumptions about what counts as a conversion and when credit should be assigned. Until your team models these differences explicitly, every budget reallocation meeting will devolve into a debate about whose numbers are "right." Spoiler: everyone's numbers are right, by their own rules. The question is whether you have a shared model that reconciles them.
Different Rules, Same Customer
The core issue is that each platform operates its own attribution logic, and those logics overlap in ways that guarantee double-counting when you sum across platforms.
Meta Ads defaults to a 7-day click window and a 1-day view window. If someone clicks your Facebook ad and converts within seven days, Meta claims it. If they merely saw the ad and converted within 24 hours, Meta still claims it. Google Ads, by contrast, defaults to a 30-day click attribution window, more than four times longer than Meta's click window.
Now picture a real customer journey: someone sees your Facebook ad on Monday, clicks your Google search ad on Wednesday, and purchases on Friday. Facebook claims the conversion because the purchase happened within its view window. Google claims it because the purchase happened within its click window. Your CRM records one sale. Both platforms are telling the truth by their own definitions. The problem is that their definitions were never designed to be additive.
This is not a rounding error. Each platform essentially plays by its own rules in assigning credit, and when you add up all the conversions each platform reports, you get a number higher than your actual sales. For a CMO presenting to the board, this creates an uncomfortable choice: report the inflated sum and look naive, report the CRM number and lose channel-level visibility, or build a reconciliation model and explain the assumptions.
The Technical Layer Underneath
Attribution windows are only half the story. The tracking mechanisms themselves introduce additional variance.
Meta primarily relies on pixel-based tracking, a small piece of code on your website that fires when someone takes an action and sends that information back to Meta's servers. Google Analytics uses a similar approach but processes the data differently. Your CRM might use server-side tracking that records conversions directly from your backend systems without relying on browser-based pixels at all.
Each method has failure modes. Pixel-based tracking can be blocked by ad blockers or browser privacy features. Server-side tracking bypasses these limitations but requires more technical setup and may miss certain touchpoints. Cookie-based systems depend on users accepting cookies and maintaining them across sessions. When someone clears their cookies or uses incognito mode, the tracking chain breaks.
Meta's documentation acknowledges this directly: third-party reporting platforms may not measure cross-device conversions well due to cookie-based measurement, while Meta's people-based measurement attempts to track behavior across devices. This sounds like an advantage for Meta, but it also means Meta will claim conversions that a cookie-based analytics tool cannot verify. Neither is wrong. They are measuring different things.
Privacy Constraints Are Making This Worse
The gap between platform-reported conversions and verifiable sales is widening, not narrowing. Apple's App Tracking Transparency framework and similar privacy regulations have made it more complex to track individual user actions across different apps and websites. Marketers must now balance consented, granular, user-level data with privacy-centric, aggregated data.
This means platforms are increasingly relying on modeled conversions, statistical estimates of conversions that could not be directly observed. Google Ads explicitly includes modeled conversions in its primary "Conversions" column. Meta uses similar techniques. These models are not fabrications; they are probabilistic inferences based on patterns in the data the platforms can observe. But they introduce another layer of divergence from your CRM's ground truth.
For B2B marketers with long sales cycles, this creates a specific problem. A 30-day attribution window might capture the initial touchpoint, but if your average deal takes 90 days to close, the platform's conversion count will systematically understate the true impact of upper-funnel campaigns. Conversely, if you are running retargeting campaigns that touch prospects multiple times before close, every platform in the mix will claim credit for the same deal.
Building a Reconciliation Model
The solution is not to pick a "source of truth" and ignore the others. It is to build a model that maps platform-reported conversions to actual revenue, with explicit assumptions about overlap and attribution.
Start with your CRM as the anchor. This is the only system that records actual closed revenue. Then work backward: for each closed deal, identify which platforms claimed credit and with what attribution logic. Over time, you will develop platform-specific conversion-to-revenue ratios that account for double-counting and tracking gaps.
For example, you might find that Meta-reported conversions convert to CRM-verified revenue at a 0.6 ratio, while Google-reported conversions convert at 0.8. These ratios are not fixed; they will vary by campaign type, audience segment, and funnel stage. But they give you a defensible basis for budget allocation that does not depend on taking any single platform's numbers at face value.
The second step is to align attribution windows where possible. If you are comparing Meta and Google, consider setting both to the same click window, even if it means deviating from platform defaults. This will not eliminate discrepancies, but it will reduce the portion of variance attributable to window differences versus tracking differences.
Third, run incrementality tests. Geo-holdouts or matched-market experiments can tell you what actually happens to revenue when you turn a channel off, independent of what the platform claims. This is the only way to validate your reconciliation model against real-world outcomes.
The Board Conversation
When you present marketing performance to the board, lead with the reconciliation model, not the raw platform numbers. Show the assumptions: which platforms are included, what attribution windows are in effect, what conversion-to-revenue ratios you are applying, and what the confidence interval looks like.
This is not about admitting uncertainty. It is about demonstrating that you understand the measurement system well enough to make decisions despite its limitations. A CFO will trust a CMO who can explain why the numbers diverge far more than one who pretends they should match.
The platforms are not lying to you. They are each telling a partial truth, optimized for their own incentives. Your job is to build the model that translates those partial truths into a forecast you can defend.
