Conversion counting: why ad platforms differ

Anyone who manages paid media knows the typical discrepancy: Google Ads reports 400 conversions, Meta another 250, and Microsoft Ads another 60. Added together, that suggests 710 purchases, while finance only books 480 real sales. The obvious suspicion is that someone is distorting the numbers. In reality, nobody is lying. Ad platforms count conversions by their own rules and therefore systematically report higher values than internal business systems. Once this counting logic is clear, deviations become easier to interpret and use.

Why platforms count generously

A central driver is commercial incentive. The more conversions a platform shows, the better its performance looks, the more trust the channel earns, and the more likely additional budget will follow. Between conservative and generous counting, platforms structurally choose the generous option. This is not a conspiracy but rational economics. Teams that understand this interest structure assess reports more calmly and avoid flawed conclusions in budget meetings.

At the same time, the number of real conversions in a period is fixed. Three platforms can claim the same purchase, but the customer still bought only once. Instead of forcing every difference to zero, teams should understand how each platform counts and work with a level of accuracy that improves decisions. Perfect cross-channel reconciliation is rarely the right standard; reliable enough for steering decisions is.

Structural reasons for mismatched numbers

When numbers must be explained to a CFO or internal controlling, concrete mechanisms help more than blanket skepticism. The main drivers of divergence sit in attribution windows, engagement definitions, view-through logic, in-platform attribution models, and the siloed nature of platforms versus analytics and CRM systems.

Attribution windows and engagement rules

Attribution windows are among the biggest factors. Meta defaults to a seven-day click window plus a one-day view window. Google Ads with data-driven attribution can look back much further and credit interactions over longer periods. These different time frames alone cause the same customer journey to be valued differently before any model change is even discussed.

What counts as an interaction also differs. On Meta, carousel swipes, video views, or shares can receive attribution as engagement. On Google Ads and Microsoft Ads, an ad click is generally required. The customer journey stays the same; the credit rules do not. Teams that miss this difference compare metrics that only share a name on the surface.

• Meta: shorter default windows, but often including view-through.
• Google Ads: longer lookback periods and data-driven credit distribution.
• Microsoft Ads: click-based logic with its own reporting structure.
• Internal systems: usually count transactions, not platform credits.

View-through conversions and YouTube effects

View-through conversions are a major source of inflation. Display, programmatic, affiliate, and especially YouTube often count conversions from users who saw an ad but did not click it. Those views are invisible to analytics, ecommerce platforms, and CRM systems. Those systems see clicks or sessions, not pure impression exposure. View-throughs can be useful for optimization and reach steering, but they should not be treated uncritically as proof of retargeting success. A modeled assessment and, where possible, validation through incrementality tests is the better approach.

In-platform models and analytics silos

Even within one attribution window, credit distribution changes the picture. Google's data-driven attribution spreads fractional credit across interactions in the Google Ads ecosystem and uses machine learning for weighting. Meta typically works in a more last-touch way and often assigns a clear single-touch credit. Identical journeys therefore produce different reported values. Simply adding platform numbers can double or triple the same purchase.

Platforms also operate in silos. Google does not see Meta touchpoints, Meta does not see Google touchpoints, and analytics or CRM systems do not see view-through impressions. Each source tells a consistent but incomplete story. That is why reconciliation is not purely a tracking problem but an interpretation problem. The right question is not which platform is lying, but which number should support which decision.

A pragmatic approach to discrepancies

In practice, a clear role model for data sources works best. Platform reports steer budget, bids, and creative tests within each channel. Analytics and CRM provide the cross-channel view for business reporting and forecasts. Finance remains the truth for revenue and contribution margin. Deviations between these layers are expected and should be documented instead of being scandalized in every cycle.

• Document attribution windows and models per platform in writing.
• Report view-through separately from click-through.
• Never equate summed platform totals 1:1 with revenue.
• Regularly mirror platform values against CRM and finance numbers.
• Investigate tracking and event definitions when outliers appear.

A shared glossary inside the team is especially helpful. Terms such as attribution window, view-through, click-through, assisted conversion, and incrementality should be defined consistently. Otherwise marketing, analytics, and finance talk past each other even though everyone appears to look at the same report. A short internal playbook with platform defaults, allowed comparison periods, and escalation rules for outliers saves measurable time in reviews and reduces bad budget-shift decisions.

Teams that treat attribution as decision support rather than accounting work more robustly. They accept that Google Ads, Meta, and Microsoft Ads use different measurement frames and still use the signals for optimization. At the same time, they protect budget decisions by using business outcomes as guardrails. That is the real value for analytics, tracking, and performance marketing: not hunting for one magical number, but reading and steering competing conversion reports correctly.