You launch the same product on Meta and TikTok. Same creative angles, same audience demographics, similar daily budget. Meta reports €15 CPA. TikTok reports €45 CPA. Either Meta is amazing or TikTok is broken — which is it?
Neither. Both platforms are reporting numbers that are partially true and partially distorted. Here's what's actually happening with attribution in 2026 and which numbers to trust for decisions.
How attribution actually works in 2026
When a customer buys from your store, multiple platforms claim credit:
- Meta claims credit if they saw a Meta ad in the last 7 days (1-day click + 7-day view window standard)
- TikTok claims credit if they saw a TikTok ad in the last 7 days
- Google claims credit if they clicked a Google ad in the last 30 days
- Email claims credit if they opened a Klaviyo flow recently
- Direct claims credit if they typed your URL directly
All of these can be true simultaneously for the same customer. The "platforms attribution sum" often exceeds 100% of actual conversions because each platform thinks they should get credit.
Meta's attribution model
Meta uses a privacy-aware attribution model since iOS 14.5 (2021). The key features:
- 7-day click window: Counts conversions within 7 days after ad click
- 1-day view window: Counts conversions within 24 hours after ad view (impression only, no click)
- Aggregated Event Measurement: Privacy-preserving conversion data from iOS users
- Modeled conversions: Statistical estimation when direct attribution is blocked
Meta tends to over-attribute because:
- View-through conversions inflate counts (people who didn't click but bought anyway get credited to Meta)
- Modeled conversions add statistical estimates that may include conversions Meta didn't influence
- Cross-device tracking helps Meta connect ad views on mobile to conversions on desktop
Typical over-attribution: Meta reports 20-35% more conversions than the brand actually got from Meta exclusively.
TikTok's attribution model
TikTok's system is younger and operates with different constraints:
- 7-day click window: Similar to Meta
- 1-day view window: Similar
- Events API: TikTok's equivalent to Meta's Conversion API
- Less modeling sophistication: TikTok has less data than Meta to model around privacy gaps
TikTok tends to under-attribute because:
- Less mature event tracking misses some conversions
- iOS attribution gaps are larger (TikTok hasn't fully closed these)
- Cross-device tracking is weaker
- View-through attribution exists but is less reliable
Typical under-attribution: TikTok shows 30-50% fewer conversions than the brand actually got from TikTok.
The 3x CPA difference unpacked
Now the math makes sense. Same campaign, same actual performance:
- Meta reports €15 CPA, actual is probably €18-22 (real CPA worse because Meta over-credits itself)
- TikTok reports €45 CPA, actual is probably €25-30 (real CPA better because TikTok under-credits itself)
The reported 3x difference might actually be a real 1.3-1.5x difference. Both platforms are profitable, but you might be optimizing the wrong one because you trust the wrong numbers.
The blended approach that works
Sophisticated brands ignore platform-reported attribution and use blended metrics instead.
MER (Marketing Efficiency Ratio):
- Total revenue / total ad spend across all platforms
- For example: €100,000 revenue / €25,000 ad spend = MER of 4
MER tells you the truth because it can't be over-attributed. Revenue is real, spend is real.
aMER (Adjusted MER):
- Same as MER but adjusts for organic baseline
- Subtract your organic revenue baseline before calculating
- Example: €100,000 total - €20,000 organic = €80,000 paid-attributable / €25,000 spend = aMER of 3.2
aMER is harder to calculate but more accurate for understanding pure paid contribution.
Incrementality testing:
- Periodic tests where you pause one channel and measure total revenue change
- Reveals the actual contribution of that channel
- Most accurate but operationally complex
Practical decisions with conflicting data
When Meta says one campaign is great and TikTok says another is bad, but blended MER suggests they're contributing equally, do this:
If blended MER is healthy and trending up: trust the overall direction, don't over-optimize individual platform numbers
If a platform shows worsening reported numbers but MER is stable: probably attribution noise, don't react
If reported numbers worsen AND blended MER drops: real issue, investigate
If you cut a "bad" platform and blended MER drops more than expected: that platform was contributing more than its reported numbers showed
Setup that produces better data
To improve attribution accuracy:
- Conversion API on both platforms: Server-side event tracking helps recover iOS attribution losses
- Proper deduplication: Same conversion shouldn't fire twice (server + client). Use proper event IDs.
- Consistent UTM tagging: Helps with cross-platform attribution validation through Google Analytics
- Server-side analytics: Triple Whale, Northbeam, or similar tools that aggregate first-party data across platforms
These don't fix the fundamental attribution problem but reduce the noise and produce more useful relative comparisons.
What to do this week
If you're comparing platform CPAs to decide where to invest:
- Calculate your blended MER for last 30 days: total revenue / total ad spend
- If blended MER is healthy: don't over-react to platform reporting
- If blended MER is weak: investigate which platform is the actual problem (run a pause test on each)
- Set up Conversion API properly on both platforms if not already
If you've been making decisions based on platform-reported CPA alone, you're probably allocating budget suboptimally. Switching to blended metrics typically reveals that platforms are closer in actual performance than reported.
Prime Scale Media supports brands running both Meta and TikTok with attribution-aware agency account infrastructure. Discuss your multi-platform attribution on WhatsApp.