TikTok's new attribution correction cuts ad cannibalization by 15%
Paid ad conversions often overstate true growth due to overlapping organic demand — now there's a fix.
In large-scale advertising, paid-attributed conversions like daily new users (DNU) can systematically overstate true incremental growth because they overlap with organic demand, brand-driven traffic, or other channels. This ‘attribution-cannibalization mismatch’ distorts ROI measurement and budget decisions. TikTok researchers present a framework that uses incrementality experiments as causal anchors to convert sparse lift measurements into daily correction estimates. They further allocate the calibrated cannibalization volume across business hierarchies under structural consistency constraints.
Offline forward-in-time validation against channel-level incrementality experiments showed the framework substantially reduces calibration error compared to raw attribution and ML baselines. When deployed across multiple global TikTok markets, the system enabled budget and traffic strategy adjustments that led to an approximately 15-percentage-point reduction in the measured cannibalization rate. This approach gives advertisers a practical, production-grade tool to distinguish truly incremental conversions from those that would have happened anyway.
- Uses incrementality experiments as causal anchors to convert sparse lift data into daily correction estimates for attribution.
- Deployed across multiple global TikTok markets, reducing measured cannibalization rate by approximately 15 percentage points.
- Outperforms raw attribution and fine-grained ML baselines in forward-in-time calibration tests.
Why It Matters
Helps advertisers allocate budgets more accurately, reducing wasted spend on non-incremental conversions and improving ROI.