Causal lift vs attribution
Attribution says what followed. Lift says what the campaign caused.
Most campaign reports answer “How much revenue happened after people received this email?” That is attribution, not necessarily causation. Some recipients would have purchased anyway, especially in a reorder audience where people already have established buying behavior.
Illustrative shapes, not client data.
- AttributionHow much revenue happened after people received this email? It counts every purchase that followed, including purchases that would have happened anyway.
- IncrementalityHow much additional revenue did this campaign cause? It compares contacted customers with a randomized holdout and counts only the difference.
How it works
A randomized holdout separates cause from coincidence.
- 1The eligible audience is identified.
- 2A small random portion is intentionally not sent the campaign.
- 3The remaining eligible customers receive the campaign through the client's existing Klaviyo process.
- 4After the appropriate purchase window, LiftMiner compares revenue and conversion between the contacted group and the holdout group.
In plain language
The number worth judging a campaign by.
Example
If the contacted customers generated $20,000 more revenue than a comparable group that did not receive the campaign, LiftMiner can estimate the campaign's true incremental contribution.
What incrementality testing is
The standard controlled-experiment approach for distinguishing a campaign's causal impact from purchases that would have happened without exposure. It compares a treatment group with an unexposed control group. Typical practical holdouts are around 5–10%, subject to available audience size and statistical power.