Article · 16 min read

Incrementality Testing: The Only Guide You Need

Olivia Kory and Erkki Paunonen · Supermetrics Free Advanced

The other half of the answer: how to actually find out whether a channel caused anything, using holdouts, geo tests and lift studies. It is also honest that click attribution flatters brand search and retargeting, which convert regardless.

  • Below roughly $5M a year in paid media you do not need geo or platform lift tests; observational experiments are enough.
  • Three methods: platform conversion lift with a holdout, geo splits, and before-and-after natural experiments.
  • Feed results back into reporting: if only 50 percent of platform conversions were incremental, double your reported CPL and CPA.
  • Re-test your big bets quarterly, or at least every 3 to 6 months, since incrementality decays.
Open on supermetrics.com ↗

The question it answers

  1. Marketing says it sourced the pipeline, sales says it did not. How do I do attribution without fooling myself?

Read with it

  1. Shopify's performance marketing lead on channel order (Google search first, LinkedIn reserved for tight B2B targeting because... Lenny's Newsletter Podcast · 75 min listen The ultimate guide to performance marketing Timothy Davis with Lenny Rachitsky
  2. The single most quotable attribution finding we know of: across 620 declared-intent conversions and 21.5 million dollars of... Refine Labs Report · 12 min read Refine Labs Study Confirms Measurement Gap in Software-Based Attribution & Releases Hybrid Attribution Framework Refine Labs
eChai Partner Brands The partnership →