Michael Kaminsky
Co-founder and co-CEO of Recast, a marketing measurement and planning company that sells an MMM. Background in statistics, causal inference and econometrics; 10–15 years in marketing science. Publishes on LinkedIn and the Recast YouTube channel.
Strong on: how MMMs fail, how to validate an MMM (parameter recovery, holdout forecasts, lift-test calibration), when a business does or does not need one.
Bias to keep in mind: an MMM vendor — though he argues against MMM for most small businesses, his validation standards also happen to favour sophisticated (paid) MMM over cheap open-source wrappers.
Sources: Mobile Dev Memo podcast (sources/youtube/mobile-dev-memo/).
Notes citing Michael Kaminsky
- Designing incrementality tests (duration, budget, CTV and brand)Practical rules for geo/holdout incrementality tests — run upper-funnel and CTV tests 4–6 weeks (two weeks shows nothing), give them enough budget not to be swamped, avoid launching into seasonality, and judge brand spend on a finance timeline.
- Evaluating a marketing mix model (and whether Robyn/Meridian are biased)Assume any MMM is wrong until it passes three checks — parameter recovery on simulated data, out-of-sample forecasting of a period with real budget changes, and agreement with lift tests; open-source Robyn and Meridian are not rigged for Meta/Google, but most MMM vendors are thin wrappers around them.
- How last-click attribution biases channel resultsLast-click flatters low-funnel, intent-capturing channels (search, Apple Ads, Google UAC standard) and under-credits view-based ones (YouTube, CTV, chatbot ads); incrementality tests have reversed channel rankings — and platform-reported numbers are biased toward the platform.
- The measurement stack — last-click attribution, incrementality and MMM togetherWhy app advertisers need both deterministic last-click attribution (for platform optimisation and daily decisions) and incrementality testing (for budget allocation), where MMM fits, and how to reconcile them against top-line results.
- When last-click attribution is good enough (and when to invest in incrementality)For small, performance-driven advertisers on one or two channels, last-touch attribution is often ~85% as good as experiments and better than a bad MMM; invest in incrementality tests as spend, channel count and brand effects grow.