Claim and evidence
Run an open-source marketing mix model and use lift or geo experiments as priors to keep its channel estimates honest.
No dated example yet. Treat this tactic as a hypothesis.
Vendor claim
A tool vendor reports numbers about its own customers or product. Useful, but not independent.
Working · 2026-11 Robyn, Meridian and Recast all document experiment-based calibration; Meta and Google are platform vendors with interests.
What the numbers show
No published number for this tactic has passed our word-for-word check yet, so there is no chart.
Why it works
MMM works from aggregate data and can be biased; experiments give causal anchors. Feeding experiment results into the model's priors or calibration step improves reliability.
How to run it
- Collect weekly spend and outcome data by channel
- Fit an open-source MMM (Robyn or Meridian)
- Run geo or conversion-lift experiments on key channels
- Feed results in as priors or calibration
- Re-run on a regular cadence and compare budget moves
Ways to run it
Design options to test, not promised results.
- Option AGeo experiments as anchors
Turn spend on or off by region for key channels and use the measured lift to calibrate the model.
- Option BPlatform lift studies
Use conversion-lift studies offered by ad platforms where geo tests are impractical.
- Option CTrust filter first
Feed in only experiments whose design you trust; leave out tests you would not act on alone.
Dated examples
No dated example yet. Treat this tactic as a hypothesis.
Does it still work?
Robyn, Meridian and Recast all document experiment-based calibration; Meta and Google are platform vendors with interests. source
Risks
- Needs analytics and data-science skills
- Bad experiments poison the model
- Vendor guidance may favor own platform
Sources
- developers.google.com
“Meridian GeoX is now available. Run geo experiments to calibrate your MMM.”
✓ quote checked word-for-word on the page - github.com
“Robyn is an experimental, AI/ML-powered and open sourced Marketing Mix Modeling (MMM) package from Meta Marketing Science.”
○ page blocks automated checks; quote as relayed - getrecast.com
“If you don’t trust the experimental results, then you shouldn’t be using them to calibrate!”
✓ quote checked word-for-word on the page - socialmediatoday.com
Change log
- 2026-11-20 — harvest v0.1 (ops-ai)