Statistical consulting specialties
Bayesian media-mix model audits
Would you move next quarter’s ad budget if you changed the model’s priors?
For marketing leaders acting on a model built in Google Meridian, PyMC-Marketing, or Stan. I reproduce the proposed budget shift and trace it back to the evidence for each channel.
Request a media-mix auditWhat the work covers
- Change the priors
- Refit the model with several plausible beliefs about channel returns. Compare what the model assumed before seeing the data with what it learned and see which budget decisions change.
- Separate the channels
- When search and social spend rise together, several explanations can fit the same sales history. I check whether the data distinguish their effects and whether the answer depends on carryover, saturation, or seasonality.
- Check against outside evidence
- Where available, I compare estimates with lift experiments the model did not use for calibration. I flag budget proposals that extrapolate beyond observed spend.