Media mix modeling
The next dollar,
not the average one.
Bayesian MMM fit to your weekly spend and revenue. Adstock and saturation per channel, posteriors on every contribution, and marginal return ranked so reallocation has a reason.
Response curves · adstock and saturation fitted per channel
Marginal return on next dollar
This ordering, not the average, is what should drive reallocation.
How it works
01
Weekly rows in
One revenue column, one column per channel, fit on your own data — not a shared benchmark.
02
Fit on a worker
Fits queue to a modeling worker and refit on schedule. The page tells you when it is waiting and why.
03
Read the posterior
Baseline share, channel contribution, and marginal return, each with its credible interval.
What you get
Adstock and saturation
Carryover and diminishing returns estimated per channel from your data, not assumed from a benchmark.
Marginal return by channel
Return on the next dollar, ranked. The Budget Simulator reads these same curves.
Intervals on everything
A contribution without an interval is a guess. Every figure shows its width.
Contract checks before a fit
Not enough weeks, not enough variation, or a broken sequence is reported before a number is produced.
What it will not do
Boundaries we state up front, so the numbers are trusted.
- It does not fit on a sample dataset. No data means no model, and the page says so.
- It does not produce a number from a series with too little variation to identify.
- It does not treat a rename as a new channel; fitted output keeps the labels it saw.
Also in the product