A glass box, not a black box.

The model shows what it believes, channel by channel: saturation curves, contributions and confidence. If a number surprises you, you can see exactly where it came from.

Book a demo See what's inside

app.odins.ai / saturation curves

Response by spend level

Last 12 months

Search Social TV

What the model estimates

Baseline first. Then every channel.

The model separates what would happen anyway from what marketing drives, then splits the marketing effect by channel.

Baseline: seasonality, price, promotions and market conditions, controlled for explicitly
Saturation curves per channel, time-varying rather than locked to a multi-year average
Adstock: TV works for weeks, search converts in hours; each channel gets its own lag
Every estimate carries a confidence range, never a false point estimate

Modeled revenue by source

Feb – Jul 2026

Baseline (not media-driven)45.0%

Social Media16.0%

Search15.0%

TV9.0%

Display Advertising8.0%

Affiliate Marketing7.0%

Baseline is demand not driven by media: seasonality, price and market conditions.

Uncertainty is information

Wide range? Test. Tight range? Optimize.

The model separates what it knows from what it doesn't. Where the data can't resolve an answer, it says so, and we propose a structured test instead of a guess.

Confidence ranges on every channel, budget and forecast
Wide bands become test plans with cost boundaries
Tight bands become reallocation you can defend

Channel contribution, 80% ranges

from the posterior

Search 14 to 16% optimize

Social 9 to 13% optimize

Display 2 to 9% test first

A tight range is a decision. A wide range is a test plan, with cost boundaries.

The model feeds everything else.

Reports, scenarios and recommendations all read from the same model, so the numbers always agree.

Data integration

Every channel, online and offline, structured into one dataset.

See data integration

Reporting and insights

Spend and business results in one view.

See reporting

Scenario planning

Test a budget before committing it, with a confidence range on every outcome.

See scenario planning

Recommendations

Where to invest more, where to pull back, and what to test, every month.

See recommendations

Retrained monthly

New data flows in daily; the model retrains monthly so saturation curves follow reality, not a three-year average.

FAQ

Frequently asked questions.

What is marketing mix modeling?

A statistical approach that measures how each channel contributes to business results, using your spend and outcome history rather than tracking individuals.

How is it different from attribution?

Attribution follows clicks and cookies, so it misses offline channels and gets weaker as tracking degrades. MMM works on aggregate data and covers every channel, online and offline.

What does the model actually estimate?

A baseline (what would happen without marketing), each channel's contribution, its saturation curve and its time lag, plus the effect of price, promotions and seasonality.

How many channels can it handle?

Typically 10 to 30 activity types per customer, depending on how the media plan is structured.

Can we see inside the model?

Yes. Saturation curves, contributions and forecast-vs-actuals are all visible in the platform, so every number can be traced to its source.

Is the methodology the same for every customer?

The approach is consistent; the model is not. Channels, priors and control variables are set up per customer, with your team.

How does it compare to Google Meridian or Meta Robyn?

Same family of methods. The difference is that Odins is operated for you: data, retraining, verification and recommendations, without staffing a data science team.

Can a channel's response curve change through the year?

Yes. In the Odins model the curve moves with the season: a channel can have a higher ceiling and a higher optimal spend in November than in July. The details, and how the open-source frameworks compare, are on a separate page: time-varying saturation.

How does the model handle brand and long-term effects?

Where a sizeable share of spend is brand-building, the model splits each channel's effect in two: sales that land within weeks, and a stock of brand value that pays out sales over months and years. The details are on a separate page: brand in marketing mix modeling.