Run your own MMM and geo experiments. Bring only what you know about the business.
Open-source marketing mix models and geo holdout experiments, running on your own data in your workspace. The platform provides the data plumbing, compute and apps; you provide the business knowledge that makes a model right.
Most MMMs are expensive, and built without the business.
MMM tools cost tens of thousands a year, and most of that pays for infrastructure. The models that come out are often built by people who never learned the business: no margins, no promotions, no experiments, so the curves look precise and mean little.
Replaces MMM vendor contracts, one-off agency models and spreadsheet incrementality estimates.
- A marketing mix model on your own data, refitted every week
- Geo holdout tests designed with market selection and power analysis, then read out
- Experiments that calibrate the model instead of arguing with it
- Response curves and a budget plan your finance team can audit
How it works
- 01
Connect spend and outcomes
Ad spend, revenue and orders by channel and geo, prepared from your connectors every week.
- 02
Add what you know
Margins, known events, seasonality and the experiments you trust, entered in the app instead of buried in a notebook.
- 03
Run geo tests
Market selection and power analysis pick test and control regions; the read-out comes back as a calibration prior.
- 04
Fit, refit, plan
The model refits on schedule; contributions, response curves and a budget plan land in the app and the warehouse.
Enterprise, with our engineers
Our engineers set up the model with you: they prepare the data, agree the business inputs and priors, design the first geo test to calibrate it, and walk your team through the results until you trust them.
Self-serve, after launch
Install Marketing Mix Modeling and Incrementality Lab, connect your ad platforms and revenue, and fill in the business inputs. Open-source models (PyMC-Marketing, Google Meridian, Meta Robyn, GeoLift) run on platform compute; you pay only for what runs.
Install from the marketplace
Everything below installs into your workspace as code. Use it as it is, or change anything.
Marketing Mix Modeling
Channel contribution for spend attribution cannot see.
App
Incrementality Lab
Geo holdout experiments that prove what a channel adds.
App
Budget Optimizer
Marginal ROAS per channel and the budget mix that maximises it.
App
PyMC-Marketing
Bayesian MMM and CLV models, the default engine for our MMM app.
Extension · PyMC Labs
Google Meridian
Google's open-source Bayesian MMM, on your own data.
Extension · Google
Meta Robyn
Meta's open-source MMM, run on your own data.
Extension · Meta
GeoLift
Geo holdout design and read-out, the engine behind Incrementality Lab.
Extension · SegmentStream (based on Meta GeoLift)
CausalPy
Causal effects from launches, price changes and regional campaigns.
Extension · PyMC Labs
Google Trimmed Match
Google's robust method for matched-pair geo experiments.
Extension · Google
Questions
Why not buy an MMM tool?
An MMM vendor typically charges tens of thousands of dollars a year, and much of that pays for infrastructure and data plumbing. Here the open-source models (Google Meridian, PyMC-Marketing, Meta Robyn) run on your own data in your workspace, and you pay only for the compute they use.
What do we need to provide?
Your business knowledge: what a conversion is worth, margins, promotions, outages and seasonality you know about, and the experiments you trust. Most MMMs go wrong because nobody supplied these; the app asks for them up front.
Can we calibrate the model with experiments?
Yes. Incrementality Lab designs geo holdout tests (market selection, power analysis), reads them out when they end, and feeds the measured lift into the MMM as a calibration prior.
Can we change the model?
Yes. The models, inputs and outputs are code and tables in your workspace. Swap PyMC-Marketing for Meridian, add a channel, or change a prior, and the next refit uses it.
Rather have it built for you? Start a pilot.
Our engineers set up the model with you: they prepare the data, agree the business inputs and priors, design the first geo test to calibrate it, and walk your team through the results until you trust them.
- One solution, on your data, in your own workspace
- First version live in two weeks
- Parity proven before anything is switched off
- Fixed scope and price, credited to your contract if you continue
Not ready for a pilot? Send the form anyway with your question. An engineer will reply, not a sales rep.
How the pilot worksYour company’s context, ready for every agent.
$ segmentstream workspace create acme --region eu
▸ workspace ready · data, objects, metrics and access rules in one place
▸ connect Claude, ChatGPT or Cursor with one connector
▸ tell your agent: “add our billing API as a source”
Our engineers build the first version on your data in two weeks, at a fixed price.
Or join the waitlist for self-serve:
Pilots are open now. Self-serve opens to the waitlist this fall.