Cassandra vs Recast: a Bayesian MMM toolset, or an AI marketing brain that does the work
Recast describes itself as incrementality measurement: a causal, Bayesian marketing mix model and GeoLift by Recast, with forecasting and budget planning. Cassandra, an AI marketing brain for growth teams, measures with a Bayesian model checked by live geo experiments and does the work.
Both feed lift results into a Bayesian model. Recast presents a planning toolset configured with its team or a certified Operator; Cassandra uses what it learns to do the work.
What Cassandra is
- An AI marketing brain for growth teams: it learns what drives growth and does the work.
- Measures with a Bayesian marketing mix model checked by live geo experiments.
- Does the work within the rules the team sets: plans, budgets, ads, creative, reports.
- Approvals: the team decides what waits; every change is recorded.
How this comparison was made
Everything about Recast comes from its own public pages, read on 12 October 2026. We have not tested their product. If anything is out of date, write to hello@cassandra.app and we will correct it.
When Recast is the better choice
- Hands-on setup: Recast says each deployment is configured for your business by its team or a certified Operator.
- Weekly refreshes: Recast says it re-estimates every channel weekly, with live out-of-sample forecast accuracy reports.
- Building on the model: Recast offers versioned public APIs, an MCP server and open-source report templates.
- Low-cost lift testing: GeoLift by Recast has a six-month free trial, as listed on 12 October 2026.
What each one says it is
- Recast's homepage lists a Bayesian MMM, GeoLift experiments, forecasts and plans, budget optimisation, APIs and an MCP server.
- Cassandra: an AI marketing brain that learns what drives growth and does the work.
- Both pair a Bayesian model of the mix with geo experiments; the difference is emphasis.
Where the methods line up
- Bayesian model: Recast says it is "Powered by a proprietary Bayesian MMM"; Cassandra uses a Bayesian MMM too.
- Geo experiments: GeoLift by Recast designs and analyses them with synthetic controls and reports confidence intervals.
- Lift tests: Recast says it calibrates estimates to every lift test; Cassandra's live geo experiments check its model.
- Attribution: Recast's GeoLift page says "Attribution is fundamentally flawed"; Cassandra checks it against measured lift.
What each publishes about pricing
Recast's documentation lists three ways to access GeoLift by Recast (as listed on 12 October 2026):
- Free trial: six months, no credit card needed.
- Subscription: $100 a month or $1,000 a year.
- With the Recast MMM: access as part of it.
Recast MMM: its site invites you to book a demo; Recast notes certified Operators may bring different pricing models.
Cassandra: start for free with your website and data, or book a demo. Paid plans scale with yearly ad spend.
A question for both: how priors are set and changed
- Why it matters: priors are stated beliefs about each channel's effect, and they move the answer.
- Recast's documentation: priors are set at onboarding from business intuition and past data; key priors are viewable.
- Cassandra: live geo experiment results check the model; every move is checked against its forecast.
- Ask both how a past geo test enters the next model, and who can see and change priors.
Limits worth knowing
- Spend history: a channel needs enough spend, and variation in it, for the model to read.
- Regions: geo experiments need regional structure and a comparison group.
- Time: tests need a few weeks; keep test regions free of other changes.
Cassandra, an AI marketing brain for growth teams, learns what drives your growth and does the work within the rules your team sets. Start for free with your website and data, or book a demo.
What moving between two similar platforms involves
- Mostly data and priors: the cleaned history and experiment results move with you.
- Record the priors: which were used, who set them and why, to reproduce or change them deliberately.
- Run both for a cycle: with similar methods, big gaps usually trace to one prior or data choice.
How to choose
- Recast: a planning toolset configured with its team or an Operator, or low-cost lift testing to start.
- Cassandra: a Bayesian model checked by geo experiments, and the work done within the team's rules, starting free.
- Undecided: ask both about priors and experiments, then run both on the same history for a cycle.
Questions
What is the difference between Cassandra and Recast for marketing mix modeling?
Both pair a Bayesian marketing mix model with geo experiments that feed it. Recast is configured with its team or a certified Operator. Cassandra, an AI marketing brain, uses what it learns to do the work.
Does Recast run geo experiments?
Yes. GeoLift by Recast designs and analyses geo experiments using synthetic controls, and Recast says lift results feed into its MMM.
How is Recast priced?
Recast's documentation lists GeoLift by Recast with a six-month free trial, then $100 a month or $1,000 a year, or access as part of the Recast MMM (as listed on 12 October 2026). For the Recast MMM, its site invites you to book a demo.
Is Recast Bayesian?
Yes. Recast says its product is powered by a proprietary Bayesian MMM, and its documentation says the Recast team sets priors with each client during onboarding.
What are alternatives to Recast?
Other vendors that pair a marketing mix model with geo experiments include Sellforte and Measured. Cassandra, an AI marketing brain, is another. Our list of marketing mix modeling software sets out who each suits.