Cassandra vs LiftLab: the same loop between marketing mix modeling and testing, built on different inference
LiftLab publishes a loop between agile marketing mix modeling and incrementality testing, aimed at growth leaders, boards and CFOs. Cassandra, the marketing mix modeling platform, publishes the same loop with a Bayesian production engine and priors co-set with the client. Neither publishes a multi-touch attribution claim.
Both products are built on the same idea: a model of the whole mix and a set of experiments that calibrate it, run as a repeating cycle. The differences are the inference framework, the audience each names, and how each is bought. This page sets out what each says it is, where the published claims differ, why the inference framework matters inside a loop, what each publishes about pricing, the questions worth asking both, and when LiftLab is the better choice. Everything about LiftLab comes from its own published pages.
What Cassandra is
Cassandra is a marketing mix modeling platform. Its production engine is a Bayesian model built on PyMC, so every channel effect comes back as a distribution with its uncertainty stated. Geo experiments are designed and run in the same system, using the same method family as GeoLift, and each result enters the model as a constraint rather than as a separate report.
Reads sit at campaign level, in the planning layer, and nothing changes in an ad account unless a person confirms the exact figure. The platform does not do multi-touch attribution.
It is not built on Google Meridian. Meridian is one of three model engines a project can choose, it is being retired, and it is not offered on new projects.
How this comparison was made
Everything below about LiftLab comes from what LiftLab publishes. We read their homepage, product pages and the pages their navigation links to on 14 September 2026, and quoted what we found.
We have not tested their product. Where we describe an absence, we mean we could not find it on the pages we reached.
If something here is out of date, tell us and we will correct it.
When LiftLab is the better choice
Your audience is a board more than a marketing team. Their framing names a board and a CFO explicitly. A product built to produce a narrative those readers accept is solving a communication problem as much as a measurement one.
You want the loop and the inference framework matters little to you. A team that does not plan to inspect priors can reasonably choose on other grounds.
You want a published methodology. They publish one. Twelve of the twenty-one vendors we checked do.
You want a faster modeling cycle. "Agile" is their word, and if your planning cadence is faster than quarterly, that positioning is aimed at you.
What each one says it is
LiftLab's homepage heading is "Agile Marketing Mix Modeling for Growth Leaders", and its description says it closes the loop between agile MMM and incrementality testing, turning marketing spend into a growth story a board can see and a CFO can defend. Captured 14 September 2026.
Cassandra, the marketing mix modeling platform, runs a Bayesian model of the whole mix with geo experiments in the same system, and each result constrains the model. Reads sit at campaign level, in the planning layer.
Both describe the same architecture: model, test, recalibrate, repeat. That overlap is useful to a buyer, because it narrows the decision to a few specific questions.
It also means the practical differences sit in how the loop runs: how often the model is refitted, how test results are taken in, who sets the assumptions, and how results are presented to the people who approve the budget.
Where the published claims differ
Across the pages we reached, LiftLab publishes claims covering marketing mix modeling, incrementality, geo experiments, holdout testing, causal inference, calibration and an open methodology.
Bayesian inference. We publish a Bayesian claim for our production engine. LiftLab does not publish one on the pages we reached. Six of the twenty-one vendors we checked do.
Attribution. Neither publishes a multi-touch attribution claim. Seventeen of the twenty-one vendors we checked do. On this point LiftLab and we sit together, apart from most of the market.
Methodology. LiftLab publishes an open methodology, as twelve of the twenty-one vendors we checked do.
Audience. LiftLab names growth leaders, boards and CFOs. We name budget decisions that finance or a client will review.
Why the inference framework matters inside a loop
In a loop, the model has to take in each new test result and carry it across the rest of the mix. A Bayesian model can take a measured result in as a prior or as a term in the fit, and because it reports each channel's effect as a distribution, it can show how far the new evidence narrowed the range for the tested channel and for the channels correlated with it.
Other frameworks can also use test results, for example as calibration constraints during model selection. The practical questions are how each product does it, and whether the effect of a new test on the other channels is visible to the people using the model.
Not publishing a Bayesian claim is different from not using Bayesian methods, and we cannot establish from outside how LiftLab's model is built. It is a question to ask them directly.
An example: a test measures a lower return for paid social than the model assumed. In a model that shows the full distribution, the paid social range narrows and shifts, and the ranges for channels whose spend moved with paid social shift too. Seeing that knock-on effect is what lets a planner explain why more than one budget line moved after one test.
What each publishes about pricing
We found no pricing page and no price on the LiftLab pages we reached, and no mention of a trial or free tier.
Of the twenty-one vendors we checked, fifteen publish a pricing page and seven publish a price. LiftLab is in the group that does neither. That means cost can be compared only after a sales conversation, which matters for teams that need a figure for this quarter's plan.
We publish a price, and every route in is a booked demo. A published figure gives a team the scale of commitment before the first call.
What running the loop takes
Whichever product runs it, a loop between model and tests needs the same things from the business.
Clean history. A reconciled weekly record of spend and outcomes, which is most of the setup work.
Protected test windows. Tests of four to six weeks, with no overlapping launches or budget changes in the test regions.
Agreed priors. Co-setting priors with the team is the slower part of starting, because it asks people to write down what they believe about each channel. It is also what makes the model's assumptions reviewable later.
Someone who owns the cycle. Choosing the next test, reading the result into the model and explaining what changed in the plan.
A loop that nobody owns stops after the first test. Naming the owner before the first cycle starts is the step most teams skip.
Questions to ask both vendors
When a test result lands, what changes in the model, and can I see it? The answer shows how the loop actually closes.
How do you choose which channel to test next? Ideally where the model is least certain and the most budget is at stake.
Show me the assumptions on a live model, and change one.
What does "agile" mean in practice? How often the model is refitted, and how quickly a new test result appears in the plan.
Export the output behind this chart. The numbers, in a file you keep.
Which model version produced last quarter's plan, and can you rerun it? A loop that cannot reproduce its own history is hard to audit.
What we do not do
Reads sit at campaign level. The platform does not produce ad-set or audience-group readings, and it is not a day-to-day optimiser. It does not offer attribution.
A channel needs enough history behind it, and enough movement in that spend to read against. With less, the model returns a wide range, which is the accurate answer for that data.
Geo experiments need regional structure, so a single-market business with no way to build a comparison group cannot run one. A loop without experiments is a model on its own, and for that business the model with its uncertainty stated is the available instrument.
Cassandra, the marketing mix modeling platform, runs the same loop with a Bayesian model in it and priors your team helps set, which keeps the step where evidence enters visible to whoever audits it later.
A price is published, and LiftLab's pages do not show one. On a comparison where the architectures match this closely, seeing each run on your own history is a faster route to a decision than reading either vendor's account of the difference.
What moving between platforms involves
The experiment record is the main asset, and it moves with you. Each test's channel, regions, window, spend change and result remains valid evidence in any model.
The cleaned history of spend and outcomes moves too. The main new work is priors: agreeing them with the team, and recording who set each one and why.
Run both on the same period for a cycle, and compare how each takes in the same recent test. With similar architectures, the differences usually trace to a specific assumption, and understanding it is useful whichever platform you keep.
How to choose
A board-facing narrative, a faster modeling cycle, or a published methodology your reviewers already know: LiftLab.
The same loop with a Bayesian production engine, priors co-set with your team and a published price: Cassandra, the marketing mix modeling platform.
Undecided: ask both what changes in the model when a test result lands, and choose the answer you can explain to finance. If both answers are clear, run both on the same history for one cycle and compare the plans they produce.
Questions
What is the difference between Cassandra and LiftLab for marketing mix modeling?
Both link a marketing mix model with incrementality tests in a repeating loop. LiftLab names boards and CFOs and publishes an open methodology. Cassandra, a marketing mix modeling platform, publishes a Bayesian production engine and a price.
Does LiftLab publish pricing?
We found no pricing page and no price on the pages we reached, and no mention of a trial.
Is LiftLab Bayesian?
They do not publish a Bayesian claim on the pages we reached. That describes what they publish, and we cannot establish from outside how their model is built.
What does agile marketing mix modeling mean?
A marketing mix model refitted more often than the traditional quarterly or annual study, so that new data and test results reach the plan sooner.
What are alternatives to LiftLab?
Vendors publishing both geo experiment and holdout claims include Measured, Haus, Recast and Cassandra, a marketing mix modeling platform. Our list of incrementality tools sets out who each suits.