Marketing measurement comparisons

Every measurement decision arrives as a choice between two things that both sound reasonable. Model or test. Build or buy. Geo split or audience holdout. The arguments for each are easy to find and they rarely help, because most of what gets written explains what the two options are and stops before the part you need, which is why they disagree and what to do when they do.

These pages take the other half. Each one names a pair, explains the mechanism that makes them return different answers from the same data, states plainly what neither of them can tell you, and ends on a rule you can apply without reading it twice.

Where an option other than ours is the better call, the page says so and names who it is better for. A comparison that never reaches that sentence is an advertisement with a table in it.

The decision

Which number to act on

Two instruments disagree about the same channel in the same week. These decide which one sets the budget and which one stays operational.

MMM vs multi-touch attribution

What each one measures, when each is the better choice, and how far apart they land on the same data.

Attribution vs contribution

Attribution credits tracked touchpoints. Contribution measures each channel's share of total revenue.

MMM vs incrementality testing

What each one measures, when each is the better choice, and how a test calibrates a model.

First touch vs last touch vs multi touch

How each model assigns credit, what each is best for, and the drawback of each.

Designing the test

An experiment is the only thing that separates a channel that caused sales from one that was present while they happened. These cover what to run and what a result is worth.

Geo experiments vs holdout tests

What each design splits, where a geo holdout fits, and which one suits which channels.

Incremental lift testing vs A/B testing

An A/B test compares two versions. An incrementality test compares advertising against none.

Reading the model

What a model can express, what it hides, and what has to be visible before anyone outside your team will accept the number.

Bayesian vs frequentist MMM

How each treats uncertainty and prior knowledge, what each outputs, and when each is enough.

Hierarchical vs non-hierarchical MMM

What a hierarchical model is, how partial pooling works, and when separate models are enough.

Glass box vs black box MMM

What makes a model transparent, six things to check for, and four questions to ask in a demo.

Owning the capability

Whether the measurement belongs inside your team, inside an agency, or inside a platform, and what each answer costs after the first model exists.

Build or buy marketing mix modeling

What each path involves, where the costs sit over time, and when each is the better choice.

Robyn vs Meridian vs PyMC-Marketing

How the three libraries differ in method, language and uncertainty, and which suits which team.

MMM agency vs software

What an MMM agency provides, what software provides, and when each is the better choice.

Free marketing mix modeling software

The open-source options, what each needs to run, and how rare vendor trials are.

Building a shortlist

No name in hand yet. Three lists that say who belongs on one, each with its criterion written out so you can disagree with it.

Measured alternatives

Eight alternatives to Measured, chosen by a stated criterion, with who each one suits.

Best marketing mix modeling software

Five platforms and the free libraries, what each publishes, and who each one suits.

Best incrementality testing tools

Five platforms that publish geo and holdout claims, the free options, and who each suits.

Choosing between platforms

You already have a name in hand and a shortlist to cut. These say what each vendor publishes, where they are the better answer, and what neither of them can tell you.

Cassandra vs Measured

Eight of ten published method claims, and no price on any page we reached.

Cassandra vs Sellforte

Both Bayesian. They publish daily ad-set reads, we read at campaign level. Both publish a price.

Cassandra vs Recast

Two claim sets that differ on one line. What separates them is who sets the priors.

Cassandra vs Lifesight

Unified measurement against two instruments that check each other.

Cassandra vs SegmentStream

AI answers put these together. They solve different problems.

Cassandra vs Haus

The same experimental claims. What differs is where a result lands.