Causality Engine vs. Measured: Incrementality Testing Compared

Causality Engine vs. Measured: Incrementality Testing Compared

Buy Measured when one large decision rides on one channel: its geo-based holdout test is the stronger evidence. Use a Causality Engine read, causal inference on the GA4 history you already have, to see what every channel caused next to what last-click gave it, and to pick which one deserves the experiment.

Attribution numbers that drift away from revenue do so for reasons that start outside your store. Last-click can't tell you whether a sale would have happened without the ad, privacy limits on Meta and Google data have made older marketing attribution models less reliable, and each ad platform still reports only on its own ads. Measured and Causality Engine both go after incrementality, what a channel actually added, from opposite directions: Measured runs new experiments and calibrates a mix model on them, Causality Engine reads the history you already have.

Measured's core offering is incrementality testing, built to answer one question: would this conversion have happened anyway without the ad? It answers it with geo-based holdout experiments. Take two similar regions, run the campaign in one and withhold it in the other, then compare sales. A well-designed geo test is strong evidence of causal lift, because randomisation removes confounders rather than adjusting for them. Its strength is experimental evidence of where budget adds sales, and it's built for enterprise brands running formal incrementality programs. Alongside the geo tests, Measured sells audience split experiments and a Causal MMM calibrated with incrementality tests.

Measured is good at proving a channel's lift, but a geo test answers one question at a time. A geo test tells you that the channel you tested was incremental in the window you tested, and by how much. It doesn't tell you what every other channel did over the same period, or what last quarter's spend caused, because that budget is already spent. Each extra answer means another test, more weeks and more spend withheld. The limit of experiment-only measurement is coverage and speed, not rigour: you get a precise answer for one channel and one window, and that's hard to reallocate a whole mix on.

You can run a rough version of Measured's method yourself, for free, before you buy either tool. Pick the channel that worries you, pause it in one region for two weeks, and compare how that region's revenue changed with how revenue changed everywhere else, against the two weeks before. Watch MER (total revenue over total ad spend) while you do it. It isn't a powered experiment, but if revenue barely moves, the channel was taking credit rather than causing sales, and you know which channel deserves a real test.

Causality Engine works from the other direction: causal inference on observational data. Instead of running a new experiment, it reads the GA4 history you already have. You upload a GA4 export (the €99 read covers whatever GA4 history you export); there's no pixel and no SDK (how a read works). For every channel in the export, paid and organic alike, it estimates what revenue would have been without that channel, and it gives every channel a data-health score from 0 to 100 that grades the read, not the channel. Where a channel is under 1% of the credit on both the engine and last-click, the read calls it too small to act on rather than forcing a verdict. The output is what each channel caused, next to what last-click gave it, with a next step for every channel, in 1 to 2 minutes. That covers the whole customer journey at once, not one tested channel at a time.

The real difference is the kind of evidence and how much it covers. Measured gives experimental evidence for the channel and window you test, which is what you want when a large decision rides on one channel. Causality Engine gives observational estimates for every channel at once, each with a data-health score, which helps you reallocate the whole mix between experiments. If you run paid social, search, email and organic at the same time, seeing what every channel caused side by side is a practical first view, and a geo test is the stronger confirmation for the one channel that matters most. Move budget in steps away from the channels that look non-incremental, and test the one you are least sure of. Email is a good example: an abandoned cart email flow can take last-click credit for orders that were already on their way, and a causal read shows what email as a channel caused, next to what last-click gave it. That's a decision about which spend to keep, not just a reporting choice.

For illustration (a hypothetical brand, no real figures): a DTC fashion brand runs campaigns on Meta, Google and TikTok, plus email and organic social built on user-generated content. A Causality Engine read on its GA4 data shows what each channel caused, next to what last-click gave it. Suppose the two agree on Paid Search, within 25%, while the engine gives Paid Social, where the Meta and TikTok ads sit, far less credit than last-click does. Now the brand knows where the disagreement is. Instead of testing every channel, it spends its experiment budget on a geo test of Paid Social alone, the channel the two methods disagree on. That tests the right thing and costs less than testing everything.

Even a well-designed test has practical costs. Clean control regions are hard to build when buyers see ads everywhere, and every test means withholding spend in the holdout regions and waiting weeks for enough data. Causality Engine works on data you already have: a read returns in 1 to 2 minutes from a GA4 export, and Pro (€299/month) runs continuously with automated GA4 ingestion, so estimates update as new data arrives. The trade-off is evidence strength: observational inference is bounded by its assumptions, which is why every channel carries a data-health score that grades the read. Speed matters when you're reacting to a trend or to customer feedback this week.

Causality Engine vs. Measured: A Methodological Comparison

Measured is built for proving a channel's impact with experiments, and for media optimization around the result. Causality Engine reads every channel at once. That helps you reallocate budget and choose which channel deserves an experiment. Used in sequence, the two cost less than testing every channel.

Most other tools in the category are multi-touch attribution (MTA) or marketing mix modeling (MMM) platforms. Triple Whale's attribution is pixel-based MTA, with MMM and incrementality testing in its Compass suite. Northbeam combines MTA with MMM and, since April 2026, native incrementality. Rockerbox adds control-group incrementality tests to MTA and MMM. Hyros tracks clicks server-side, and Cometly and WeTracked pair a first-party pixel with server-side tracking. Multi-touch credit shows what came before a conversion, not what caused it. Measured and Causality Engine both go after the causal question: Measured with experiments and a mix model calibrated on them, Causality Engine with inference on the GA4 history you already have.

Benchmark Data: What We Do and Don't Publish

The beta users' results, with the figures each one agreed to publish, are on the case studies page.

What causal inference changes is the question behind the number: from which channel was present when a sale happened to which channel caused it. That's what makes a reallocation defensible in a budget meeting, and it's why the read sets last-click beside every number, gives every channel a data-health score and checks every number against your export, instead of handing you one confident-looking number.

A geo test gives you that estimate for one channel; a causal read gives it for all of them, graded by a data-health score. Whichever you use, ask how the numbers were checked: for a test, its power and design; for a model, its validation against randomised experiments.

Conclusion

Measured sells the stronger evidence: a well-powered geo test for the channel and window it covers, worth buying when a large decision rides on one channel and there's budget to run it. Causality Engine sells coverage and speed: what every channel caused, next to last-click, from a GA4 export, with a data-health score on each, for €99.

FAQ

Can Causality Engine replace my existing attribution tool?

For the budget question, often yes; for day-to-day reporting, no. The read shows what each channel caused, next to last-click, from a GA4 export, which is the question attribution tools try to answer. It works on channel groups, not campaigns or creatives, and it isn't a daily dashboard, so keep your current tool for reporting if you use it that way.

How sure is a Causality Engine estimate?

Every channel gets a data-health score from 0 to 100 that grades the read, not the channel: under 90, move budget in steps. Last-click, rebuilt from your own export with Direct left out, sits next to every number; within 25% the two agree, and a channel under 1% on both is too small to act on. Receipts check every number against your export. Ask any vendor, this one included, how its method was checked against randomised experiments.

Is Causality Engine suitable for small to medium-sized DTC brands?

Yes. There's no minimum spend, revenue or team size. The €99 read runs on whatever GA4 history you export, and Pro is €299/month, billed monthly, cancel any time. The read needs enough GA4 history to work with, and it grades itself: every channel gets a data-health score from 0 to 100, and a channel under 1% on both methods is called too small to act on instead of given a confident number.

How long does it take to see results with Causality Engine?

A €99 read returns results 1 to 2 minutes after you upload a GA4 export, and Pro runs continuously with automated GA4 ingestion. Unlike incrementality tests that take weeks per test, it works on history you already have, so nothing has to be paused to get a first answer.

What kind of data does Causality Engine integrate with?

Today, Causality Engine reads a GA4 export; there's no pixel and no SDK. Every channel in the GA4 export is modeled, paid and organic alike. Shopify, Meta, Google, TikTok and other platforms are not connected directly.

Related Resources

Agency vs In House Attribution Numbers: Who Is Right

Last Click vs. Data-Driven Attribution: Which Should You Use?

Attribution Model Comparison Template: Side by Side Analysis

Causal Inference Vs Rule Based Attribution