Causality Engine vs Measured: When to Run the Experiment

The short answer: the budget call is next week, and the ad platforms between them claim more revenue than the store booked last month. Start with Causality Engine. The first finding from your GA4 export costs nothing, and a read of every channel is €99 once, refundable within 30 days (pricing, checked on 27 October 2026). Choose Measured when an enterprise media portfolio needs a standing programme of holdout experiments, with an analyst assigned and budget to withhold media while each test runs. The cheap read is a good way to decide which experiment deserves the expensive weeks.

Disclosure: I founded Causality Engine and own it, and this page recommends it. Every claim about it links to Causality Engine's own pages.

Before you book a demo:

1. Put the calendar on paper. Measured's own holdout guide calls 30 days a good standard length, and longer where buyers take longer to decide. For example, four channels tested one after another at that length is 4 × 30 = 120 days. That is a third of the year before the fourth answer arrives.
2. Use the history you already hold. GA4 has logged converting paths since it went on the store. The read takes twelve months or more of them as one CSV, installs nothing and returns in 1 to 2 minutes, per the how-it-works page.
3. Choose the number you will check afterwards. Find the channel where the platform's claim and the caused figure sit furthest apart. That gap is what to test, and the revenue line to watch once budget moves.

Both use the word causal. Measured means an experiment: take a channel away from some people or regions and count what was lost. Causality Engine means an estimate from paths GA4 already logged. One is slow and strong. The other is fast and graded.

The two, attribute by attribute

Where they part ways

One channel per test, or the whole mix at once

A Measured holdout answers one question well: how many conversions disappear if this channel stops. Its guide splits by customer list for email, catalog and SMS, and by geography for prospecting, CTV and paid search. Causality Engine's incrementality page concedes the point. It calls a geo holdout rigorous, then names the cost: slow, expensive, and fragile when regions are not comparable. The read covers every channel in one pass, with weaker evidence per channel. That is why each one carries a data-health score.

What it costs to find out

A holdout works by not advertising to people who would otherwise see the ad. Which of your working channels would you go dark on for a month? Measured's guide says to begin with the tactic that takes the most spend, where a result moves the most money. It puts email low on the list, because email is cheap whatever its lift. Sensible. It also means the first test lands on the channel you can least afford to get wrong. The read pauses nothing, and already proposes the cautious version of a test: raise, cut or hold, tried in one region first.

When an experiment is worth its weeks

Run one when the read and last-click disagree sharply on a channel whose data-health score is under 90. That is the line below which the how-it-works page says to move budget in steps. Run one for media GA4 never sees, such as TV or catalog. Run one when a month of withheld media is cheap next to being wrong. Otherwise, act on the read.

At The Two Sisters, last-click argued for cutting Pinterest, and on that report cutting was the reasonable call. The beta causal read found it driving sales last-click missed, and the brand kept it live. The case-studies page reports a 60% revenue lift for the brand in 2024. That is the company's own published result, not an independent finding or a forecast for your store.

Prices side by side

Causality Engine, checked on 27 October 2026 on its pricing page: €99 for one read, paid once. Pro is €299 a month, billed monthly, cancel at month end. Enterprise is a custom 90-day pilot, and VAT is excluded. Measured, checked the same day: no price on measured.com. Every product page leads to a demo request.

Measured pairs every customer with an analytics expert and quotes only through a demo. So the default on offer is a standing testing programme, sized in a sales conversation. That fits a portfolio with a test on the calendar most months. It is more than a store with one budget call a quarter will use.

Free first, then paid

Three free steps show whether you need either tool. Add up what each ad platform reports for last month and set the total beside the store's revenue. Causality Engine's home page uses that check as one of five reasons to run a read. Take Measured's free readiness assessment, three minutes by its home page. Then drop the GA4 export on the start page for the free first finding, worked out in your browser. If the gap is real, the €99 read prices every channel. With eight or more channels and weekly budget moves, the pricing page points to Pro, which adds automated GA4 ingestion and a chatbot over your own attribution data. Check any figure the chatbot gives against the read's receipts before it reaches a budget sheet.

Choose Causality Engine if

• Your job is this month's split across Meta, Google, TikTok, email and organic, from history you already have.
• You cannot pause a working channel for a month to learn.
• You want every channel judged at once, including the sales sitting under Direct.
• You want a short list of which channel to put through an experiment first.

Choose Measured if

• Your job is a formal experiment programme across a large media portfolio, with an analyst on the account.
• Much of the spend sits where GA4 cannot see it: TV, CTV, catalog or retail.
• You want a mix model calibrated on your own tests and refreshed weekly, monthly or quarterly, per its MMM page.
• Procurement asks for SOC 2 Type II and ISO/IEC 27001:2022, which Measured's site states it holds.

Often it is both, in sequence: the read ranks the channels, the holdout settles the biggest one. Causality Engine's own comparison with Measured, the company's view, argues the same. For holdouts inside wider validation, see enterprise attribution and how to validate it.

Read the history, then pick the test

Causality Engine, which I founded, gives the first finding free on the start page. The full read is €99 per upload, with the 30-day refund on the pricing page. Read which channel's two numbers disagree most, and take that one to an experiment. Background: the Causality Engine glossary entry.

FAQ

Is a causal read as strong as a holdout test?

No, and most weekly decisions do not need it to be. A well-powered holdout is direct evidence for one channel in one window. The read estimates every channel from past data and grades each. Under a grade of 90, the how-it-works page says to move in steps, and a holdout is the natural next step.

How long does a Measured test take?

Measured's holdout guide treats 30 days as a good standard, and at least the length of your average consideration period. Tests launch in days, its incrementality page says. Results take the full run.

Does Causality Engine replace Measured?

It is not built to. Its home page calls it a layer, not a migration, and it measures nothing GA4 does not record. A brand on Measured can use the read to pick the next channel to test.