Experiments designed, run and read by independent measurement experts. Find the true lift of a channel or campaign, then use it to sharpen every budget decision that follows.
Run your marketing for one group and not for a matched group, then compare. The difference is the sales your marketing caused: an answer to "if we switched this off tomorrow, what would we lose?"
A controlled experiment, not a correlation
We don't sell media, so we don't mark our own homework
TV, out-of-home and radio as well as search and social
Results calibrate the model, so every update gets sharper
We recommend the design that fits your channel, budget and markets
Turn media up, down or off in selected regions and compare them with matched control regions. Works for almost any channel, online or offline.
Hold back an audience or customer group from a campaign, such as CRM, retargeting or brand search, to see what happens without it.
Conversion and brand lift studies run inside ad platforms, set up with clear success measures and read independently against your own data.
When a launch has to go national, we build a counterfactual from comparable markets or series to estimate what would have happened anyway.
From the question to a budget decision in five steps
We start from your biggest decisions and, where you have MMM, the channels with the widest ROI ranges, where a test is worth the most.
Method, markets, duration and spend, sized with a power analysis so the test can detect the effect you care about.
We work with your media team or agency on set-up and check in flight for leakage, delivery problems and outside shocks.
Incremental sales, ROI or CPA, with the uncertainty around them, presented to the people who need to act.
Agree the budget change, and feed the result into your Bayesian MMM so future estimates reflect what the test found.
Illustrative geo test: test and control regions track together until the campaign goes live.
Testing is as much about process as statistics. We handle the parts that usually go wrong: choosing comparable markets, sizing the test and keeping it clean while it runs.
Tests and models work best together
A test measures one channel at one moment. MMM measures everything, all the time. We use each to check the other.
No media to sell and no platform to protect, so the readout is about your business, not anyone's targets.
Tests sized for your budget and markets, so you don't go dark nationally to learn something a few regions could tell you.
Results explained in plain language, with ranges, so marketing and finance can make the call together.
How an independent test compares with other ways of judging a channel
| Platform-run lift study | Before-and-after comparison | Last-click attribution | ||
|---|---|---|---|---|
| Measures causal impact | ||||
| Independent of the media seller | ? | |||
| Measures your salesall channels, including offline and marketplaces | ||||
| Works for TV, out-of-home and radio | ? | |||
| Set up the test with accuracy | ? | |||
| Reports uncertainty | ||||
| Calibrates your MMM | ? |
yes · ? depends on set-up · no. Already running a platform lift study? We can read it against your own data.
Most tests run for several weeks, plus time afterwards to capture delayed effects. The power analysis at the design stage sets the duration, so you know before launch how long it needs.
No. A test answers a focused question on its own. With MMM in place we can pick the tests that matter most and carry each result into every future model update. See Always-on MMM and In-house MMM.
Not necessarily. Many designs scale spend up in test regions rather than turning it off, and we size the test so you only hold back what's needed to get a clear answer.
Yes. We can review the design and data from past platform lift studies or geo tests, tell you how much weight the result can bear and, where it holds up, use it in your MMM.
More on measuring incrementality
Have any project on mind?
For immediate support: