3rd-party validated lift for creator and retail media spend.
Did the marketing spend actually move demand?
The only way to show your media caused the outcome, rather than merely happened before it, is a counterfactual test: build the version of events where the campaign never ran, and measure the gap. That is what separates what your media did from everything else moving demand at the same time. The outcome itself also has to come from an independent observer rather than from either party to the transaction.
We build that counterfactual two ways, depending on whether your exposure can be geo-targeted.
The Counterfactual Edge: Two ways to build the version of events where the campaign never ran.
The approach we use comes down to a single question: can your exposure be geo-targeted?
Not geo-targetable: creator content, and any campaign that reaches an audience by interest rather than by market. No campaign viewers missed it based on location, so there is no geo-based comparison group waiting to be used. We build the comparison from historical data instead, and measure branded search demand: how many people go looking for the brand by name. It is public, it moves for real reasons, and it is not a sales figure.
Geo-targetable: in-store screens, forecourt media, and onsite retail media that ran in some stores and not others. The comparison group already exists, so we read sales at the exposed stores against matched stores in the same chain that ran no media.
Read A: The public marketing data read
For: Creator content, and any campaign whose audience is distributed by interest rather than by market.
- Outcome
- Branded search demand, from public data
- Comparison
- Comparison trends that move with the brand's demand for reasons unrelated to the campaign
- Estimator
- Bayesian structural time series
- Dose
- Campaign reach
- What we need from you
- The upload or flight date
- Integrity
- Every input is public. There is no input to adjust.
Read B: Matched-store read
For: In-store screens, forecourt media, and onsite retail media that ran in some stores and not others.
- Outcome
- Unit and dollar sales of the advertised item and its category, at store level
- Comparison
- Stores in the same chain that ran no exposure, weighted to reproduce the exposed stores' pre-campaign path
- Estimator
- Difference-in-differences, or a staggered-adoption estimator where the rollout was phased
- Dose
- Impressions, loops, or screen-hours per store
- What we need from you
- The flight dates, the store list, and weekly sales from the retailer
- Validation
- The comparison store list and the specification are fixed before the outcome window closes, and the fee does not move with the finding. You see every result before anyone else does.
We ran one on ourselves and published it.
A direct-to-consumer apparel brand ran paid integrations inside creator content between June 2024 and August 2026. We found 46 of them from public sources alone, read the 19 with a clean pre-campaign baseline, and published the result: a reach threshold below which the estimator finds nothing at all, a placebo row that comes back at minus 3.0% with a p-value of 0.80 against 0.001 for the real dates, and a weekly persistence of 0.52. Nobody asked us to do it and nobody paid for it, which is rather the point.
Our Commitment to Transparent Analytical Methods
The control set or the control store list, the pre-period, the estimator and the outcome window are fixed in writing and dated while the result is still unknown. Choosing the specification after seeing what each one produces is the failure that makes most observational marketing results worth nothing.
Fixed, agreed before the read, not a share of measured lift and not renewable on the strength of a positive one. A number that can only come back positive is not a fair measurement, and no media planner treats it as such.
Every read is run against dates and trends where nothing should be found. If the method finds effects where none can exist, the read is not delivered. On public-data reads the results are published either way. On client reads you see them first.
Public inputs and a published specification are only worth something if somebody can actually run them. Anyone who wants to check a read can have the code and the specification it ran on, including anyone who expects to disagree with the result.
We compute the smallest effect the design could have detected and publish it alongside the number. It is the only way to tell a campaign that did nothing from a campaign too small for the design to see.