Your study runs as a lightweight 300×250 survey creative served through your DSP, in parallel with your campaign. It asks up to five short questions — recall, understanding, consideration — with no pixel and no cross-site tracking.
The same survey is trafficked as two line items: one to your campaign's exposed audience, one to an identical audience with exposed users suppressed. Because the only difference between the two groups is whether they saw your campaign, the gap between them is the campaign's effect.
For each metric we compare the percentage giving the target answer (e.g. recalling your brand, or top-2-box consideration) in each group. The difference is the absolute lift in percentage points; we also report relative lift.
Each lift is tested with a two-proportion z-test at the 95% confidence level. A metric is only flagged "significant" when the observed lift is unlikely to be due to chance (p < 0.05). We show n, p-value and confidence intervals so you can judge for yourself.
Responses are anonymous. We never collect names or full IP addresses — IPs are masked at collection (host bits dropped). Data is written server-side and isolated per account. See our Privacy Policy.
We balance question rotation to hit a target sample per question per cohort, and report a live fill %. Because more people answer the first question than the last, later metrics collect somewhat fewer responses — the dashboard shows this so you never over-read a thin sample.