How it works
Same survey, two matched groups. The difference is the lift.
Agents do the work, researchers sign it off, and the Evidence File proves it.
01
Verified audience
Screened for the target and checked for fraud.
02
Exposed and control
People who saw the campaign, and matched people who didn't.
03
Same short survey
Awareness, recall, favourability, consideration, intent.
04
Lift
Exposed minus control, in points, with a 95% range.
Study designs
Three designs, picked by the question.
| Design | Exposed group | Control group | Best for |
|---|---|---|---|
| Creative Test | Target audience, randomly shown the ad inside the survey | Same pool, randomly shown a different ad or none | Pre-launch creative decisions |
| Campaign Uplift | People who saw the campaign in market, confirmed by recognition, a decoy ad and media data | Matched people from the same window who didn't see it, or a pre-campaign wave | Proving a campaign worked |
| Brand Tracker | Monthly waves of the target audience | Pre-campaign baseline and competitor brands | Brand health between campaigns |
Sample size
Smallest lift we can reliably detect.
95% confidence, 80% power, 30% baseline response.
- 150 per cell14.8 pts
- 200 per cell12.8 pts
- 300 per cell10.5 pts
- 400 per cell9.1 pts
300 per cell is our standard. Most real campaign effects are 3 to 10 points, which is why 200 per cell so often ends in “not significant”.
Respondent journey
One journey, whichever channel someone arrives from.
Every step writes to the Evidence File: who was invited, through which channel, who was excluded and why.
01
Invite
Meta, QR, publisher, panel, WhatsApp or email
02
Screen
Qualifies people and sets the quota cell
03
Consent
Purpose, privacy notice, opt-in to re-contact
04
Survey
Short and mobile-first, web or WhatsApp Flow
05
Check
Fraud, attention, exposure and decoy checks
06
Reward
Cash or neutral voucher, then the panel
The agent roster
Ten agents carry a study from brief to Evidence File.
Agents draft, recruit, check and write. Deterministic code does the statistics. Real people give every answer.
- 01IntakeTurns a booking into a study spec
- 02IntelligenceBuilds the context brief
- 03QuestionnaireBuilds surveys from templates
- 04ChannelsRuns recruitment on every channel
- 05VerificationScreens fraud, bots, false claims
- 06FieldworkRuns invites, quotas and rewards
- 07AnalysisRuns the locked stats plan in code
- 08ReportingWrites up from computed tables
- 09EvidenceAssembles the Evidence File
- 10NormsBuilds benchmarks from every study
Run as a managed service by our delivery partner, Agentic MSP: change control, monitoring, a researcher desk on every gate, incident response, UK GDPR controls and agreed service levels.
Human gates
Agents move a study forward. People sign the gates.
- Gate 1Researcher approves spec and survey
- Gate 2Analysis plan locked and hashed
- Gate 3Researcher signs the report
Draft → Approved → Fielding → Closed → Analysed → Signed off → Released. Every transition is logged with who, what and when. No study reaches Released without all three gates signed.
Trust and compliance
Six rules keep it independent and lawful.
Real people only
Reported results come from verified humans. Synthetic answers only pre-test wording, and are labelled.
Research, not marketing
No sales message in any invite or survey. The ICO treats marketing dressed as research as direct marketing.
Opt-in first
WhatsApp needs opt-in before any business message. Privacy notices name every channel we use.
Clean incentives
Cash or neutral vouchers only. MRS rules bar a client's own goods or services as incentives.
Locked analysis
Method, KPI and sub-groups fixed before fieldwork. Nobody edits results, and flat results are reported.
UK GDPR by design
Consent, data minimisation, retention limits and processor agreements across every channel.