Insight Engine Intake (Onboarding)

Insight Engine Intake (Onboarding)

1. 01

   Write the client's decision question in one sentence before opening any research tool

   If the question cannot be stated in a single sentence, the engagement is not ready for data collection and the kickoff needs another pass with the account lead.

2. 02

   Split the question into what people say and what people do, then assign a capture method to each half

   Self-reported capture through Typeform or Qualtrics covers stated preference; behavioral capture through Hotjar or session replay covers observed action. Both halves get a named owner.

3. 03

   Set the sample floor and the disqualification rules in writing before the first response arrives

   For B2B panels, a 40 to 80 qualified response floor is a workable starting point for directional reads; document who is excluded (competitors, employees, existing customers) so the client cannot relitigate the sample later.

4. 04

   Map every question to the deliverable it will inform

   Questions that map to nothing get cut. This keeps a 25-question survey from becoming a 60-question survey that respondents abandon halfway through.

5. 05

   Confirm data handling and consent language with the client's legal contact

   Server-side memory and off-device processing now appear in mainstream AI infrastructure, so agencies should disclose where response data is stored and which subprocessors touch it before launch.

6. 06

   Load the question set into the chosen platform and run a five-person internal pilot

   Pilot respondents flag ambiguous wording, broken logic jumps, and mobile rendering problems that cost far more to fix after launch than before.

7. 07

   Publish the analysis plan, including the statistical test or coding scheme, on day one

   An analysis plan written after data collection invites cherry-picking. Writing it first makes the eventual recommendation defensible when a client challenges it.