Primary Data Collection Gate (Onboarding)
A checklist with 7 steps: Write the decision statement before selecting any collection method.
By InnovaAI ResearchPublished
What are the steps?
Primary Data Collection Gate (Onboarding)
- 01
Write the decision statement before selecting any collection method
Name the exact client decision this research will inform (pricing test, onboarding flow, positioning claim) and the threshold that would change the recommendation. If no decision hinges on the answer, cancel the study.
- 02
Split the question set into behavioral and self-reported halves
Self-reported items go to a conversational form builder such as Typeform; behavioral items require session recording, funnel drop-off, or event data. A study that only asks people what they did is a hypothesis, not evidence.
- 03
Cap the instrument at 12 questions and 6 minutes for external respondents
Completion rates fall sharply past that point, and partial responses are worse than fewer complete ones. Move demographic and qualifying items to the end so abandonment still leaves usable answers.
- 04
Define the sample frame and the minimum usable n before launch
State who qualifies, how they will be reached, and the count below which the agency will report directional findings only. For B2B panels, budget 3 to 6 weeks of field time and expect a 5 to 15 percent response rate on cold outreach.
- 05
Run a 5-person pilot and time it end to end
Recruit pilot respondents who match the target profile, not agency staff. Fix ambiguous wording, double-barreled questions, and any item where two pilot respondents interpreted the prompt differently.
- 06
Log the collection plan in the client-facing project file
Record method, sample frame, field dates, incentive, and known limitations in one page. This becomes the methods appendix that keeps the eventual readout defensible when a client challenges a number.
- 07
Schedule the observational data pull to overlap the field window
Behavioral analytics and survey responses collected in the same two-week window can be cross-referenced; data pulled a quarter apart cannot. Align the dates before either collection starts.