Primary Data Collection Gate (Delivery)

Primary Data Collection Gate (Delivery)

1. 01

   Define the decision the study must inform before writing a single question

   Write one sentence naming the client decision at stake, such as which landing page variant to ship or which lead segment to prioritize. If no decision depends on the answer, cancel the study and reallocate the budget.

2. 02

   Split the research budget between self-reported and observed data

   Allocate roughly 60 percent to behavioral signals (session recordings, funnel drop-off, heatmaps) and 40 percent to surveys or interviews. Self-reported answers tell you what people claim; observed behavior tells you what they did.

3. 03

   Draft questions against a fixed response schema and pilot with five internal respondents

   Use a conversational form builder like Typeform for adaptive branching, or a structured survey engine for longitudinal tracking. Run the pilot to catch ambiguous wording, double-barreled questions, and answer options that do not cover the real distribution.

4. 04

   Set the minimum sample and fielding window before launch

   For B2B research, 30 to 50 qualified responses per segment is a working floor; consumer studies need 200 or more. Publish the fielding window to the client in writing so nobody extends it mid-flight to chase a preferred result.

5. 05

   Instrument the collection layer for source and completion tracking

   Tag every response with acquisition channel, device, and completion timestamp. Without source tags you cannot separate a genuine segment difference from a channel artifact.

6. 06

   Run a mid-fielding quality check at the 40 percent mark

   Scan for straight-lining, sub-30-second completions, and duplicate IPs. Remove flagged responses before analysis rather than after, and log every removal with a reason.

7. 07

   Pair every headline finding with at least one behavioral corroboration

   If survey respondents say pricing is the top blocker but checkout recordings show form abandonment at the address field, the behavioral signal wins. Report both and state which one you weighted.

8. 08

   Archive raw responses and the analysis script in the client workspace

   Store the unedited dataset alongside the cleaning logic so a future analyst can reproduce the numbers. Retainer clients frequently ask for a re-cut six months later.