Designs a rigorous primary research protocol for collecting new data - research question, hypotheses, method selection, sampling with a power analysis, instrument design, a named control for every bias threat, and a pre-specified analysis plan. Use when someone asks "design a study to test this", "how many respondents do I need", "should this be a survey or interviews", or must produce defensible first-hand evidence rather than cite existing work. Do NOT use for synthesizing already-published sources - use deep-research or literature-review instead; for writing the discussion guide for a single expert conversation, use expert-interview instead; for detailed survey question wording and scale design, use survey-designer instead.
Click to play with sound.
---
name: Primary Research Planner
description: Designs a rigorous primary research protocol for collecting new data - research question, hypotheses, method selection, sampling with a power analysis, instrument design, a named control for every bias threat, and a pre-specified analysis plan. Use when someone asks "design a study to test this", "how many respondents do I need", "should this be a survey or interviews", or must produce defensible first-hand evidence rather than cite existing work. Do NOT use for synthesizing already-published sources - use deep-research or literature-review instead; for writing the discussion guide for a single expert conversation, use expert-interview instead; for detailed survey question wording and scale design, use survey-designer instead.
---
# Primary Research Planner
Use this when someone needs to collect new data, not review existing literature. The costly failure this skill prevents is the study that produces an answer nobody can trust: underpowered samples, leading instruments, and analysis decisions made after seeing the data - each one quietly converts research spend into confident noise. The goal is a plan that produces valid, defensible answers before a single respondent is contacted.
## Inputs to collect
1. **The decision** the research informs, and what answer would change it. Research with no decision attached gets descoped or declined.
2. **The vague goal as stated** - to be reframed into an answerable question.
3. **Timeline and budget**, since they constrain method and sample size.
4. **Access to the population**: existing panels, customer lists, recruiting budget.
5. **Prior evidence**: what is already known, so the study tests what is genuinely open. If an input is a guess, label it a guess.
## Operating procedure
The steps run in this order because each constrains the next: the question type dictates the method, the method dictates the sample, and the analysis plan must be locked before data arrives or it is not confirmatory.
### Step 1: Clarify the research question
Reframe the vague goal into a specific, answerable question. State whether the aim is **exploratory** (generate hypotheses) or **confirmatory** (test one) - these demand different designs, and pretending an exploratory study was confirmatory is the root of most overclaimed findings.