Use when you know the conversion question but not which research method would actually answer it.
conversion-research-method-picker.md
You are choosing the cheapest research method that would genuinely answer a conversion question. You are not running the research or designing a test.
THE QUESTION I NEED ANSWERED: {{RESEARCH_QUESTION}}
WHAT I ALREADY KNOW AND WHERE IT CAME FROM: {{EXISTING_EVIDENCE}}
TRAFFIC ON THE PAGE OR STEP: {{TRAFFIC_VOLUME}}
TIME AND BUDGET AVAILABLE: {{CONSTRAINTS}}
TOOLS AND DATA I CAN ALREADY ACCESS: {{AVAILABLE_TOOLS}}
Output a ranked table: Method | What it tells us about {{RESEARCH_QUESTION}} | What it cannot tell us | Sample or time needed | Feasible within {{CONSTRAINTS}}? | Cost (Low/Med/High).
Consider at least: an analytics query, session recordings, an on-site survey, customer interviews, a review of support and sales tickets, a usability test with five people, a tree test, a painted door, and an A/B test. Rank by cost to answer, cheapest first, not by rigour.
Then state:
- The single method to start with, and the specific first thing to look at.
- Whether {{EXISTING_EVIDENCE}} already answers part of the question, so that part needs no new research.
- Whether {{TRAFFIC_VOLUME}} makes an A/B test viable here. If it does not, say so directly and stop recommending one.
- Which decision changes depending on the answer. If no decision changes, say the research is not worth doing.
Constraints: do not recommend a method needing a tool absent from {{AVAILABLE_TOOLS}}. No em dashes.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When do I pick a method rather than run the one already proposed?
When you have a conversion question and the discussion has already jumped to a method, normally a test or a survey. It ranks by cost to answer rather than by rigour, and it will say directly when your traffic cannot support an A/B test and then stop recommending one, which ends a familiar argument.
What do I need in front of me before running it?
One clear question, what you already know and where it came from, traffic on the page or step, your time and budget, and the tools you can actually access. Write the question as a single sentence first, since a vague one makes every method look reasonable. It will not recommend a tool you do not have.
What comes back, and which part is worth acting on?
A ranked table of roughly nine methods with what each cannot tell you, sample or time needed and feasibility, then the method to start with and the specific first thing to look at. The final check is the one that saves money: which decision changes with the answer, and if none does, skip the research.
What is the mistake that costs me here?
Skipping the check on what you already know. A good part of most research questions is already answered by existing evidence, and paying for a study that confirms it is a common way to lose a month. Cheap and directional beats rigorous and late when the finding only has to point you at a page.