%202.png)
Research Method Selection
Matches the method to the question, explains why the alternatives were rejected, and states plainly what the chosen design cannot answer.
Free. Works on Claude, ChatGPT, Gemini or any assistant that accepts a skill file.
What this skill does
The method, encoded.
Method is the most consequential decision in any research project, and the one most often made for the wrong reasons: habit, budget shape, stakeholder preference, or whatever is quickest to buy. The result is rarely a study that visibly fails. It is a well-run study that answers a question nobody needed answered, with the findings then stretched to cover the question that mattered.
This skill starts from the shape of the question rather than a menu of methods. It splits a brief into its component question types (how many, why, how, what would happen if, which of these, is it changing), works out what claim each one needs to support, and only then selects from the full repertoire: surveys, depth interviews, focus groups, AI-moderated interviews, diary studies, ethnography, observation, social listening, desk research, experiments, concept and usability testing, conjoint, MaxDiff, tracking and mixed-method designs.
It also makes the recommendations a supplier rarely volunteers: that the answer is already in your own data, that the question is unanswerable as posed, that a small qualitative study beats an under-powered quantitative one, and that sometimes no research should be commissioned at all.
Best used for
- Testing whether the method named in a brief actually fits the question
- Splitting a multi-objective brief into the methods each objective needs
- Deciding whether any primary research is required at all
- Choosing the least-bad design under a fixed deadline or budget
- Reviewing a proposed design before commissioning
- Justifying a method choice to a stakeholder who prefers another
Typical inputs
What you give it.
The business decision the research serves, with its owner and deadline, Research questions or objectives from 01.02, Population of interest and known route to reach it, Hard constraints: deadline, budget order of magnitude, analysis capability, Hypotheses from 01.03 (optional), Inventory of existing internal or published evidence (optional), Tracker history and prior instrument, where continuity applies (optional)
Typical outputs
What you get back.
Method Recommendation Note, Question shape analysis table, with a verdict per objective, Specified design: population, scale, period, channel, stimulus, analysis, Rejected options record with reasons, A list of what this design cannot answer, Constraint and compromise register, priced in conclusions, Effort profile across design, fieldwork and analysis, Confidence statement on the recommendation itself
Method coverage
What the skill works through.
- Why method choice goes wrong
- Start from the shape of the question, not a list of methods
- The seven question shapes
- What each method can and cannot answer
- Measurement methods: surveys, tracking, experiments
- Depth methods: interviews, groups, AI-moderated interviews
- Behavioural methods: observation, diary studies, ethnography
- Trade-off methods: conjoint and MaxDiff
- Evaluative methods: concept testing and usability testing
- Existing evidence: desk research and social listening
- The four honest verdicts before you commission anything
- Choosing a channel and knowing what it costs you
- When the ideal design is unavailable
- What to write down: the method recommendation note
Download
Free skill. One file.
Enter your email once. Every skill you download after that takes a single click.
How to install
Add the skill file and the five kernel protocols to a Claude Project, a ChatGPT Project, a Gemini Gem, or paste them at the top of any assistant conversation. Then give it your real research material, not a description of it.
Download skillQuestions
Common questions.
How do I choose between qualitative and quantitative research?
By the claim you need to make, not the topic. If the sentence you need at the end contains a proportion or a comparison between groups, you need measurement. If it explains why something happens, or reconstructs how a decision was made, you need depth. Briefs usually contain both, which means two strands in sequence, not one method stretched across both.
How do I know if I need primary research at all?
Run four checks. Is it already answered by internal records or published sources? Is the question answerable by any evidence at all? Would a quantitative design actually be powered for the subgroups you intend to report? And would a different result change what the organisation does? A "no" on the last one means no study is worth commissioning.
What is the most common research method mistake?
Using a survey to answer a "why" question, usually with an open-ended box added to cover the gap. Surveys measure well and explain poorly. The second most common is setting the sample size from the budget rather than from the smallest subgroup you have to report on, which quietly makes the intended analysis impossible.
Can AI choose a research method for me?
It can structure the choice, hold the full repertoire in view, and stop you from skipping the uncomfortable verdicts. It should not settle a trade-off under constraint. Deciding what to sacrifice when the ideal design is unaffordable, and accepting what that sacrifice costs the conclusions, is a professional judgement with accountability attached.
How many depth interviews or focus groups do I need?
There is no fixed number, and any answer given without knowing the variation you need to cover is guesswork. Sample in qualitative work follows the range of people and situations the question spans, and adequacy is judged during fieldwork by whether new sessions are still producing new material, not fixed in advance for a proposal.
What is the difference between conjoint and MaxDiff?
MaxDiff ranks a list of items by relative priority. Conjoint estimates how people trade attributes against each other, including price, and can simulate configurations. Both share one critical limitation: anything left out of the attribute or item list does not exist in the results, so both need prior work to build that list.
How do I explain to a stakeholder why their preferred method is wrong?
Do not argue about the method. Write the sentence they want to be able to say at the end, then show what evidence would be needed to say it honestly and whether their method produces it. Record the rejected options and the reasons in writing at the design stage, so the conversation happens once rather than after fieldwork.
What should a method recommendation actually contain?
The decision it serves with its date, each objective mapped to a question shape and a verdict, the specified design (who, how many, over what period, in what channel, analysed how), the options rejected and why, the compromises forced by constraints and what each costs, and an explicit list of the questions the design will not answer.
The skill chain
Works well with.
Research where people already are.
Analyse it where you already work.
Yazi helps researchers conduct surveys, AI interviews and longitudinal research directly through WhatsApp.
