Research design · Lesson 1 of 3

Read the shape of the question

Name the shape first. It rules out most designs immediately and tells you what the rest would each cost.

9 min read · Free — no account needed to read

Most research questions are one of a small number of shapes, and the shape is what makes some designs appropriate and others incapable.

What exists? — descriptive. You are mapping something not yet documented. Descriptive surveys, observation, content analysis, archival work.

How common, or how much? — measurement. You need a sample that supports generalising to a defined population: surveys, secondary data, registry analysis.

Does A relate to B? — correlational. Cross-sectional if one moment will do, longitudinal if order matters.

Does A cause B? — causal. Experiments where assignment is possible; quasi-experiments, natural experiments and longitudinal designs with careful control when it is not.

How do people experience it? — interpretive. Interviews, phenomenology, ethnography.

Why does this happen — by what mechanism? — explanatory. Often qualitative or mixed, because mechanism is usually about process rather than magnitude.

How has it changed? — temporal. Longitudinal, cohort, historical, archival, time series.

What does the existing research collectively show? — synthesis. Literature review, scoping review, systematic review, meta-analysis.

Notice that a single topic contains several of these shapes, and they are different projects. "Remote work and productivity" splits into how many teams went remote, whether remote teams differ in output, whether going remote caused the change, how developers experience it, and what the literature collectively shows. Each is legitimate; only one of them is yours.

See it happen

You want to understand whether a new onboarding process improved retention.

The read that arrives on its own

Compare retention before and after the change — the numbers will show whether it worked.

The same situation, with this idea

You read the shape: this is causal. A before-and-after comparison is a design, and it cannot separate the onboarding from everything else that changed in the same period — hiring, the market, the team. Knowing that, you look for what would help: a group that did not receive it, a staggered rollout, or at minimum a plot over enough time to see whether the change coincides with the intervention or with something else.

After this, you can

name the shape of a question before reaching for a way to answer it.

The takeaway

Name the shape first. It rules out most designs immediately and tells you what the rest would each cost.

Easy to take the wrong way

That one project should cover several shapes. Splitting into the shape you can actually answer is usually what makes a project finishable.

Try this

Write your topic, then list every shape of question inside it. Circle the one you are actually asking.