What is Formative Research?

Formative research is evaluation conducted during the development of a product, service, or programme, with the purpose of informing and improving it before it is finished: identifying usability problems in a prototype, testing whether a concept is understood, discovering unmet needs a design should address, checking whether messaging lands. The term comes from educational evaluation (Michael Scriven's 1967 distinction between formative and summative assessment) and the famous shorthand captures it: when the cook tastes the soup, that's formative; when the guests taste the soup, that's summative. Summative research measures the outcome of a finished thing against goals or competitors (benchmark studies, standardised scores, A/B verdicts, launch evaluations); formative research exists to make the summative result better before it's measured.

How Formative Differs in Practice

The purpose changes the design at every level. Sample: formative studies are small (five to eight participants per round) because their job is finding problems, not estimating prevalence, and problems recur; summative studies are larger because their job is measurement, and measurement needs precision. Method: formative work is qualitative-leaning and observational: think-aloud usability tests, concept tests, interviews, prototype walkthroughs; summative work is quantitative-leaning and controlled: fixed tasks, measured outcomes, comparison against a baseline or benchmark. Artefact: formative studies test rough things (sketches, wireframes, clickable prototypes), because roughness invites honest criticism and costs nothing to change; summative studies test the real thing. Output: formative findings are a prioritised list of problems and opportunities with recommended changes, delivered fast; summative findings are metrics with confidence intervals, delivered as a verdict. Cadence: formative research is iterative by nature, rounds following fixes; summative research happens at milestones.

Running It So Findings Form the Product

1. Test early enough to matter.
The value of a formative finding is inversely proportional to the cost of acting on it; a problem found in a sketch costs an hour, in code a sprint, in production a quarter. Start before the team feels ready.

2. Recruit the audience, roughly.
Formative work tolerates looser representativeness than summative work, but not the wrong audience; problems that only experts hit are invisible to novices and vice versa. Screen for the segment whose problems you're hunting.

3. Prioritise by severity, not by count.
Problems are scored on impact and frequency (does it block the task? how many hit it?) so that the fix list orders itself; a rubric keeps prioritisation consistent across rounds.

4. Report as recommendations, fast.
Findings framed as "change X because Y" within a day or two of the sessions, ideally with the clips attached, so the design changes before the next round; a report three weeks later is summative by accident.

5. Close the loop with a re-test.
The next round checks the fixes and finds the next layer; formative research without re-testing is an opinion about what might have helped. Unmoderated prototype tests with screened participants, run and re-run as fast as the prototype changes (a Ballpark study is designed for exactly this cadence), keep the loop tight enough that the product is genuinely formed by the evidence rather than merely audited by it.

The Common Confusion

Teams regularly run a formative study and read it summatively ("80% task success" from five participants treated as a score), or run a summative study when formative was needed (a large benchmark on a design that had never been through a single round of problem-finding, producing a precise measure of a product with obvious flaws). The cure is naming the purpose before designing the study: are we improving this, or judging it? Everything else follows from the answer.

The Takeaway

Formative research tastes the soup while it's still cooking: small, fast, qualitative, iterative studies on rough artefacts, delivering prioritised fixes in time to act. Summative research judges the finished dish with measurement. Name which one you're doing, design accordingly, test early, prioritise by severity, report as recommendations, and re-test the fixes. The point of formative work is not to know how good the product is; it is to make it better before anyone measures.