What is evaluative research?
Also called: Evaluation research, Validation research
Evaluative research tests an existing design, prototype or product against a goal, to find out whether people can use it and whether it does what the team intended. It checks a solution, where generative research looks for the problem.
Updated
What it's for
Evaluative research answers narrow questions about something that exists: can new admins finish setup without help, which of two pricing page layouts do buyers understand, did the redesign cut support tickets about exports? Because the scope is tied to what's been built, it's good at showing where a design fails and less good at telling you what to build instead.
Evaluative research methods
- Usability testing, moderated or unmoderated: participants attempt tasks while you watch where they struggle.
- Prototype testing: the same idea on a clickable mockup, before engineering starts.
- Concept testing: checking whether an idea is understood and wanted, which sits between generative and evaluative.
- Tree testing, first-click and five-second tests: narrow checks on structure, layout and first impressions.
- A/B testing: comparing two live versions on a metric such as conversion.
- Expert inspections, such as a heuristic evaluation or cognitive walkthrough, when you need a fast read before recruiting.
Formative vs summative evaluation
- Formative evaluation happens during design, to find and fix problems. Small samples, qualitative findings, quick rounds. Most product usability testing is formative.
- Summative evaluation happens on a finished product, to measure it against a benchmark or a competitor. Larger samples and metrics such as task success, time on task or SUS score.
How to set up an evaluative study
- Write the decision the study supports, such as "ship the new onboarding or keep iterating", and the result that would change it.
- Write tasks around goals, not interface labels: "Invite a teammate to this project" instead of "Click Share".
- Recruit people who match the audience for the design, including new users if you're testing a first-run flow. Your most experienced customers will breeze through it.
- Decide in advance what you'll measure: problems found, task success, or a score.
Go deeper
- UX research methods guide
How evaluative methods compare with generative ones.
- Usability testing guide
The core evaluative method, from tasks to analysis.
- Concept testing guide
Evaluating an idea before there's a design.
- Unmoderated usability testing template
A ready test plan for a quick evaluative round.
- Usability testing in Great Question
Run usability tests with your own customers.
Questions
Is evaluative research qualitative or quantitative?
Both. A moderated usability test with six participants is qualitative evaluative research. An A/B test or a benchmark study with task success rates is quantitative. Formative evaluation leans qualitative; summative leans quantitative.
Can you do evaluative research with your own customers?
Yes, and for changes to an existing product they are usually the right participants. For evaluative research with potential buyers, or a first-run flow aimed at people who've never used the product, you'll also need participants from outside your customer base.
Run the study, not just the method
Great Question handles recruiting from your own customers, scheduling, incentives, interviews, tests and analysis in one place.
No credit card required.
Related terms
Full glossary- Generative research
Generative research is research done before there is a solution to test, to understand people's needs, behaviours and problems well enough to decide what to build. It produces opportunities and questions, not a verdict on a design.
- Time on task
Time on task is a usability metric that records how long a participant takes to complete a task, from the moment they start to the moment they finish or give up. It's used to measure efficiency and to compare designs or versions.
- Cognitive walkthrough
A cognitive walkthrough is a usability inspection method where evaluators step through a specific task in an interface and, at each action, ask whether a new user would know what to do, find the control and understand the result.