What it is
Two expert reviews needing no participants. In a heuristic evaluation, several evaluators independently inspect an interface against recognized principles, such as Nielsen's ten heuristics, then combine and rate findings. A cognitive walkthrough asks, at each task step, whether a new user would know what to do. Both predict problems; neither observes users.
Why use it
Expert review finds many problems cheaply before users see the design and focuses scarce testing on what experts cannot answer. Several evaluators matter: each misses problems others find.
Use this when
- A hi-fi prototype or build exists and user testing is not yet possible
- You want to clear obvious problems before usability testing
- Time, access, or budget limit user research
- A task flow needs a first-time learnability check
You may not need this when
- User testing is planned soon and can find the same problems
- No evaluators with usability expertise, other than the designers, are available
- The question is about needs or preferences, which expert review cannot answer
Dependencies — what you need
- A prototype or build covering the flows to review
- Evaluators with usability expertise (commonly three to five), ideally not the designers, reviewing alone first
- A heuristic set, tasks, and user types
- A severity scale agreed beforehand
People involved
- Design
- Research
- Product
Typical outputs
- A consolidated problem list tied to heuristics or walkthrough steps
- Severity ratings and suggested fixes
- Walkthrough findings by step
- Issues to confirm with users
Template
Heuristic Evaluation template
A fill-in structure for running this method: what to prepare, what to do, what to record, and how to read the result.
How AI can help
AI as augmentation — not replacement
- Generate a review checklist from the chosen heuristics
- Draft walkthrough questions for each step
- Consolidate duplicate findings for evaluators to confirm
- Draft write-ups that carry the evaluators' own ratings
AI prompt
A prompt for any AI assistant. It works only from material you paste in, keeps assumptions apart from findings, and leaves the decisions to you.