Research

Watching people attempt real tasks, which is the most direct evidence available about whether a design works.

Research method · Evaluative

Experts reviewing an interface against a defined set of principles, which is fast and cheap but not a substitute for users.

Research method · Evaluative

Structured conversations about what people actually did, which is far more reliable than what they say they will do.

Research method · Generative

Good for measuring how widespread something is, and poor for discovering what that something is.

Research method · Quantitative

Asking users to group content, which reveals their mental model of a domain rather than your org chart.

Research method · Information architecture

Testing whether people can find things in a structure, with no visual design to confound the result.

Research method · Information architecture

Measuring where people click first, which predicts task success better than almost any other single measure.

Research method · Evaluative

Showing a design briefly and asking what people remember, which tests hierarchy and clarity of message.

Research method · Evaluative

Comparing variants with real traffic, which tells you what happened without telling you why.

Research method · Quantitative

Measuring what people do at scale, which tells you where problems are and never why they exist.

Research method · Quantitative

Measuring progression through a sequence, where the largest drop is rarely the biggest opportunity.

Research method · Quantitative

Watching people work in their own environment, which surfaces the context that interviews miss.

Research method · Generative

Participants record their own experience over time, which captures things a single session cannot.

Research method · Generative

Turning raw observations into findings, which is where most of the value and most of the bias enters.

Research method · Analysis

Organising, labelling and structuring content so people can find it, which determines navigation rather than following from it.

Research method · Information architecture

Fast, informal testing with whoever is available, which is far better than no testing and worse than proper recruitment.

Research method · Evaluative

Measuring where people look, which is expensive, frequently over-interpreted, and occasionally irreplaceable.

Research method · Evaluative

Framing user needs as the progress someone is trying to make, which separates the need from any particular solution.

Research method · Generative

Study how others solved the same problem to learn the conventions your users already hold, not to assemble a feature list.

Research method · Evaluative

Group individual observations until themes emerge from the data instead of from the categories you brought with you.

Research method · Analysis

Structured feedback against stated objectives, which is a different activity from a decision meeting.

Workflow · Practice

Getting a design built correctly, which works better as a conversation than as a document.

Workflow · Practice

Checking the built thing against the intent, before release rather than after.

Workflow · Practice

Working out what problem to solve before deciding how, which is where most product failure is actually prevented.

Workflow · Practice

Choosing how finished a prototype needs to look, which is decided by the question you are asking rather than by taste.

Workflow · Practice

Deciding what to fix first, using evidence rather than the volume of the most recent complaint.

Workflow · Practice

Turn a pile of research into one statement narrow enough to design against and wide enough to allow more than one answer.

Workflow · Process

Generate enough distinct options that the first idea has to compete, then converge deliberately rather than by exhaustion.

Workflow · Process

Map the decisions, states and exits before drawing screens, because most missing work hides in the branches nobody drew.

Workflow · Process

Work out structure and priority while changes are still cheap, and stop before the fidelity starts answering questions you have not asked.

Workflow · Process

Pick one behavioural measure per goal, pair it with a quality measure, and decide the threshold before you look.

Workflow · Process