User experience
and interface design
in Cambridge

A persona built without research is a guess wearing a name tag. It looks convincing. It describes the person a team hopes exists rather than the person who actually opens the product each morning. Interviews, support transcripts, and session recordings hold better raw material than a brainstorm in a conference room. The work is pattern recognition, not imagination.

The first draft of a persona should read like a case file, not a magazine profile. It needs goals, constraints, and the tools already in use. It needs the moment a person decided to look elsewhere. A name and a photo add nothing. What actually matters is a short list of behaviors a designer can point to when a feature debate starts running in circles.

Teams often want one persona to cover an entire market. It never works. A product used by a solo founder needs a different picture than one used by a twelve person operations group. Merging the two produces a persona nobody recognizes, and screens that serve no one well.

Quotes pulled directly from interviews carry more weight in a review meeting than any summary paragraph. A stakeholder who hears the actual words a customer used argues less with the direction a design takes. Paraphrasing loses that specific language. And that language often becomes the copy on a screen later.

A persona needs an expiry date. Markets shift. A product adds a segment. Competitors change what people expect by default. A document written eighteen months earlier starts describing a company that no longer exists. Revisiting personas on a fixed schedule keeps the document from becoming decoration on a wall.

Behavioral data from analytics should sit next to the qualitative picture, not replace it. Numbers show what people do. Interviews explain why they do it. A drop in a particular flow is a fact. That fact still needs a reason attached before a team can act on it with any confidence.

The test of a good persona is whether it changes a decision. A document can sit for months, unopened. If nobody opens it before a roadmap meeting, the research behind it was still valuable, but the artefact itself was not. A persona earns its place by showing up in arguments about scope, priority, and tone, not in a kickoff deck alone.

Some teams skip personas entirely. They rely instead on a handful of named real customers, referred to by first name in every meeting. That approach can work at a small scale, where everyone actually remembers who these people are and what they said. It tends to break down once the team grows past a size where shared memory can no longer substitute for a written reference.

A critique session without structure turns into whoever speaks loudest deciding the direction. That helps no one. Design work gets stronger when the room agrees on the problem a screen is solving before anyone comments on color or spacing. Starting with the goal keeps small opinions about button shade from derailing time that should go toward the real decision.

Most critiques work better with a small, fixed group rather than an open invitation. Five or six people who understand the project give sharper feedback than fifteen people skimming a screen for the first time. A larger group can review it later. The working session benefits from people who already carry context.

Feedback framed as a question moves a critique forward faster than feedback framed as a verdict. Asking what a layout is trying to achieve invites the designer to explain reasoning that might already answer the concern. A flat statement that something looks wrong gives the room nothing to build on. It just restates a personal preference.

A timer sounds mechanical. It keeps a critique honest. Ten minutes per screen forces a room to raise its most important point, rather than working through a list of small preferences. Screens that need more discussion can get a second, dedicated session, so that time does not eat into the rest of the review.

Not every comment deserves equal weight. Pretending otherwise slows a team down. A note about a genuine usability problem belongs ahead of a note about a font weight. Someone in the room needs the authority to make that call out loud. Otherwise a critique produces a flat list, where the smallest issue reads the same size as the largest.

Written notes captured during the session save a designer from reconstructing memory the next morning. A shared document, updated in the room rather than after it, also guards against a common complaint. The decision did not quietly change between the session and the follow up. Everyone leaves having seen the same summary.

A critique held only once, right before launch, arrives too late to change anything structural. Early sketches are cheap to redirect. Reviewing them lets a team catch a wrong direction while it still costs an afternoon to fix, rather than a sprint. Later reviews then focus on refinement, not on rescuing a screen that solved the wrong problem from the start.

Remote teams lose some of the value of a critique held in person, since body language and the pace of a room carry information a video call flattens. Screen sharing with a live cursor helps. So does a short silent period where everyone writes a comment before speaking. Together they recover some of that lost signal without asking anyone to travel.

Moderated testing puts a researcher in the room, or on the call, while a participant works through a task. Unmoderated testing sends the same kind of task to a participant who completes it alone. A platform records the screen and any comments spoken aloud. Each format answers a different question. Picking the wrong one wastes a budget.

A moderator can follow up on confusion the moment it happens. If a participant hesitates over a button, the moderator can ask what they expected next. That single question often explains an entire pattern of drop off seen later in analytics. Unmoderated sessions cannot ask that question. Ambiguity in the recording often stays.

Unmoderated testing scales in a way moderated sessions cannot. A study can reach dozens of participants across different time zones within a day or two. Nobody has to watch every session live. That volume matters when a team needs a number. A useful example is how many people complete a task without help, rather than a rich explanation from one person.

Early concepts, where the interface is still rough, tend to need a moderator in the room. A participant looking at a low fidelity sketch needs occasional context only a person can supply. A fully unmoderated session risks something else. A participant may abandon the study out of confusion that has nothing to do with the actual design question.

Late stage testing on a near final interface often works well unmoderated. The screen should already explain itself, without a guide standing by. That removes the moderator as a variable. A participant who might soften criticism in front of a live researcher tends to write blunter notes alone.

Cost and turnaround time push many teams toward unmoderated tools by default. That default is not wrong for the right question. The mistake is skipping moderated sessions entirely. That happens once a team gets comfortable with the faster format, and some questions about reasoning simply do not survive translation into a written log.

A blended plan often works best. A handful of moderated sessions, followed by a larger unmoderated round, covers both needs. The moderated round surfaces the questions worth asking at scale. The unmoderated round then supplies the sample size needed to trust the pattern.

Recruiting quality affects both formats, but it hides more easily in unmoderated studies. That is a real risk. A participant rushing for an incentive payment looks the same, in the data, as someone genuinely confused. Screening questions and a minimum time on task help filter out sessions that would otherwise quietly distort the results.