Realistic tasks
Goals, not instructions: “book an appointment with a GP before Friday” or “ask for a refund on the order that arrived yesterday”. Never “click the menu in the top right corner”.
Service · Software testing and QA
Usability and UX testing
The question is not whether the button works. It is whether anybody spots it. A technically flawless appointment portal can shed one patient in three at the step where a reason for the visit has to be picked, purely because the list borrows the vocabulary of the practice software instead of the words people say out loud. Sessions run remotely: each participant sits at home with the laptop or the handset they always use, on their own connection, inside their own habits. Real conditions, then, rather than a laboratory.
Five people
enough for a first wave
Recording
the screen and what they say aloud
Mobile
where most journeys are decided
Ranking
ordered by effect on the goal
Where this service reaches
The method is cheap and plain: the participant receives something to achieve, and whoever runs the session says nothing at all, whatever the temptation.
Observation
We note the hesitations, the steps retraced, the label read twice over, the moment somebody abandons the attempt. Thinking aloud is the one thing we ask of them.
Devices
A handset carries most of the sessions. Whatever sits comfortably on a twenty-seven inch monitor tends to end up tucked under a collapsed menu on a six inch screen.
Accessibility
Keyboard navigation, NVDA or VoiceOver, contrast, the size of tap targets, field labels. WCAG 2.1 AA in general, the RGAA 4.1 framework where a public body is involved; whether your own activity falls under the European Accessibility Act is a question for your legal adviser.
Forms and payment
The most treacherous stretch of any journey: surplus fields, error messages that never say what to correct, typed entries wiped out by a step backwards, an account forced on somebody before they are allowed to pay.
Ranked report
Each problem placed by its effect on the goal being measured, with a video clip of a few seconds and a rough cost in repair work. Fifteen seconds of footage ends more arguments than a full meeting.
The way an engagement runs
A wave of five participants fits inside a short project. When the report lands is stated in the plan, preparation of the tasks included.
01
The opening question
We settle what the study has to answer, for instance why so many visitors walk away at the point of choosing a slot.
02
Recruitment and tasks
Participants drawn from your real audience, and tasks written as goals rather than as the route you hope they take.
03
Remote sessions
Conversations over Google Meet or Teams, screen shared and recorded, in which we deliberately give no help.
04
Analysis and recommendations
One-off reactions get separated from the obstacles several people run into, and specific changes follow.
Statistics give you the effect, watching gives you the cause. Analytics report that half of all visitors stop at the delivery step. What they never report is the “Continue” button hiding below the map of pickup points, with nobody thinking to scroll. Five minutes with somebody thinking aloud settles it.
Questions and answers
Among your current customers or service users; a newsletter and a small thank-you usually does it. Failing that, colleagues from another department who have never seen the product will serve nicely. Anyone who designed the interface is disqualified: the hand moves to the right spot on reflex.
This is in fact the cheapest moment. On a clickable Figma mock-up the whole structure can be rethought while a change costs a few hours of design work, as against three weeks of development and another round of checks afterwards.
For one wave, yes. The first few trip on identical obstacles and the notes soon start repeating themselves. Five people, then repairs, then five more will teach you more than twenty in a single sitting.
Every participant signs a consent form covering the purpose, the people who see the material and how long it is kept. Recordings stay on servers inside the European Union, access goes only to those working on the study, and the lot is deleted on the agreed date. The matching entry joins your record of processing activities.
More from this area
Service
Functional testing
Proof that the product behaves the way the brief says
Service
Load and stress testing
Finding out what happens when everybody arrives together
Service
Test automation
Repetitive checks handed over to the machine
Service
Penetration testing
An intrusion attempt run under contract and written consent
Your service, seen from their side
Name the step where your figures show the heaviest drop-off. The sessions get run and the real causes come back to you.
When we are around
Weekdays, 8:00 to 18:00 CET; answers land inside one working day
Weekdays, 8:00 to 18:00 CET; answers land inside one working day
Talking it through
A call on Teams or Google Meet, whenever writing is not enough
A call on Teams or Google Meet, whenever writing is not enough
Message received
An answer follows inside one working day. Report an outage that is stopping people working and it moves ahead of everything else.
- Anything unclear gets asked. Where a figure needs detail we do not have, expect an e-mail or an offer of a short call on Teams or Google Meet.
- A proposal follows. What is covered, the figure in euros excluding VAT, and a start date that will hold. Nothing hides beneath an asterisk.
- Then it is your call. The proposal sits in your inbox for as long as you need. Query any line of it, and decide once you are satisfied.