User testing, watched.
Also called: Usability testing · In-context trials · Field testing · Real-user observation
Putting a working prototype in real users’ hands, in their own setting, then watching where they stumble without stepping in to help.
Give a real user a task, not a tour, in their own kitchen. Then sit on your hands. The urge to explain is the enemy: every rescue hides the exact confusion you came to find. Watch where they stumble, write it down, change the product not the user.
What user testing is
User testing is the moment you stop guessing how people will use the thing and watch them actually do it. You hand a real user a working prototype, set them a genuine task, and observe. The product is on trial, not the person. Everything they fumble is a fault in the design, never a fault in them.
The discipline is harder than it sounds, because every instinct pulls the wrong way. You want to demo the clever bits. You want to explain the control before they touch it. You want to leap in the second they look puzzled. Resist all three. A demo teaches them your mental model and erases theirs. An explanation papers over the exact gap you came to measure. A rescue ends the test at the most useful moment. Give a task, not a tour, then go quiet.
Task, not tour
“Have a look at this and tell me what you think” gets you politeness. “It’s Friday night, you want bread for the weekend, get a prove going in this” gets you behaviour. A real task with a real goal surfaces what a guided walkthrough never will, because the user has to navigate the product on their own terms, in the order their head invents, not the order your slide deck imposes.
Do not rescue
When a user goes quiet and frowns at a dial, the silence is unbearable and the temptation to help is enormous. Sit on your hands. That ten seconds of confusion is the most valuable data in the whole session. If you rescue them, you have learned that the product works fine with you sitting next to it, which is not how anyone will own it. Let them struggle. Note exactly where, and how they try to recover.
A user test on the proofing box
For the proofing box we did not run a lab. We lent working units to a handful of Sourdough School bakers and let them use them overnight, in their own cold kitchens, for a fortnight. Here is the shape of what we set up and what we watched, so you can see a real test rather than a generic template.
The fix only existed because nobody jumped in. Had we phoned each baker the first night and talked them through it, the lid-opening would never have shown up, and we would have shipped a box that quietly failed in every kitchen colder than ours.
- You show them how the rotary works first.
- They nod along to your mental model.
- You jump in the moment they pause.
- You learn the product works with you in the room.
- You hand it over with a task and stay quiet.
- They build their own mental model in front of you.
- You let the ten-second frown happen and note it.
- You learn how the product behaves when you are gone.
How it fits the bigger picture
User testing is activity 08.20.04 in the framework, inside Stage 08 Develop. It takes the working prototype and exposes it to real users before the product hardens. What it surfaces feeds straight into the CE/UKCA test (08.20.05), where the design is held against the BS EN 61010 safety requirements before it can carry a UKCA mark.
What it can do
It shows you how the product behaves in the wild, in a real kitchen, with a real user who has no idea what you intended. It catches the confusions a spec review never will, because it measures what people do rather than what they say they will do. A fortnight of real overnight use teaches you more than a month of opinions.
What it can’t do
It can’t tell you the design is safe or compliant. That is the job of the CE/UKCA test that follows. User testing finds usability faults; it does not certify that the heater and low-voltage electronics meet the standard. And five bakers are a signal, not a verdict, so treat strong patterns as worth fixing and single odd reactions as worth watching, not chasing.
See the full 10-stage process →
Try it yourself
Find three to five people who match your real user, not your friends and not your team. Give each one a genuine task in their own setting, hand over a working prototype, and then say nothing. Watch where they hesitate, what they touch by mistake, what they never find. Write down the stumble, never the excuse. Change the product, not the person. Start the Free Sprint → if you want a structured way to define what you’re testing first.
Your user-testing checklist
Project notes: the lid they kept lifting
▸ From the notebook · optional reading
A fortnight of lent prototypes in Sourdough School kitchens, and the silent week-one mistake that nearly made me pick up the phone.
3 min read · click to open
We had five working units and a clean theory: at £149 the box had to earn its place on a kitchen counter, so the rotary and OLED needed to be obvious enough that a baker would set 26°C, leave it, and sleep. To find out, we lent the boxes to five Sourdough School bakers and asked them to use them overnight, in their own kitchens, for two weeks. The only instruction was the task: prove a loaf the way you normally would.
From my kitchen in Stockport I could see the temperature logs the next morning, and the first week was agony. One baker’s log was a sawtooth: 26°C, then a cold dip, then a slow climb back, over and over through the night. My instinct was to phone her and ask what was wrong. Anna stopped me. “If you call her, you’ll fix it for her and learn nothing. Let it run.”
What the silence taught us
So I sat on my hands. At the end of week one I asked each baker, gently, to talk me through their night. The sawtooth baker said it plainly: “I didn’t trust it. I kept lifting the lid to check the dough hadn’t gone cold.” Every lift dumped the heat the box had just built. Two of the five were doing it. The OLED showed a number, but a number alone did not reassure a nervous baker at 1am.
If I had rescued her in week one, that would never have surfaced. The fix was small and the test paid for it: we added a steady “holding 26°C” line to the display, so the box visibly told you it had the situation in hand. We pushed updated firmware to the units for week two, and the lid-lifting stopped. The bread, by their own reports, improved.
The whole lesson cost us nothing but the discipline to stay quiet. The hardest part of user testing is not setting it up. It is keeping your hands off it while it runs.
— Develop stage, project notes, 2026
&mda