Test it where the work happens
A usability session in a quiet room with good lighting, a stable connection and a participant giving you their full attention will tell you whether your interface is comprehensible. It will not tell you whether it is usable, because comprehension is not the thing that fails in the field.
I have had two prototypes test perfectly indoors and fail completely within ten minutes of leaving the building. Both times the failure was environmental, and both times no amount of additional lab testing would have found it.
The first was an inspection app for utility technicians. It tested beautifully. Task completion was high, participants found it clear, one described it as obvious, which is about the best word you can hope for.
Then we ran a session in a car park in February, at minus nine degrees, with the actual insulated gloves the technicians wear for eight months of the year. Two core interactions failed outright. A swipe-to-complete gesture that gloves cannot perform reliably, and a set of forty-four pixel targets that are perfectly adequate for a bare thumb and useless for a gloved one.
Environment is a design constraint, not a variable to control for
Lab testing works by removing variables to isolate the interface. That is exactly right when the question is does this make sense. It is exactly wrong when the question is does this work, because the variables you removed are the conditions the product will actually live in.
Direct sunlight destroys contrast ratios that pass in a meeting room. Cold makes touchscreens erratic and fingers imprecise. A queue behind someone changes what they are willing to type. A phone held one-handed while the other hand holds a torch removes half your interface from reach. None of that shows up on a desk.
The second failure was social
On a lending product we tested the application flow in a branch lobby rather than a room. Participants who had been perfectly willing to enter salary and debt figures in a private session became visibly uncomfortable doing it where other customers could see the screen. Two turned the phone away from the room. One stopped entirely and said she would finish at home, which in the real funnel is indistinguishable from abandonment.
That is not an interface problem in any conventional sense. It is a problem about where the interface is used, and it changed what we put on a screen at the same time.
If you only test where testing is comfortable, you will only learn about the conditions where the product was already going to work.
How to do it without a research budget
- Spend one day observing before you design anything. Not interviewing. Watching, and writing down what the environment does to the work.
- Run at least one session in the actual place, with the actual constraints, even if it is only one participant.
- Test on the hardware your users hold, which is usually older and slower than yours.
- Photograph the working conditions. A photo of a phone in direct November sun settles a contrast argument in a way a citation cannot.
- Add the environmental constraint to your specification explicitly, so it survives handoff. Sixty-four pixel minimum, gloved use.
What lab testing is still the right tool for
Comprehension, language, information architecture, whether somebody can form a correct mental model of what your product does. That work is genuinely better in a controlled setting, and field sessions are worse at it because there is too much noise.
Use both. Test comprehension in the room, then take the thing that passed and go stand in a car park with it. The second session is usually cheaper, always shorter, and it has caught more genuinely product-threatening problems in my work than any other single method.