Field Research

For a long time research happened in meeting rooms. You invited a user to your office or opened a video call, asked questions for an hour, and wrote up the answers. It was efficient — the questions could be prepared in advance, and three sessions fit into one day.

The trouble is that people tidy their own work up when they describe it. Ask how someone finds a document and most will say they search for it. Go to that person's desk and you find a sticky note along the edge of the monitor with the addresses of the documents they open most, and a spreadsheet of links parked permanently on the second screen. It was too ordinary to be worth mentioning, so it never came up.

Sticky notes are not the only thing that goes missing. A teller cannot spend a minute on a screen while a customer is standing in front of them. A plant supervisor wearing gloves cannot hit a small button. Asking the person at the next desk is faster than reading the help page. None of those facts surface unless you are standing there. Field research was invented to look at the place where answers get made, instead of collecting the answers themselves.

The work runs in six steps. Unlike an interview, what you prepare is an observation plan rather than a question list.

1. Write down what you want to see, in one sentence. Not "understand the site" but something you can confirm with your eyes, like when and how a teller opens our product while a customer is waiting.
2. Pick the places and the people. Two companies in the same industry work differently at different sizes, so mix two or three. Visiting two branches and one head office separates what is true of the industry from what is true of that one company.
3. Schedule the visit for the hours when the work actually happens. A quiet Tuesday afternoon shows you nothing. Go at the lunch rush, or at the month-end close, because friction only appears when the place is busy.
4. On site, let people work instead of assigning them tasks. Handing someone a task belongs to the usability test, where you watch a person attempt a defined job. Here you watch the job they already had. Save your questions for the end rather than interrupting a hand mid-motion.
5. Split the recording between two people. One follows the hands and the screen changes, the other photographs the desk and how things are arranged.
6. Come back, organize what you saw into scenes, and connect them to a decision. If the same scene showed up at two of the three sites, write down what changes because of it.

The method pays off most when the product is not used at a quiet desk. Checkout counters, warehouses, hospital front desks, factory lines — places with noise, queues and equipment — impose constraints that decide the outcome no matter how good the screen design is. It also fits when half the workflow happens outside your product. If someone moves between your tool, a paper ledger and the company chat app, those round trips are invisible from inside your own analytics.

Anything where words and behavior are likely to diverge belongs here too. Someone who says they follow the manual may be skipping two steps, and only observation catches that.

The reverse case is short use that can happen at any hour of the day. The scene has to occur inside the two hours you happen to be standing there. Subjects like that suit a diary study, where users log their own activity for about two weeks. Visits cover work that runs long in one place; diaries cover work that is scattered and brief.

There are three limits. First, it's expensive. One site takes half a day including travel, and six interviews fit in the same slot. Second, people behave differently when they are watched. For the first half hour they follow the rules more carefully than usual, and their real habits appear after that, so don't judge from the opening stretch. Third, what you saw at three sites is the story of three sites and cannot be turned into a percentage. What you get is a hypothesis; how common the behavior is gets confirmed separately, through a survey that puts the same question to many people, or through usage logs. And sometimes a client's security policy simply won't let an outsider stand in front of a working screen, so the visit never happens.

"I think we need to go see this in person. Interviews keep circling the same answers."

The interviews have stopped producing anything new, so switch to observation.

"I've booked two distribution centers for next week, during the morning peak."

A visit is happening, and the speaker has already decided that a quiet hour would be worthless.

"They all prop the tablet on a stand. We designed it assuming people hold it."

One observed fact has overturned a design assumption.

When Pagedeck started selling its document tool to plant supervisors, Mia's first move was a visit. Every supervisor on a video call said the tool was working fine, yet actual usage fell by half after the first month. Three factories made the picture simple. Nobody carried a tablet around the floor. They collected a day's worth of notes in a pocket notebook and typed everything into a desktop computer in the office at the end of the shift.

The conclusion was not to build more features but to change the entry screen so several rows could be filled in at once. That direction was unlikely to come out of another round of video calls. Nobody mentions a pocket notebook.

The most common mistake is assuming the visit itself produces the result. Watching looks easier than interviewing, so it seems fine to show up unprepared. But someone who goes without deciding what to look for comes back with twenty interesting scenes and no way to say why any of them matters. Nothing connects to a decision, and the request for a second visit doesn't get approved.

Confusing it with a sales visit is also common, because going to a customer's office looks the same from the outside. The purposes are opposite. Sales goes to deliver your side of the story; research goes to collect theirs. Mix the two into one trip and the host stops working normally and starts demonstrating how well they use the product, which makes every scene from that day unusable.