Contextual inquiry: how to observe and interview people while they work

Contextual inquiry is a research method in which you watch someone do their real work, in the place they normally do it, and ask questions as they go. Nielsen Norman Group describes it as a type of ethnographic field study that combines in-depth observation and interviews with a small number of users to understand their work practices and behavior (Nielsen Norman Group). Karen Holtzblatt developed the technique, and it forms the core of Contextual Design, the method she and Hugh Beyer set out in their book of the same name (Elsevier).

The reason to do it is that people leave things out when they describe their routines from memory. In an interview they give you a tidy summary. While they work, they can show you the extra steps, workarounds and habits the summary leaves out, and explain each one at the moment it happens.

When it's worth the effort

NN/g says contextual inquiry suits complex systems, detailed processes and the work of expert users, and that teams usually do it early, while deciding what to build. Its article gives an example from insurance software. Specialists who entered vehicle data all said in interviews that they pasted it from a spreadsheet. When the researcher watched them, they were also looking up missing details in a second system, and clicking save after every entry even though the software saved automatically.

The method is less useful for simple screens, such as a signup form, where a usability test answers the question faster. When people can't be interrupted, as with doctors seeing patients, NN/g suggests silent direct observation with questions afterward.

The apprentice model and four principles

The method borrows the relationship between a master craftsperson and an apprentice. The participant is the expert doing the work, and you learn by watching and asking. NN/g summarizes four principles behind it. In plain terms:

• Context. Go where the work happens, not to a conference room.
• Partnership. The participant helps steer the session; you don't control every topic.
• Interpretation. Say what you think is happening and let them correct you.
• Focus. Know what you're there to learn, so you know what to watch for.

GOV.UK's service manual makes a related point about visiting people at work: the familiar setting makes it easier for them to remember details, and you see behavior you can't capture in a lab. It also notes that some people are uncomfortable having researchers at their workplace (GOV.UK).

A worked example

This example is hypothetical. A startup is building scheduling software for companies that maintain heating, cooling and electrical systems in office buildings. In interviews, dispatchers described their job simply: a request comes in, they check who is free and they assign the job.

The team arranges to sit with two dispatchers for a morning each. They watch one dispatcher take a call about a broken air-conditioning unit, and then:

• Check a paper list taped to her monitor showing which technicians are certified for that building's equipment
• Text a technician to ask whether he has the right part in his van
• Call the building manager back to confirm access, because the request didn't say who would let the technician in
• Re-enter the job into the billing system by hand

None of these steps came up in the interviews. The paper list and the parts check suggest the real scheduling problem is matching skills and parts to each job, not just finding someone who is free.

How to run a session

NN/g describes a four-part structure:

1. Primer. Introduce yourself, explain what you hope to learn and that you'll want them to correct your interpretations, discuss confidentiality and get permission for any recording. Ask for a quick summary of the work they'll be doing.
2. Transition. Say clearly that from now on you'll watch them work as they normally would and interrupt when you see something you want to understand. Ask them to tell you when a moment is a bad time to interrupt.
3. Contextual interview. Watch, then stop to ask when they do something you don't understand, or to check an interpretation: "So the paper list is where you check certifications, not the scheduling system?" Notice the other tools, notes and people they rely on.
4. Wrap-up. Summarize what you learned and let them correct you.

Sessions can run from an hour or two to several days, depending on how much of the work you need to see.

Some practical points. Ask the participant's employer for permission before visiting a workplace, and agree in advance what you may photograph or record. If the work involves customer or confidential information, agree on what you will leave out of your notes. How to ask for recording consent covers that conversation. GOV.UK also advises not to go alone to a location away from your office and to let your team know where you are. Take notes on what people do as well as what they say; how to take interview notes suggests a format.

If you can't visit, watching someone do the work over a screen share still shows the steps on their computer, though you lose the paper lists, phone calls and conversations at the next desk.

Mistakes to avoid

NN/g lists four common risks:

• The participant slips into interview mode and starts summarizing instead of working. A clear transition helps; if it happens, remind them you want to see the work done as usual.
• The session becomes a list of complaints about current software. Redirect them to the work and the reasons behind each step.
• You bring your own assumptions and see what you expected. Treat every observation as equally important until you've compared sessions.
• You change their behavior by asking about it. Ask open questions, and save most of your interpretations for the wrap-up.

After the sessions

Meet with everyone who observed soon afterward and pool what you saw. Beyer and Holtzblatt's book includes structured interpretation sessions and an affinity diagram for consolidating what many participants showed you, and NN/g also suggests affinity mapping to find themes. Affinity mapping walks through that step. In the facilities example, the team would put the two dispatchers' steps side by side and mark which workarounds both used.

Your next step

Write down the one process you most need to understand and the job title of the person who does it every day. Ask two or three of those people whether you can watch for an hour or two while they work, and list beforehand what you'll look for.

If you don't yet know anyone who does the work, a short call can come first: ask them to walk you through the process, then ask whether they'd be open to a longer observed session. Instant Expert can find people who match a description you write, such as "dispatchers at commercial facilities maintenance companies." You review who it finds, it sends your invitations, and you pay only for calls that get booked. The operations professionals in facilities management directory page is one place to start.