Contextual Inquiry: interviewing people inside their work

Context. Go where the work happens and study ongoing, concrete instances — this order, this claim, this patient — not generalised descriptions. Artifacts on the desk, sticky notes on the monitor, and interruptions are data.

Partnership. The master–apprentice stance: the participant leads with their work; the researcher alternates watching with asking, adjusting the traditional interviewer power balance so the participant teaches.

Interpretation. Facts are not enough; the method requires assigning meaning — and, distinctively, sharing interpretations with the participant in the moment (“so you keep that spreadsheet because you don’t trust the system’s totals?”) so they can confirm, correct, or refine. Participants are reliably accurate at validating interpretations of their own work, even when they could never have volunteered them.

Focus. A declared project focus steers attention without a script — the inquiry follows the work, but with a defined lens (say, handoffs and rework) that keeps sessions comparable and findings actionable.

The method was developed by Hugh Beyer and Karen Holtzblatt as the data-gathering foundation of Contextual Design; the canonical treatments are their book Contextual Design (Morgan Kaufmann) and the compact statements in Contextual design and the earlier CHI paper Contextual design: an emergent view of system design.

Its intellectual debt is to ethnography — fieldwork, artifacts, work-as-lived — industrialised into sessions of one to three hours that product teams can actually run, with interpretation sessions and consolidated work models turning field data into design input.

The method’s scope has widened with work itself: “context” now includes home offices, mobile work, and screen-mediated tasks, where the inquiry happens over screen-share with the same principles intact.

• Designing or replacing systems that must fit existing work — the method exists to surface the real workflow, including the workarounds the new system will otherwise destroy.
• When accounts and behaviour are known to diverge: compliance-heavy settings, expert work, anything people describe with “it’s pretty straightforward” — the tell that it is not.
• Early discovery, before requirements exist, when the team needs the work’s actual structure rather than validation of a concept.
• Not for rare or unschedulable events (emergencies, quarterly closes happening off-cycle) — use retrospective walkthroughs anchored to artifacts instead, and say so.
• Not for attitudes, markets, or motivation beyond the work itself; the method studies practice, superbly and narrowly.

An insurer preparing to replace its claims platform ran contextual inquiries with ten processors across two offices, focus declared as “where claims wait and why”. The conference-room account, gathered earlier, described a clean six-step flow with occasional supervisor referrals.

The desks told the truth. Every processor maintained a personal shadow system — spreadsheets, printed queues, annotated sticky hierarchies — whose common function emerged through in-session interpretation: the official system’s work queue hid claims that were waiting on external documents, so processors exported and re-sorted their queues to keep promise-dates visible. One master apprenticed the researcher through her “Friday sweep”, a weekly ritual reconciling three tracking surfaces; asked why, she said the sentence that reframed the project: “the system knows what’s in it, I know what’s owed.”

Interpretation sessions consolidated the models: the breakdown was not step four (as the replacement’s requirements assumed) but the system’s inability to represent waiting-on-external states. The programme rewrote its requirements around commitment tracking, and the shadow spreadsheets — slated in the old plan for “elimination through training” — were recognised as the spec, written by the users in the only medium they had.

• Conference-room “contextual” interviews. Same questions, wrong room, none of the artifacts or interruptions that carry the data.
• Demo capture. Letting participants perform an idealised version; the corrective is concrete current work — “do today’s, not the training example”.
• Watching without interpreting. Passive observation defers meaning-making to the analysis room, where the participant can no longer correct it.
• Interrogating flow. Interrupting at the wrong grain until the work becomes an interview; apprentices watch sequences complete.
• Focus creep. Sessions wandering off the declared lens, producing rich data about nothing comparable.
• Skipping interpretation sessions. Ten field visits reduced to one researcher’s recollections — the method’s team-scale value discarded.

Presence changes work: participants tidy, slow down, narrate. Prolonged sessions and concrete-task discipline dampen the performance, and the remaining reactivity should be reported, not denied.

The method samples occasions, not calendars — it sees the work that happened while you were there. Rare events, seasonal peaks, and night-shift realities need deliberate scheduling or honest exclusion.

And it is resource-intensive per data point: skilled researchers, on-site hours, team interpretation. The economics only work because the alternative — building on imagined workflows — is more expensive still; but for questions a survey can answer, this is the wrong tool.

Participant observation is the longer-immersion ancestor; contextual inquiry industrialises it into product-team sessions.

In-depth interviews collect accounts when the work cannot be visited — weaker on practice, necessary at scale.

Usability testing evaluates a designed artifact against tasks; contextual inquiry discovers the work the artifact must fit.

Diary studies extend observation across time where sessions cannot follow.

Contextual inquiry produces messy, mixed evidence — session recordings, photos of artifacts, verbatim corrections, interpretation notes — and the consolidation step is coding across it: breakdowns by type, workarounds by function, quotes by model. Evidano transcribes session audio, codes the corpus against the project focus with quote-level links, and lets interpretation sessions query “every waiting-related breakdown across all ten sites” instead of relying on whoever remembers.

The apprenticeship itself — the noticing, the in-the-moment interpretations, the artifact chasing — is craft that happens at the desk. The tool guarantees what was learned there survives contact with the team.