Voice of the customer

Gemba interview protocol

Gemba interview protocol is licensed CC BY 4.0. Attribution: Katafacts (katafacts.com).

1 · What it is

What it is

A gemba interview happens where the work actually happens, watching it as it's done rather than asking someone to describe it from memory afterward. This protocol drafts questions built to reference what you're seeing in real time — 'I noticed you just...' — plus reminders specific to observing without disrupting the work you came to see.

2 · When to use it

When to use it — and when not to

Use it when

You can actually observe the work happening — sitting with a support agent, watching a customer use the product live, following a process end to end.
You want to catch workarounds and small frictions so routine the person doing them wouldn't think to mention them unprompted.
You want evidence grounded in what actually happened, not a recalled and possibly polished account of it.

Not when

You can't actually observe the work — that's a general discovery interview, working from recall instead of observation.
The conversation is about a decision that's already been made, not an ongoing workflow — that's the win-loss interview guide.
Your presence would meaningfully change how the work gets done and there's no way to account for that — note the gap directly if it happens rather than pretending it didn't.

3 · How to fill it in

How to fill it in

Research goal: What do you want to see, specifically? 'How support agents actually triage tickets' drafts a sharper protocol than 'understand support.'
Topics and questions: Drafted from your research goal — includes a 'setting the scene' opener and prompts built to reference what you observe live, not just planned questions.
Interview technique reminders: Drafted for gemba-specific technique — watching before asking, naming what you saw neutrally rather than what you inferred.
How to run this interview: Drafted from the research goal — how to observe without disrupting the work, and what to listen for.

4 · What good looks like

What good looks like

The example below is a gemba session watching a support agent triage a normal shift of tickets: questions built around real-time observation, a prompt for checking whether behavior changed because of the observer's presence, and reminders specific to watching work without interrupting it.

Same example, as a downloadable xlsx workbook.

Download .xlsx

Interview guide

Support agent gemba — ticket triage workflow

Marcus Lee, Support · 2026-02-15

Team: Dana Whitfield, Product · Sponsor: Alex Romero, VP Product

Research goal

See the real triage workflow as it happens — which steps take longer than they should, where a workaround has quietly become the norm, and what an agent notices that never makes it into a ticket note.

Topics and questions

4 topics · 8 questions

Setting the scene

Before we start, can you just work as you normally would — I'll ask questions as things come up.
What are you working on right now, and what triggered it?

In-the-moment observations

I noticed you just [specific action] — walk me through why.
What were you expecting to happen there, and is that what happened?
Tell me more about that pause — what were you checking or thinking about?

Workarounds and friction

Is this how you normally do this step, or are you doing something different because I'm watching?
What's the part of this you'd change first if you could?

Wrap-up

Was there anything today that was unusually smooth or unusually rough compared to a typical day?

Interview technique reminders

Watch first, ask second — don't interrupt an action in progress unless you'll lose the moment.
Ask about a specific action you just saw, not a general "how do you usually..." — the specific version surfaces real detail.
Name what you observed neutrally ("I noticed you...") rather than what you inferred ("I noticed you were struggling") — let them characterize it.
If they say they're doing something differently because you're watching, ask what they'd normally do instead — that gap is itself a finding.

How to run this interview

The value of a gemba interview is catching what never makes it into a written report — a workaround so routine the agent doesn't think to mention it unprompted. Resist the urge to fill every silence; the pauses where someone is deciding what to do next are often the most informative moments to ask about.

5 · Common mistakes

Common mistakes

Interrupting an action in progress to ask about it.
You lose the moment and the natural flow of the work — wait for a natural pause, or ask about it immediately after rather than mid-action.
Not asking whether behavior is different because you're watching.
People often work differently when observed — asking directly turns that gap into a finding instead of a blind spot in what you recorded.
Describing what you inferred instead of what you actually saw.
"I noticed you were struggling" is your interpretation stated as fact — "I noticed you paused there" is neutral and lets them characterize it themselves.

6 · What it connects to

What it connects to

upstream

downstream

VOC evidence log
Gemba observations are exactly the kind of raw material the evidence log tallies — log them as entries, source type 'gemba.'
WIP aging report
A workaround observed in a gemba session is often the human explanation behind a stalled-item pattern the WIP aging report would otherwise only see as a number.

7 · Where AI helps

Where AI helps

Judgement — stays yours

Deciding when to interrupt versus wait for a natural pause
Judging whether an observed workaround is a real finding or a one-off

Analysis — AI helps

Drafting real-time-observation questions from a stated research goal
Drafting gemba-specific technique reminders

Drudgery — automated

Counting topics and questions
Exporting to xlsx in the house format

9 · Rate this kata