Implementation Guide
Gemba Walks.
A gemba walk is a leader going to the place where work actually happens to understand it through direct observation and questions. It studies the process, never the person, and the walker's job is questions, not answers: solving problems on the spot feels helpful and quietly teaches people that thinking is the boss's job. Go see, ask why, show respect, in that order.
What a gemba walk is
Gemba is the Japanese word for the real place: in manufacturing, the place where value is created. A gemba walk is a leader going there, regularly and on purpose, to understand how work actually happens by watching it and asking questions. Not how the ERP says it happens, not how the work instruction describes it, but what the operator actually does at 9:40 on a Tuesday when the label printer jams.
The definition sounds obvious, which is why the practice is so often confused with things that look identical from a distance. A walk is not an audit: an audit verifies a specific standard against a checklist and produces findings. It is not an inspection tour or management by walking around: presence without a method produces impressions, not understanding. All three involve a leader moving through the plant. They differ in what the leader carries, what comes out, and what the people being visited learn to expect.
| Gemba walk | Layered audit | Inspection tour / MBWA | |
|---|---|---|---|
| Purpose | Understand how work actually happens and where the process fights the people doing it | Verify that a specific standard is followed, and still followable | Be visible, show presence, catch whatever happens to be in view |
| What the leader carries | A theme and three open questions | A short checklist tied to a written standard | Opinions and a bit of spare time |
| Output | Observations, one owned follow-up, feedback to the area leader within a day | Findings, each with a disposition: fix, train, or change the standard | Impressions, and the occasional instruction shouted over a machine |
| Cadence | Scheduled and themed, part of leader standard work | Scheduled by layer, daily to monthly | Irregular, driven by mood and calendar gaps |
The first two columns are both legitimate work. The damage comes from mixing them: a walk that quietly turns into an audit teaches people that open questions are traps, and an audit dressed up as a friendly walk produces findings nobody agreed to be measured against. Decide which one you are doing before you leave the office, and say it out loud when you arrive. This page is about the first column.
Where gemba walks sit in the transformation roadmap
On the TeamGuru deployment roadmap gemba walks appear twice. They start in the Diagnose stage as a core method of the current-state assessment: before anyone picks improvement targets, leaders walk the flow to see how the plant actually runs, alongside value stream mapping, which turns that seeing into a measured picture. Then they never leave. Once daily management is running, walks move into leader standard work as a scheduled, checkable routine, and they stay there for as long as the leader has the job.
- You are here Current-State Assessment
- In parallel Value Stream Mapping
- After Daily Management
- After Leader Standard Work
When walks do harm
Walking is not automatically good. Every walk teaches the organization something about its leaders, and it teaches fast: one careless visit can undo months of honest boards. A badly done walk is worse than no walk when it looks like any of these:
- Verdict tourism. The leader sweeps through, pronounces judgment on what a snapshot seems to show, and leaves. The area learns to manage appearances, and the next walk sees even less of the truth.
- Entourage walks. Six managers, a clipboard each, zero listening. Nobody speaks honestly to a delegation. If more than two or three people are walking, it is an event, not a walk.
- Outsourced to-do lists. The walk ends with a list of actions for the area, written without the area, from an hour of context. The people who live with the process get homework from someone who visited it.
- Walking only when something is wrong. If leadership appears at the line only after a bad month, presence itself becomes a threat signal, and every question lands as an accusation.
The common thread is that all four study the person instead of the process, or hand the area conclusions it had no part in reaching. If a walk cannot yet be done without those habits, it is better to fix the walker's method than to increase the walking.
How to walk
A walk needs a theme, a structure and a discipline about time. The theme narrows what you look for, because looking for everything means seeing nothing. Good themes are one lens wide: one waste from the seven wastes, following inventory or waiting for the whole walk; one standard, comparing the actual work against the standardized work sheet at two or three stations; or one flow, following a single order or part door to door and logging every wait and handoff.
The walk structure card
The card below is the whole method on one page. Print it, fold it, take it with you.
| Phase | What to do |
|---|---|
| Before, 10 minutes | Pick the theme and a route of two to four stops. Read the area's board first: yesterday's numbers, open escalations, running problems, so your questions start where the team already is. Tell the area leader you are coming and what the theme is. |
| During, 30 to 45 minutes | At each stop, watch one full work cycle before saying anything. Then ask three questions from the bank, listen, and write down what you saw and what you heard. Note follow-ups, but assign nothing to anyone on the spot. |
| After, within a day | Choose one action you will own personally, and do it. Give the area leader feedback within a day: what you saw, what you learned, what you will do. No verdicts in passing, no surprise task lists arriving by email. |
The least natural part is standing still. The instinct on a walk is to keep moving, because movement feels productive and standing feels awkward. Taiichi Ohno famously made trainees stand inside a chalk circle on the floor until they could describe what was actually happening around them. You do not need the chalk or the hours, but you do need the principle: a glance sees a snapshot, and snapshots almost always look fine. Only a full cycle, and ideally two or three, shows the variation: the reach that is different every time, the third attempt at a fitting, the walk to fetch something that should be at hand. Given 30 minutes, three stops watched properly beat ten stops passed through.
Asking without judging
The question bank below is organized by intent. Pick three questions per walk, matched to your theme, and let the answers use the time: three questions fully listened to teach more than ten asked while checking a phone. You are not collecting answers to grade. You are learning what the process looks like from inside it.
| Intent | Questions to pick from |
|---|---|
| Understanding the work |
|
| Standards |
|
| Problems |
|
| Improvement |
|
How a question lands is the walker's responsibility, not the listener's. The same words can be curiosity or interrogation depending on tone, and the power gap amplifies everything a manager says at the line. Two rules cover most of it. First, ask about the process, never the person: what makes this step slow, not why are you slow. Second, when someone shows you a problem, thank them, visibly, every time. A problem shown to you is an act of trust, and one bad reaction to bad news can cost you a year of honest answers.
Cadence by level
Cadence follows level: the closer to the floor, the shorter and more frequent the walk. The pattern below fits a mid-size plant; adjust the numbers, keep the ratios.
| Level | Cadence | Length | Typical focus |
|---|---|---|---|
| Shift supervisor | Daily | 10 to 15 minutes | Own area: the board, one standard observed, yesterday's deviations |
| Area or value stream manager | Weekly | 30 to 45 minutes | One theme across the area, rotating: a waste, a standard, a flow |
| Plant manager | Weekly | 45 to 60 minutes | A themed walk of one area with full cycles watched, plus the tier boards |
| Executives and visiting leadership | Monthly | 60 to 90 minutes | A real walk with the plant manager: one flow followed end to end, not a corridor tour |
The way to make any of this survive a bad quarter is to write it into leader standard work: the walk becomes a scheduled routine on the leader's card, completion is visible, and a skipped walk is a miss that gets discussed like any other missed standard. Walks that live only in good intentions die in the first busy week.
Common failure modes
Gemba walk programs rarely fail loudly. They fade, in four predictable ways, and each one has a countermeasure that works better as a standing rule than as a reminder.
Walks get canceled first
Under pressure, the walk is the easiest thing in the calendar to drop, and everyone sees it go. Canceling walks in a crisis teaches the plant that going to see is a fair-weather activity, which is exactly backwards: the crisis is when reality matters most. Protect the slot like a customer visit, and if it truly must move, walk shorter, not never.
Notes that go nowhere
Observations accumulate in a notebook that nobody, including the walker, opens again. The people who answered questions never hear anything back, conclude the walk was theater, and start giving theater answers. The countermeasure is small and strict: one owned action per walk, closed and reported back, and every other follow-up handed to a channel with an owner.
Fixing on the spot
The senior person sees a problem, knows an answer, and gives it. It feels like leadership and it works exactly once. The team learns that thinking is the boss's job, waits for the next visit, and a fix built on ten minutes of context usually creates the next problem. Ask what the team would try, then coach afterward through the area leader. The one exception is safety: an unsafe condition stops the walk and gets fixed now.
The polished route
Same route, same day, same time, every week. Within a month the route is staged: the area ahead of you is being tidied while you park. Vary the route and the hour, follow a part instead of a corridor, and treat a spotless station with suspicion. A gemba that looks perfect is hiding its problems somewhere.
What happens next
A walk produces raw material for three systems, and it is worth being explicit about which observation goes where. What you learned about how the plant actually runs feeds the current-state assessment and the priorities that come out of the Diagnose stage. Improvement ideas raised by operators go into the kaizen pipeline with the originator's name on them, not into your notebook. Deviations from standard become either training or a standard change through standardized work, and recurring problems get a proper 5 whys at the process instead of a corridor diagnosis.
The bookkeeping is where walk programs quietly leak: follow-ups promised at the line and forgotten in the office. This is the part TeamGuru operationalizes: walk findings become owned, dated follow-ups in project and action management, and recurring themed walks run as scheduled checklists and audits, so completion, findings and photos live in one place and the feedback loop back to the area closes visibly instead of on trust.
On the roadmap, what walks surface during Diagnose becomes the baseline for daily management: the boards and tier meetings that turn seeing problems into managing them every day. Once that system runs, the walk changes character, from a diagnostic expedition into standard leadership work, which is where it spends the rest of its life.
