Nobody schedules an hour to write up a Gemba walk. It just takes one. Here is where it goes, roughly, for a 45-minute walk that produced twelve observations.

The actual breakdown

  • ~15 min — transcribing notes. Turning shorthand scribbles into sentences other people can read. Some notes no longer make sense.
  • ~10 min — chasing photos. Getting images off a phone, cropping them, working out which photo belongs to which observation.
  • ~10 min — formatting. Getting things into the template, fixing the table that broke, resizing images.
  • ~10 min — classifying. Deciding waste type, severity and area for each item, several days of habit removed from the moment of observation.
  • ~10 min — assigning. Working out who owns what, finding email addresses, writing the covering note.
  • ~5 min — actual thinking. Noticing that three of the twelve items are the same underlying problem.

Fifty-five of those sixty minutes are clerical. The five minutes of pattern recognition at the end is the only part that needed an operations leader rather than an intern, and it is the part that gets cut when the day runs short.

The report is not expensive because writing is hard. It is expensive because it reconstructs, at a desk, information that was complete on the floor.

The reconstruction tax

Every one of those clerical steps exists for the same reason: the record was incomplete at the moment of observation. You wrote three words on a pad, took a photo you did not label, and left the classification for later. Everything after that is reconstruction.

Reconstruction is lossy. The photo you took to remind you of something now reminds you of nothing. The note that said "wrap station — walking" made sense at 06:20 and is ambiguous by Friday. You end up writing a version of what you saw that is slightly less true than what you saw.

Fixing it at the source, not the sink

There are two ways to attack the hour. The obvious one is to make the writing faster: better templates, dictation, a tidier export. That gets you maybe twenty minutes back and does nothing about the accuracy loss.

The other is to make the record complete at capture. If, standing at the wrap station, you attached the photo, tagged the waste type, scored impact and frequency, and named the area — then there is nothing to reconstruct. The report becomes a selection problem, not a writing problem: which of these complete records go in, and who owns each one.

The test
A good capture flow is one where walking away from the issue is the last action you take on it. If you are carrying anything in your head to the next stop, the record is not finished.

What still deserves a human

Removing the clerical hour should not mean removing judgement. Three things stay with the walk owner, and no amount of automation should take them:

  • What matters. Twelve observations are not twelve priorities. Somebody has to say which three are worth capacity this month.
  • Who owns it. Assignment is a management act. It carries a promise to the person who raised the problem.
  • What it means together. The five minutes where three separate items turn out to be one layout decision. That is the whole reason to walk.

Cut the fifty-five minutes and the five minutes get better, because you arrive at them with accurate records instead of a reconstruction. Most teams find their reports get shorter and more useful at the same time, which surprises them.