This article is for plant managers, production managers, continuous improvement leads and maintenance managers who run or sponsor kaizen events and then have to answer a simple question a month later: did the gain actually happen, and on which machines? It explains how to set up a before and after comparison built on machine stop data, so the result of a kaizen week is proven equipment by equipment rather than estimated from a few observations and a good feeling at the report-out.
- A report-out estimate is not yet a proven result.
- Measure stops on each machine two to four weeks before the event.
- Compare the same indicators on the same machines after it.
- Review machine by machine at 30 days, including what did not improve.
A kaizen week is a promise, and someone will ask for the proof
Adam Lawrence of Process Improvement Partners tells the story of his first contracted event in My First Kaizen Event as a Consultant. It was a value stream mapping event at a consumer brands plant with three independent value streams, and he was the only external facilitator. He trained two internal leaders to run two of the streams, took the third himself, and found that while every team was walking the process, he could not see how the other two were progressing. The event succeeded, yet his conclusion is a sober one: “Good intentions don’t guarantee good outcomes.” He also draws a scoping lesson that applies to any improvement week, and afterwards ran these events one value stream at a time.
That lesson about visibility does not stop at the facilitation. The same blind spot appears after the event, when the team has gone back to daily work and nobody can see, machine by machine, whether the new standard is holding. The method belongs to the kaizen practitioner. What the plant can add is measurement: a record of when each machine stopped, for how long and why, captured before the week and kept running after it.
Why a kaizen report-out is not yet a result
Most kaizen events end with a report-out that lists actions, a revised layout, sometimes a new standard work sheet and an estimated gain. The estimate is usually built from a stopwatch study during the week, a few interviews and the team’s judgement. None of that is wrong. The problem is what it is compared with. If nobody recorded how the machines behaved before the event, the “before” is a memory, and the “after” is a short observation taken while everyone was paying attention.
Three things typically blur the result:
- The baseline is an average. A line figure hides the fact that one machine improved, one stayed flat and one got worse because the new sequence moved a bottleneck.
- The observation window is too short. A week of attention changes behaviour. The real test is the third and fourth week, on night shifts and during changeovers.
- Stops are declared, not measured. Short stops of a minute or two rarely make it into a manual log, yet they are often exactly what the event targeted.
Build the before and after on stop data, machine by machine
The method below fits around a standard kaizen calendar. It does not replace the event structure; it gives the team a factual baseline and a factual follow-up.
- Start measuring two to four weeks before the event. Record every stop on each machine in scope, with start time, duration and a reason code. Two weeks is a minimum to see a normal mix of shifts, products and changeovers.
- Agree the indicators at the kick-off, not at the report-out. Pick a small set per machine: number of stops, total stopped time, short stops under a set threshold, changeover duration, and OEE if the team already uses it.
- Use the data inside the week. Bring the stop Pareto for each machine to the gemba walk. It helps the team choose which losses to attack and stops the debate about what “usually” happens.
- Freeze the baseline. Save the pre-event figures per machine with the date range and the product mix. That is the reference the sponsor will compare against.
- Keep measuring for at least four weeks after. Compare the same indicators, on the same machines, with a comparable mix of shifts and products.
- Review at 30 days with the sponsor. Show the result machine by machine, including the ones that did not improve, and decide which actions need a second pass.
What a machine level comparison looks like
The table below is an illustration with fictional figures, for example a three machine cell after a kaizen week focused on short stops and changeovers. It shows why a cell average is not enough.
| Machine | Stops per shift (before / after) | Short stops under 2 min (before / after) | Changeover (before / after) | Reading |
|---|---|---|---|---|
| Machine A | 14 / 8 | 9 / 4 | 42 min / 28 min | Gain confirmed, new standard holding |
| Machine B | 11 / 10 | 6 / 6 | 35 min / 33 min | No real change, actions not applied on night shift |
| Machine C | 9 / 12 | 4 / 7 | 30 min / 30 min | Worse: now starved by the faster upstream machine |
In this example the cell average looks acceptable, yet only one machine delivered the gain. Machine B needs coaching on the night shift, and Machine C needs the flow to be rebalanced. Without per machine stop data, the plant would have declared a success and moved on.
Keep the gain alive after the team disbands
The hardest part of a kaizen event is the second month. The team is back in its usual roles, the facilitator has left and the standard starts to drift. A few habits help:
- Put the per machine stop figures on the daily shop floor meeting board, not only in the monthly report.
- Name an owner per machine for the follow-up actions, with a date.
- Set an alert when a machine’s short stops return to the pre-event level for several days in a row.
- Use the 30 day review to decide the next event’s scope, based on where the losses are now, not where they were.
This is also where scoping comes back. A follow-up that tries to cover ten machines with one spreadsheet will slip. Starting with the machines the event actually changed keeps the review honest and manageable.
Where TEEPTRAK fits
TEEPTRAK measures stops, short stops and OEE automatically and in real time. The hardware is a module wired in the machine’s control cabinet and a sensor at the line side, installed in about one hour per machine, with no connection to the PLC. OEE is available within 48 hours, which is short enough to install the system a few weeks before a kaizen event and have a usable baseline when the team meets. TEEPTRAK is used in more than 700 plants in more than 30 countries, and customers see on average a gain of 10 to 15 OEE points and 30% fewer unplanned stops. It is a measurement tool, not a method: the kaizen practice, the facilitation and the decisions remain with the team and its coach.
Further reading
- My First Kaizen Event as a Consultant (Adam Lawrence, Process Improvement Partners)
- Kaizen and continuous improvement: how to digitalize your CI programme
- Measuring lean project results: set an OEE baseline before you start
FAQ
How long should the baseline period be before a kaizen event? Two weeks is a practical minimum, four is better. The period should cover every shift and the usual product mix, including changeovers, so the comparison after the event is fair.
Which indicators prove a kaizen gain at machine level? Keep it short: number of stops, total stopped time, short stops under a fixed threshold and changeover duration, plus OEE if the plant already tracks it. Compare the same indicators on the same machines before and after.
What if one machine gets worse after the event? It is useful information, not a failure. It often means the bottleneck moved or the new standard was not applied on every shift. The 30 day review is the place to decide a second round of actions for that machine.
0 Comments