Key takeaways
The standard kaizen event runs for three to five days, with a cross-functional team focused on one process. The team maps current state, identifies waste, designs future state, runs an experiment, and at the end of the week presents the improvement to the plant. Numbers are usually impressive, cycle time down 30%, defects down 50%, changeover time halved.
The numbers are also usually short-lived. Three months later the process has drifted halfway back to baseline. Six months later, in many plants, it is hard to distinguish the post-kaizen process from the pre-kaizen one. This pattern is so consistent that it stops being a quality-of-event problem and becomes a structural problem with how kaizen events plug into the rest of the plant.
At the end of a kaizen the new standard exists. It is in the wrap-up presentation. It is in the team's heads. It is rarely in a written document with a named owner who can be held accountable when the standard is not followed. Without that owner, the standard is unenforceable. The first time someone runs the process the old way, nobody corrects them; the second time, the standard is already half gone.
The fix is simple: every kaizen produces a one-page standard with a single name at the top. The owner is the person who is on the hook when the standard slips. Without the name the standard does not exist.
Most kaizen events measure outcomes, cycle time, defect rate, OEE, and declare success when the outcome moves. Outcomes lag the process by days or weeks. By the time the outcome regresses, the process has already drifted, and the team has already lost the thread.
The fix is to define one leading indicator per change: a count, a frequency, a percentage compliance that moves the same day the process moves. For a changeover-reduction event, it might be "share of changeovers using the new sequence." For a defect-reduction event, "share of operators using the new inspection point." A leading indicator that holds is what protects the lagging gain. The deeper treatment is in the article on manufacturing KPIs.
Many kaizen outputs end up as preventive maintenance tasks, work order types, or quality checks. If they do not get written into the CMMS as standing rules, they exist only in the heads of the people present at the event. Six months later the maintenance team has rotated, the operators have rotated, and the rule is gone. The article on work order management systems covers the rule structures the kaizen output should plug into.
Most kaizen events have a "30-day check-in." Few have a 90-day, 180-day, or 365-day check-in. The decay curve, however, is shallowest in the first 30 days and steepest in the next 90. Without a multi-point review cadence the team only catches drift after it is far advanced. Plants that schedule reviews at 30, 90 and 180 days catch drift while it is still cheap to correct.
A separate structural issue: most kaizen events assume the existing measurement system is reliable. In many plants it is not. Cycle time is captured by one team in one tool; defect rate is logged by another team in another tool; downtime in a third. If the three do not agree on the same shift, the kaizen team is comparing pre-event and post-event numbers that may differ by a factor as large as the improvement itself. The result is a measurable improvement that is unfalsifiable because no one knows what the baseline really was.
The piece on production loss analysis covers the trusted-downtime-register problem in more detail. The short version: run the measurement-system fix before the kaizen, not as part of it.
The fix is procedural, not philosophical. Every kaizen event must produce, before the team leaves the room:
An event that ends without all four is a workshop, not a kaizen. The improvement may still happen; it just will not last.
A plant that runs eight kaizen events a year and ships all four artefacts on every one ends the year with eight standing rules, eight leading indicators, eight owners, and 24 scheduled review checkpoints. The cost of this discipline is roughly half a day of paperwork per event. The return is that the eight improvements are still in place at year-end instead of three of them.
The article on root cause analysis connects to this directly: the leading indicators from past kaizens are the same signals that surface the next round of root causes. The structural discipline is what makes the improvement program compound rather than oscillate.
The four artefacts are tool-agnostic, they can live in a binder. Where a unified OEE + CMMS platform helps is in artefacts 2 and 3: the leading indicator and the CMMS rule. Both need to live in the same system as the OEE event stream so the leading indicator is calculated against current data and the rule fires against current asset state. Fabrico is built so the kaizen output is captured as standing rules rather than as a separate report. To see how a kaizen would close on your live data, book a demo.
For a mid-market plant with 50-100 assets, six to ten focused events per year is a sustainable rate. Above that, the four-artefact discipline tends to slip; below that, the improvement engine cools.
Research on kaizen sustainability has found that a large share of measured gains, on the order of 70%, can decline within a year without standing rules to hold them in place. With the four artefacts, far more of the gain holds, because structured, documented standards retain materially more improvement than verbal handover. The framework does not eliminate decay; it slows it enough that the next event compounds on a real baseline.
The plant manager, with named event owners reporting in. A review cadence owned by no one is a calendar entry that never happens.
That is the most useful outcome the review schedule can produce. A 90-day review showing the leading indicator hit target but OEE did not improve is a clear signal that the wrong target was chosen. The named owner closes that event explicitly and the next event chooses a different target. The discipline is what makes that judgement possible.
Treating the wrap-up presentation as the end. Wrap-up is when the four artefacts are produced. The presentation is one slide of those four, the team has not finished until the standard is written, the indicator is defined, the CMMS rule is in place, and the review schedule is on the calendar.
Вход
Блог
