Improvement engine: how a plant chooses, closes and keeps its improvements
This pillar asks whether improvement in your plant is a managed system or a collection of good intentions. It looks at how projects are chosen and resourced, what happens to agreed actions, how standards change, how gains are counted and whether they hold a year later.
What the improvement engine covers, and why it decides whether other gains stick
Most plants are not short of improvement ideas. What they lack is a mechanism that picks the right ones, gives them real time, closes them properly and keeps the result in place after the project team has moved on. That mechanism is what this pillar measures.
The other pillars find losses and react to them: performance visibility shows where the time goes, daily management acts within the shift and problem-solving finds the cause of a specific loss. The improvement engine is where larger changes are planned and staffed, and where a fix becomes the normal way of working. When it is weak, the plant solves the same problems every year with different people.
We score five things: how the programme is organised (ci1), what happens to agreed actions (ci2), how work standards are managed (ci3), how gains are measured (ci4) and whether results last (ci5). The last two carry most of the weight in results. A plant can run a busy calendar of projects and see no change in cost per part if gains are counted loosely and slip back within a quarter.
Pull the list of improvement projects closed in the last twelve months. Next to each one, write the measure it was meant to change and the value of that measure today. If you cannot fill both columns for most of the list, you already have a fair idea of where you sit on this pillar.
The five levels as you would see them on the floor
| Level | What you see on the floor | What the numbers look like in meetings |
|---|---|---|
| 1 Reactive | Improvements happen when someone has spare time or after a crisis. Instructions, where they exist, sit in a binder nobody opens. Two operators on the same station do the job two different ways. | Nobody reports improvement results. Actions agreed last month are not mentioned again. |
| 2 Aware | A kaizen event or a project now and then, often triggered by a customer complaint or an audit. Actions live in meeting minutes. Some instructions exist but describe an older version of the process. | Project leaders quote their own savings estimates. The same open actions reappear in the minutes week after week, and nobody can say how many are late. |
| 3 Structured | An annual improvement plan linked to plant targets. One central action list with owners and dates. Standard work instructions for key operations, updated when a problem exposes a gap. | Gains are calculated with one common method. Major projects are followed up after closure. The action list is visible, but the overdue share is not yet measured. |
| 4 Proactive | A portfolio ranked by the value of the loss each project addresses, with people and hours assigned. Standards are under version control, and closing an improvement means updating the standard. Key measures of closed projects stay on the line board. | The weekly review opens with overdue items and the closure rate. Controlling has validated the gains of the main projects. A measure that slips reopens its project, and this happens openly. |
| 5 Excellent | Plant objectives are broken down into line-level projects. Standards are reviewed on a cycle and good practices move between lines and sites. Sustainment is part of layered audits. | Improvement gains reconcile with the budget. Results are checked again at 3 and 12 months. A regression is analysed like any other problem, with its own root cause. |
Place yourself question by question rather than overall. A plant can have a Level 4 action list and a Level 2 approach to sustainment, and the assessment scores the two separately for that reason. The levels page explains the scoring and the weakest-pillar rule.
The five questions, and how to check your own answer this week
ci1. How is continuous improvement organised? We are asking whether the plant chooses its improvement work on purpose and backs it with hours. A kaizen calendar is not a programme if nothing ties it to the plant's targets or to its largest losses.
The usual over-rating is to claim Level 3 or 4 because an annual plan exists, when it is a January slide whose projects have no hours assigned and were half displaced by urgent work by March. The usual under-rating comes from smaller plants with no CI department. A plant manager who keeps a ranked list of five projects and reviews it monthly with the team meets the Level 4 description: the question is about the mechanism, not the size of the department.
Take your top five projects. For each, write the loss it addresses in hours or money, the named owner, and the hours the owner actually spent on it last month. If the hours column is empty or guessed, the portfolio is not resourced, whatever the plan says.
ci2. What happens to actions once they are agreed? We are asking whether an agreed action reliably becomes a finished action. This question often hides behind a tidy spreadsheet.
Plants over-rate in two ways. They claim a central list while parallel lists run beside it (the quality team's, the audit findings, the maintenance backlog). Or they re-date items at every review, so nothing ever shows as overdue.
Count the actions whose due date has been changed more than twice, and the actions open for more than 90 days. Then take the minutes of your last three meetings and check that every action in them appears on the central list. If any of these counts is hard to produce, that is your answer.
ci3. How are work standards managed? We are asking whether the documented method matches what people actually do, and whether improvements change the documents. A standard nobody follows is worse than no standard, because it gives false comfort in audits.
Plants with ISO 9001 document control often score themselves at Level 4 because every instruction has a revision number, yet the operator may not follow the current version at the station. The opposite also happens: a plant with a few clear one-page visual standards that are genuinely followed under-rates itself because it has less paperwork. The model does not reward volume of documents.
Pick three stations. Take the instruction posted at each one, watch a full cycle and mark every difference between the document and the work. Then look at your last five closed improvements and check how many changed a standard. If none did, the improvements are living in people's heads.
ci4. How are the gains from improvement measured? We are asking whether a claimed gain is money or time the plant can see in its results. The project leader's estimate at closure is the starting point, not the answer.
The common over-rating is a shared template called a common method while each project leader picks their own cost per hour. Another is counting freed labour hours as savings when nobody was redeployed and output did not change. Both inflate the total and erode finance's trust in every figure the CI team presents.
Take the largest gain claimed last year and ask controlling to find it in the relevant cost centre's results. Then compare the cost per hour used in three different project files. If controlling cannot point to the gain, or the three rates differ, you are at Level 2 or 3 on this question.
ci5. Do improvement results last? We are asking what happens to the measure after the project closes and attention moves on.
Over-rating usually means 'major projects are followed up' has come to mean a tracker status marked closed. Some plants have results that hold without formal tracking, but the model scores what you can show, so an unchecked result still counts as Level 2 or 3.
Revisit the last five closed projects. Go to the line, read today's value of each key measure and compare it with the value at closure. At the same time, check whether the standard each project introduced is still followed at the station.
What Level-4 plants do differently
Level-4 plants do not run more projects. They run fewer and see them through, and they write their own rules for what finished means. These are the practices we look for.
- One list, one review. Every action from meetings, audits, projects and problem-solving goes on one list with an owner, a due date and its source, reviewed at a fixed weekly slot with overdue items first. A common rule: a due date may be moved once, with a written reason; a second move goes to the next level.
- A portfolio capped at capacity. Projects are ranked by the value of the loss they address, using a cost per hour and a cost of scrap agreed with finance. The plant sets its own cap, for example two projects per project leader, and a new one starts only when another closes or is stopped.
- A written closure definition. An improvement closes only when the loss data shows the change has held over a period agreed at the start, the standard has been updated and people trained on it, and the key measure is on the line board with a threshold that reopens the project.
- Fast standard changes. Each standard has a named owner and approver. A change prompted by an improvement is approved in days, operators are trained before the new version goes live, and the skills matrix is updated at the same time.
- Controlling in the room. Finance attends the monthly portfolio review. Gains are shown as validated or expected, and only validated gains are reported upwards.
- A reopen trigger set in advance. For each closed project the team writes the level at which it reopens, for example the measure moving halfway back to its old value for two consecutive weeks.
- Stopping projects. The portfolio review ends projects that no longer address a top loss.
A 90-day plan to move up one level
This plan assumes you sit at Level 2 or 3 on most of the five questions. Start with your lowest-scoring question: the report from the assessment gives a specific next move for each gap, and it takes priority over the general sequence below.
- Weeks 1–2Take stock. List every improvement project and every open action across all lists, then merge the actions into one list with owner, date and source. For each project, write the loss it addresses and the target it serves. Stop anything with no owner or no link to a target. Revisit the last five closed projects and record today's value of their key measures.
- Weeks 3–6Set the rules. Agree a cost per hour of downtime and a cost of scrap with finance, and rank the projects with them. Fix the weekly action review in the calendar. Write the closure definition on one page. Put the standards for the bottleneck line under version control, each with an owner and an approver.
- Weeks 7–12Run the system. Hold the weekly action review and record the on-time rate from week 7. Hold two monthly portfolio reviews with the leadership team and controlling. Close every finished project against the closure definition, with its key measure and reopen threshold on the line board. In week 12, repeat the self-checks above and retake the assessment.
| Role | Owns during the 90 days |
|---|---|
| Plant manager | Chairs the monthly portfolio review, decides which projects stop and protects the hours of project owners. |
| CI lead, or the production manager where there is no CI role | The single action list, the weekly review, the closure definition and the on-time rate. |
| Finance or controlling | The agreed cost per hour and cost of scrap, and validation of gains on the top projects. |
| Engineering or quality | Version control of standards and fast approval of changes. |
| Team leaders | Key measures of closed projects on the line boards, training operators on changed standards and raising the alarm when a measure slips. |
Evidence a jury looks for in verification
If you apply for verification, you upload artefacts and then walk two jurors through them in a 45-minute video interview. They look for records produced by the routine over time, not documents assembled for the application.
Artefacts that count:
- An export or photo of the central action list showing creation, due and closure dates, including overdue items.
- The on-time closure rate over several months, with a line explaining how it is calculated.
- The ranked project portfolio with loss values and assigned people, or minutes of recent portfolio reviews.
- Two or three closed project files with before-and-after data and, where a financial gain is claimed, the controlling sign-off.
- A work instruction with its revision history, and the improvement that triggered the latest revision.
- A line board photo showing a closed project's key measure and its reopen threshold.
- A sustainment record: a re-measurement at a set interval after closure, ideally including a result that slipped and what was done about it.
What does not count: a CI strategy slide or policy with no records of its use; before-and-after photos without data; savings totals with no method or finance sign-off; action lists without dates, or where every item is shown as closed; standards written in the weeks before submission with no revision history.
A project reopened because its measure slipped is good evidence: it proves the trigger works. Plants sometimes leave these out to look better, which removes their most convincing artefact.
Pitfalls that keep plants stuck on this pillar
- Events mistaken for a programme. A kaizen week produces energy and a long action list. Without an owner for that list afterwards, most of it stays open.
- More projects than people. A portfolio larger than the hours owners really have means every project moves slowly and none finishes.
- Closing by date instead of by data. A project reaches its planned end date and is marked closed whether or not the loss has changed.
- Rolling due dates. Re-dating overdue actions every week keeps the on-time rate high and hides the backlog.
- Paper gains. Freed hours that were not redeployed, or scrap savings priced at a list price nobody pays. Finance stops believing the totals.
- Standards written in the office. Instructions drafted at a desk and never tried at the station are ignored from the first shift.
- Improvement as someone else's job. When a CI department owns every project, line managers treat improvement as a service they receive.
- Sustainment by tick-box. A follow-up marked 'still holding' by someone who never went to the line to read the measure.
Where to go next
Take the assessment to see your index for this pillar alongside the other five, and the specific 90-day move for each gap. The improvement engine rests heavily on people and frontline ownership: standards and small improvements only hold when team leaders and operators treat them as their own, so that is the pillar to read next.
Questions
Do we need a dedicated continuous improvement department to score well?
No. The questions describe a mechanism: a ranked and staffed portfolio, one action list, controlled standards, validated gains and a sustainment check. A small plant can run all of this through the plant manager and the production lead. A large CI team without these routines will score lower than a small plant that has them.
How long should we wait before closing an improvement?
Long enough for the loss data to show a change that holds. For frequent losses such as stops on the bottleneck, two to four weeks of data is usually enough to see a stable shift. For rare events, set a longer period. Decide it at the start of the project, not at the end.
Our finance team will not validate improvement gains. What can we do?
Start by agreeing the rates, not the gains: a cost per hour of downtime and a cost of scrap that finance accepts. Then ask them to validate only the top two or three projects. Once they have seen the method applied consistently, extending validation is a much easier conversation.
Is ISO 9001 document control enough to reach Level 4 on standards?
It covers version control, which is part of Level 4. The question also asks whether improvements become the new standard and whether adherence is checked at the station. Plants that meet these through their quality system score well; plants that only have the revision numbers do not.
Are kaizen events still worth running?
Yes, as a format inside the portfolio. An event aimed at a ranked loss, with an owner for its actions and a closure definition, is a good way to move quickly. An event chosen because it is on the calendar, with no follow-up, is what the model describes at Level 2.