Improvement theatre: when the management system exists and nothing changes

In February 2024 an expert panel told the FAA that Boeing had a safety management system that matched the standards on paper, and a workforce that did not understand its own role inside it. Twenty-seven findings, fifty-three recommendations, and not one of them about a missing procedure. This is the most expensive failure mode in improvement work, and almost every organisation has a version of it.

I have been handed a lot of folders. Process documents, RACI charts, policy statements, a governance model with committees drawn in boxes, and a certificate on the wall. Everything the standard asks for, produced to a decent quality, signed by somebody senior.

And then I walk the floor and nobody recognises any of it.

This is not fraud. Almost nobody sets out to build a management system that does not work. It happens because the improvement programme was scoped around producing artefacts, measured on producing artefacts, and closed when the artefacts were produced — and because changing what people actually do is a different and much harder piece of work that nobody budgeted for.

It has a name worth using out loud with a sponsor: improvement theatre. The performance of improvement, staged competently, to an audience that wants to believe it.

The case

In 2024 an Organization Designation Authorization Expert Review Panel, convened under a mandate from the United States Congress, reported to the Federal Aviation Administration on the effectiveness of Boeing’s safety management system. The panel reviewed around 4,000 pages of Boeing documents, ran seven surveys, interviewed more than 250 employees and visited six Boeing locations. Its report ran to roughly fifty pages and contained 27 findings of concern and 53 recommendations.

Read the findings as an improvement practitioner rather than as an aviation person, and something jumps out immediately.

The panel did not find that Boeing lacked a safety management system. Boeing had SMS procedures that aligned with FAA and ICAO standards. The documentation existed and was, on its face, correct.

What the panel found was that the company “failed to ensure all employees understood their role in maintaining safety”; that the system was confusing to employees and not fully implemented as a safety culture; that “the procedures and training are complex and in a constant state of change, creating employee confusion” across work sites and employee groups; and — the finding I would put at the top of any improvement diagnostic — “a lack of awareness of safety-related metrics at all levels of the organization.”

There is more. The panel found that the system “still allows opportunities for retaliation to occur, particularly with regards to salary and furlough ranking”, and that Boeing sometimes failed to make safety reporting non-punitive, despite a published Speak Up policy. It found that human factors integration had deteriorated through reorganisation, decentralisation, downsizing and a headquarters relocation. It noted that some interviewees appeared to treat the panel as an audit to be survived rather than a collaboration, and mentioned a briefing from Boeing legal before their interviews.

Boeing has since made substantial changes, and the panel itself acknowledged steps the company had taken. The point here is not to relitigate Boeing’s safety record, which is far outside my competence and well covered by people whose competence it is.

The point is the shape of the finding. A management system that satisfies the standard. A workforce that cannot operate it. And a set of organisational decisions — reorganisation, downsizing, a pay and furlough ranking system — that quietly worked against the thing the management system was for.

That shape is not unique to aerospace. I have seen it in banks, in telecom operators, in pharmaceutical manufacturing and in IT service organisations, usually about eighteen months after a certification.

The six tells

Improvement theatre has recognisable symptoms. If three or more of these are true, the programme is producing artefacts rather than change.

The deliverables are all documents. Look at the plan. If every milestone is a document, a workshop or a sign-off, nothing in the plan commits anybody to behaving differently. A real improvement plan has at least one milestone that reads like “the night shift stops doing X”.

Nobody below the steering committee can describe the change. Ask three people two levels down what is changing and why. If you get three different answers, or a reference to “the transformation”, the change has not landed. This takes twenty minutes and it is the single most informative thing you can do.

The metrics live in a pack, not in the work. The Boeing finding about awareness of safety metrics at all levels is the generalisable one. A measure that only exists in a monthly pack is a reporting artefact. A measure that is visible where the work happens, that the people doing the work recognise and can influence, is a control.

The reward system was never touched. This is the big one and it is almost always true. The improvement programme redesigned the process; the appraisal, the bonus, the ranking and the utilisation target stayed exactly as they were. People follow the reward system. They always have. If the two disagree, the reward system wins, and your process document becomes a thing people work around politely.

Training was the implementation plan. A course is how you tell people about a change. It is not how you make one. When “roll out training to 400 staff” is the final milestone, the programme has confused informing with implementing.

The programme closed on time. Improvement programmes that genuinely change behaviour almost never close cleanly, because the hard part — the part where the new way survives contact with a bad week — happens after the plan says it is finished. A programme that closed exactly on schedule with all deliverables green is usually a programme that was scoped to be closeable.

Why competent people build it

It is tempting to treat this as a failure of seriousness. It is not, and treating it that way guarantees you will not fix it.

Improvement theatre is what you get when you take a genuinely well-run project discipline and point it at a target that is easy to specify. Documents can be specified, estimated, assigned, tracked and signed off. Behaviour cannot, at least not with the same instruments. So a good project manager, doing a good job, will naturally decompose “improve incident management” into a set of deliverables — because that is what the toolset does — and will deliver them.

The second reason is that certification creates a deadline with a document as its acceptance criterion. If you are going for ISO/IEC 20000 or ISO 9001 or ISO 27001 by a date, the thing being tested at that date is evidence. A well-run certification project produces evidence. Whether the evidence describes the organisation is a separate question, and one the certification timetable does not force anybody to ask.

I say this as someone who holds lead auditor qualifications for several of those standards and has sat on both sides of a certification audit. A standard is a good thing. A certificate obtained by building a parallel paper organisation alongside the real one is worse than no certificate, because it inoculates the board against the suspicion that anything is wrong.

The third reason is sunk cost in reverse. Nobody wants to be the person who says, eight months into a programme, that the documents are done and the behaviour has not moved. The honest report is career-expensive and the reassuring one is free.

What actually changes behaviour

Three things, in my experience, and only three.

Change what gets measured where the work happens. Not in the pack. At the point of work, visible to the people doing it, in a form they can influence that day. If you cannot name the measure a team sees every shift, you have not changed anything they experience.

Change the reward. Including the informal one — who gets praised in the morning meeting, who gets the interesting assignment, whose queue people quietly help with. If the formal reward cannot be moved this year, at least stop the active contradictions. The Boeing panel’s finding about retaliation risk sitting inside salary and furlough ranking is a precise example of a reward system working against a stated value, and it is the kind of thing that is invisible from a steering committee and obvious from the floor.

Make the old way harder than the new way. This is the one improvement programmes almost never do, and it is the most reliable. People take the path of least resistance, which is a feature of competent humans and not a flaw. If the new process is three more clicks than the workaround, the workaround wins forever. Remove the workaround, or make the new path genuinely faster. One deleted spreadsheet does more than a dozen awareness sessions.

Everything else — the documents, the training, the communications plan — is necessary support. None of it is the mechanism.

Scoping a programme so it cannot become theatre

Prevention is cheaper than the diagnostic, and it is mostly a matter of how the programme is written down at the start.

Four changes to a terms of reference, none of which cost anything:

Name the behaviour, not the artefact. Every objective gets a sentence of the form “X will do Y instead of Z”. “Produce an incident management process” becomes “shift leads will classify priority at the point of logging instead of at the morning review”. You can still produce the document; it is now in service of something.

Put the measurement baseline before the design. Measure the current state for a month before anybody designs anything. This is resisted, because it delays the visible work, and it is the single highest-return month in the programme. Without it you cannot prove improvement, you cannot size the problem, and — most usefully — you frequently discover that the thing everyone agreed was broken is not the thing that is broken.

Make the sponsor own one behavioural measure personally. Not the programme’s delivery measure. One operational measure of the changed behaviour, reported to the board under the sponsor’s name, for twelve months after the programme closes. This has a remarkable effect on how the programme is scoped, because a sponsor who will still be answering for the measure next year asks much better questions in month two.

Budget the tail. Reserve a quarter of the effort for the period after the formal close — the clinics, the follow-up, the fixing of the things that only appear under load. If the budget does not have a tail, the programme will end at the handover and the organisation will revert, and everyone will conclude that improvement does not work here.

How to find out whether you have it

A diagnostic you can run yourself in two days, with no consultant and no budget.

Pick one process the improvement programme claims to have changed. Just one.

Day one: read the documented process. Then watch it happen four times, with four different people, at least one on a shift that is not the day shift. Do not ask questions while watching. Write down every point where what happens differs from what is written.

Day two: ask each of those four people three questions. What are you measured on? What happens if you do this the documented way and it takes longer? Who would you tell if you found a better way, and what would happen next?

The gap between day one and day two is your answer. If the documented process and the observed process agree, and the four people give compatible answers to the three questions, the improvement landed. If the documented process is a work of fiction that everybody is too polite to mention, you have theatre — and now you have evidence, which is the only thing that makes the conversation with the sponsor possible.

I have run this in organisations that had just been certified. It is uncomfortable for about an hour and then it becomes the most useful meeting of the quarter, because for the first time the discussion is about a specific observed gap rather than about whether people are committed enough.

A note for Indian organisations

Two local pressures make this pattern more likely here, and both are worth naming without defensiveness.

The first is that a great deal of Indian service delivery is contracted against evidence. A client or a parent organisation asks for proof of a process, and the commercially rational response is to produce proof of a process. Over a few years an organisation can accumulate a complete documented management system that exists primarily as an answer to audit questions. The people inside it are not cynical; the system was simply never designed for them.

The second is scale and attrition. When a function turns over a significant share of its people every year, the documented process is doing more work than it can carry — it is simultaneously the standard, the training material and the institutional memory. Under that load it drifts towards being comprehensive rather than usable, and a process document that nobody can finish reading is a process document nobody follows.

Neither pressure is an excuse and both are manageable. But a programme that ignores them will produce theatre no matter how well it is run, because it is working against the grain of how the organisation actually operates.

The question to ask on Monday

If you are a sponsor, or a director, and you want one question that cuts through an improvement status report, it is this:

What is somebody doing differently today that they were not doing six months ago — and who can I ask about it?

Not what has been delivered. Not what percentage is complete. What is somebody doing differently, and may I go and speak to them.

A programme that is working answers that question immediately and is pleased to be asked. A programme that is producing theatre will offer you a document instead, and the offer itself is the answer.

Sources

• Forbes, Boeing made safety confusing, FAA expert panel finds, 27 February 2024 (panel scope, 27 findings and 53 recommendations, quotations on employee confusion, metrics awareness, retaliation and the ODA unit) — forbes.com
• CNN, Aviation safety panel finds Boeing culture included safety ‘gaps,’ fear of retaliation, 26 February 2024 — cnn.com
• Axios, Review finds “disconnect” between Boeing management and staff on safety, 27 February 2024 — axios.com
• Airport Technology, FAA report finds Boeing safety culture is “inadequate and confusing” — airport-technology.com

The panel’s report concerns Boeing’s safety management system as assessed in 2023–24. Boeing has made significant changes since, and the panel acknowledged steps already taken. The findings are used here as a well-documented illustration of a general pattern in improvement work, not as a current assessment of any company.