Even good schools are subject to entropy

Why is it that schools drift? What happens to cause so many once successful schools to lose, over time, whatever it was that led to their success? Is there a form of institutional entropy that naturally pulls schools down to a lowest common denominator? Must successful schools constantly expend energy just to resist the gravitational pull of mediocrity?

An easy answer to these questions is place the blame at the feet of school leaders. It can seem pretty obvious that if a good headteacher leaves and is replaced by a less good one then decline will be inevitable. Of course, sometimes that’s exactly what happens. There’s no doubt that principals make an enormous difference and poor appointments can cause considerable damage. But this explanation might be a little too easy, conveniently turning the trajectory of a school into a neat narrative about the success or failure of individuals, heroes and villains, good leaders and bad ones. If standards slip, it’s as easy to blame the incoming head as it is for them to claim credit for anything that starts to improve.

But this neat story is both unfair and insufficient. Incoming heads inherit organisations already travelling down established tracks. They don’t just start driving a train, they also have to contend with the rail network the train’s travelling on. As a new head you have to get your head around existing staffing decisions, budgetary constraints, accumulated exceptions, unresolved problems, established routines, hidden dependencies and layers of institutional memory. The consequences of these decisions — made long before a new head is appointed — may not become visible for years. A headteacher can accelerate decline or arrest it, but, unless they’re launching a brand new school, they never begin with a blank canvas.

Schools aren’t set in amber. Staff leave, routines drift, temporary fixes become permanent and the reasons behind established practices are forgotten. None of these changes matters much in isolation, but together they accumulate. None of these shifts matter individually and can all be justified reasonably but they accumulate. And this accumulation leads, inexorably, to entropy.

Of course, schools aren’t closed thermodynamic systems and a headteacher isn’t one of Maxwell’s demons.1 Thinking in terms of entropy captures schools’ seemingly natural tendency to slide towards disorder unless energy is continuously spent maintaining order.

But deterioration sis hard to see in the present. A school can seem stable even though parts of the puzzle are being slowly taken away and put in the wrong places. If you walk the corridors behaviour seems mostly fine. The school’s still oversubscribed and, crucially, results remain within the acceptable range. From the outside, nothing appears to have changed.

Meanwhile, for those on the inside, getting anything done requires more and more effort. In most schools, competent people compensate. Because the deputy understands the timetable well enough, she can patch it whenever staffing changes. Heads of year increasingly have to intervene personally in incidents that would have been routine. Department leaders spend more time explaining procedures that used to well understood by everyone in the team. But good people can conceal weak systems for only so long.

We tend to imagine stability as the absence of change. If behaviour’s good, then presumably the behaviour system is stable. If the curriculum is coherent, the job’s done. As soon as expectations seem to be embedded, we turn our attention elsewhere. Sadly, however, this isn’t how schools (or any other human organisations) work.

Biological systems offer a more useful analogy. The human body is able to keep its temperature within a narrow range because a vast number of processes are constantly adjusting to changing conditions. Blood vessels dilate and contract, sweat evaporates, muscles shiver and hormones alter metabolic activity. What appears stable is actually homeostasis, the process by which living systems maintain relatively stable internal conditions despite changes in its environment.

For a school’s corridors to remain calm, staff must actively monitor students’ behaviour. Although students get used to corridors being a particular way — what we see around us becomes what we think is normal — as existing staff and students are replaced by incomers, expectations are tested. Stability depends on staff repeatedly noticing, reinforcing and repairing. What we accept becomes acceptable. What we permit we promote.

Culture is perishable. Shared expectations only remain shared if they’re continually communicated and renewed. For schools to achieve homeostasis, a huge amount of regulation and maintenance must go on. Remaining excellent after losing twenty members of staff is a spectacular — but usually uncelebrated — success story

Different elements of what makes schools successful decay at different rates. While knowledge encoded in policies can survive for years,2 tacit knowledge — the assumptions, habits and expectations people acquire simply by spending time in a school — is more fragile.

Success is perishable, but unevenly so. Some things need renewing constantly, others can coast for much longer. Each of the the different ingredients of a successful school (student behaviour, precise focus on effective teaching routines, staff expertise, professional trust etc.) don’t experience entropic forces in the same ways or at the same speed.

Good behaviour can evaporate in a matter of weeks if students get the the impression that expectations aren’t being reinforced but this isn’t how it usually happens. Many staff and some departs continue to operate like benign dictators in a failed state. They uphold orderly disciple with warm ferocity. Some corridors will become chaotic imediately whilst others remain a pocket paradise. As each new cohort tests what will actually be tolerated, standrds decline year on year unless it is actively arrested.

Trust can erode quickly or slowly after repeated inconsistency. One of the keys to maintaining professional trust is under promise and over deliver. It’s not enough to be honest with staff, you also have to do what you’ve committed to. Staff appreciate it when their leaders stand up and admit they’ve made a mistake, but they quickly get intolerant if they keep making them.

Although staff turnover is an obvious source of entropy, we often think about it primarily in terms of vacancies: when one teacher leaves, another needs to be recruited to fill the gap. But teachers aren’t interchangeable units of labour. An experienced head of department knows which bit of the curriculum new teachers usually find difficult to teach, why a unit’s place in the sequence was moved, which routines are essential, which parents need a phone call rather than an email and which colleagues can solve problems on their own and which need some hand holding.

Most of this knowledge is tacit. It’s not that anyone deliberately decide not to write it down, it’s more that to attempt to do so would be incredibly inefficient. I remember a line manager telling me I should record more of what I did because “what if I was knocked over by a bus tomorrow?” I could see her point but the stuff I knew and did had taken years to accumulate through experience. I knew things because I was there when problems emerged, watched others solve similar problems and encountered enough cases to recognise patterns. How would you even start writing all this down?3

When experience staff leave, schools lose compressed experience. A replacement may be excellent — they may even possess greater expertise in some areas — but they don’t automatically inherit the accumulated knowledge of how this particular school works. As with institutional resilience, this loss of tacit knowledge is hard to see. It’s usually not til it’s gone that it’s absence hits us.

Imagine a department of eight teachers. One teacher leaves and another arrives. Nothing may have changed on the staffing spreadsheet but the lived experiece of the team might be very different. Suppose the departing teacher had taught every examination specification used by the department, mentored several colleagues, remembered the reasoning behind the curriculum and was the person everyone asked when an unusual problem arose. Their replacement is in their third year of teaching. If recruitment endangers organisational knowledge maybe we should put more of a premium on retention.

Schools tend to think about succession in hierarchical terms: who might become the next headteacher, deputy or assistant head? But succession risk exists at every level. The most consequential departure may be that of someone whose importance has gone largely unnoticed by senior leaders.

How many schools deliberately survey who possesses vital institutional knowledge? If only one person knows how an important process works, a school doesn’t really possess organisational knowledge. In actual fact, it has a dependency it might not even be aware of.

As I explored in my last post, schools lose the reasoning behind decisions as well as staff who made them

Most decisions are made in response to particular problems. Over time, knowledge of what that problem was erodes despite the ‘solution’ continuing to cast its shadow. As a result, schools often either preserve practices whose purpose has long been forgotten, or remove them without understanding what useful functions they may still perform.

Schools posses huge volumes of information. Policies documents and handbooks tell staff what to do but rarely record why the rule exists. Shared drives fill with documents and lesson slides, while their context disappears. When I visit schools I regulary ask questions about what I notice — Why are you teaching that there? Or, Why is that teacher teaching this so differently to another? — and listen to explanations that are, at best, garbled. I wait as people search through their laptops to try and find the version of a document they want if the certain knowledge that even if they can find it, were they to leave, no one else would even know to look for it.

Institutional memory, the continuing ability to explain why important things are done, is one very good reason why we develop rituals: memory survives socially. Through repetition, we hear messages we’ve heard many times before, ingraining ideas and ways of behaving ever deeper into consciousness. But what’s repetition for some is always new for others. New staff and students have their school taught to them through its rituals. And if this chain is broken meaning drains away.

Once schools start to forget why they do what they do, it becomes increasingly expensive to maintain the status quo, with time and attention devoted to servicing decisions whose original purpose is no longer understood. In software engineering, teams incur what they refer to as technical debt by choosing quick fixes that create extra work later. As shortcuts accumulate, increasing effort is spent maintaining the compromises they’ve created.

Whenever they shortcut the hard work of resisting entropy through workarounds, temporary fixes and sensible exceptions, schools incur a similar debt. Those responsible may fully intend to return when time allows and replace the bodge with something more durable, but we all know what paves the road to hell. People move on, circumstances change and temporary solutions endure. Eventually, all that remains is something clunky enough to just about work, whose upkeep is now someone else’s problem.

Sometimes, though, maintenance debt begins with something much more reasonable: an exception. Every exception to a rule can be justified. Let’s say the rule is that if a student misses a lunchtime detention this automatically escalates to an afterschool detention. What happens if a student has a medical appointment, or can only get home on the school bus? There are always excellent reasons to depart from the accepted process.

But what happens when reasonable exceptions accumulate? Staff become increasingly uncertain about what should happen. Policy and practice drift apart, and teachers start creating their own ways of making the system work. The same process appears elsewhere: a class is exempted from a common assessment becuase their teacher has been off long-term, a staffing workaround caused by an ECT walking out midway through the first term complicates the timetable, or the Head of Year 8 creates a parallel spreadsheet on students causing concern because the MIS is too unwieldy. Each decision makes sense in context, but together they make the organisation harder to predict, understand and maintain.

Exceptions should always be exceptional. As soon as they become routine, they start degrading the system and creating a maintenance debt that’s easy to miss. Staff end up spending time and attention on chasing kids to go to intervention sessions, keeping track of who’s allowed to wear trainers because or their “sore feet” and which parents have been promised either that their children don’t have to complete homework and that tbey should be given more of it.

The dilemma is that if we preserve too we get bogged down by the accumulation of unnecessary complexity. On the other hand, when too much is stripped away, we risk removing something that turns out to be vital. Robust schools can absorb long-term absence, a difficult cohort or an unexpected shock without the wheels coming off, but more fragile schools depend on everything going right all the time. Each new year, term and week is another spin of the roulette wheel. Sooner or later, a school’s run of luck will end and the debt they’ve piled up will need to be paid.

Schools often increase their vulnerability by trimming away whatever looks surplus. Duplication seems inefficient, unused capacity is expensive and so any role that isn’t at capacity is ripe for trimming. Financially led curriculum planning intensifies this pressure because staffing models are built backwards from what schools can afford: contact ratios rise, smaller classes become harder to justify and apparently spare leadership or administrative capacity is squeezed out. On paper, the model becomes leaner but in practice, each saving removes some of the capacity schools rely on when maintainance debts are due.

Organisational researcher, Louis Bourgeois described organisational slack as, “that cushion of actual or potential resources which allows an organization to adapt successfully to internal pressures for adjustment or to external pressures for change in policy.” He continues by adding that slack also allows organisations “to initiate changes in strategy with respect to the external environment.” Although this thinking is already embedded in schools in the form of prudent financial reserves, we’re much less comfortable applying the same logic to staffing, time and attention.

When considering the available slack in your school, what matters is how easy it is to access and redeploy. Spare money, spare time, duplicated expertise and an underloaded member of staff may all count as slack, but they offer very different kinds of protection when circumstances change. You can’t just redeploy a MFL teacher to successfully teach chemistry. Mao and colleagues found that both ‘absorbed slack’ (resources already tied up in existing operations) and ‘unabsorbed slack’ (resources able to be redirected more freely) were associated with greater organisational resilience. But just having spare capacity isn’t enough. Obvously just because you have £40,000 in your staff budget doesn’t mean you’ll be able to hire an experienced teacher halfway through October. Slack only has value if we can be used to learn, adapt and recombine it in response to new problems.

Interestingly, Essuman et al found that slack also improves resilience indirectly, through organisational attention. Spare resources gave organisations the capacity to notice emerging problems, make sense of them and act before they devolved into crises. The more time and attention staff have, the more likely they are to notice unusual patterns in, for instance, attendance data or assessment outcomes. What can look like waste in the short term — spare time, additional staffing or overlapping expertise — may prove costly to eliminate.

For obvious reasons, industries like aviation or nuclear power don’t pursue efficiency at any price. They know that reliability depends partly on the capacity to absorb error and surprise. Research on high-reliability organisations suggests that dependable performance under difficult conditions depends less on assuming that systems will work than on assuming they could fail and actively working to forestall potential failure.

Because their associated risks of failure seem so much less dramatic, schools can easily move in the opposite direction. Roles disappear because another staff member with a couple of frees on their timetable can absorb the work. It’s common for a department to lose its second in charge and for them not to be replaced because the head of department “appears to be coping.” Maybe they were, but is coping the limit of ambitions? And, now their support has been pared down will they continue to cope? In a system obsessed with efficiency, teaching allocations creep ever upwards and administrative tasks are redistributed whenever someone leaves. Each individual decision makes the staffing model look more efficient on paper but every such gain in efficiency reduces the margin for error.

Inefficiency may be the price of reliability. What looks like slack under normal conditions becomes resilience as soon as anything goes wrong.

Schools have another vulnerability: we always want to improve them, no matter how successful they are.. Dylan Wiliam’s famous aphorism captures the impulse neatly: “Every teacher needs to improve, not because they are not good enough, but because they can be even better.”4 On the face of it that’s hard to argue with but Voltaire’s warning that “perfect is the enemy of good” points to the potential risk. As soon as continual improvement becomes an expectation, perfectly sensible ways of doing things can start to look like problems just because they’ve stayed the same for too long.

This impulse creates a bias towards what psychologists call subtraction neglect. When trying to improve something, we’re much more likely to think about what we could add than what could be removed.5 New priorities — iteracy, retrieval practice, attendance, adaptive teaching, oracy, metacognition, belonging, AI — are stacked onto existing ones because they’re obviously important, The problem is that trying to focus on them all will likely lead to most being done badly. Over time, ‘improvement’ becomes synonymous with doing more. Inevitably, attention fragments, embedded procedures are disrupted and maintenance is seen as complacency. Ironically, the drive to improve can make schools worse.

Every school improvement initiative is heralded as the thing that will lead to rocketing outcomes, tranquil classrooms and harmoniously coordinated teaching. Of course, this rarely comes to pass. Whoever launched the initiative leaves to inflict their ideas on another school and another shiny new initiative replaces the last. But old initiatives don’t just disappear. Traces of them clog up shared drives and defunct laminated posters continue cropping up in classrooms for years, a phenomenon Heusinkveld and Benders describe as organisational sedimentation. Sometimes this sedimentation is so entrenched that it leads to schools doing some very strange things. Incredibly, I worked with several schools over the past six months who are still training their heads of department for deep dives that will never happen. Usually the discrepancy between what’s current and what’s defunct is less stark, but staff still have to work out which layers are current and which are the fossilised remains of school improvement plans past.

William Ocasio’s attention-based view of organisations explains why this is can be a serious impediment. As he puts it, “What decision-makers do depends on what issues and answers they focus their attention on.”6 Attention is channelled by the rules, resources, relationships and routines of an organisation. Every new initiative therefore changes not only what staff have to do, but what they’re expected to notice, discuss and act upon.

Every new initiative makes a claim on a school’s collective attention. As initiatives accumulate, attention fragments across more and more demands making it ever harder to get relatively simple things done. The effort required to produce dependable outcomes keeps on rising.

In healthy schools, low effort produces high reliability. Behaviour systems work, assessments generate broadly trustworthy data and new staff quickly understand what’s expected. As entropy increases, the same level of reliability requires more effort.

Hard work, so often seen as a proxy for effectiveness, might actually suggest too much effort is being spent overcoming unnecessary friction created by accumulated complexity.

If your effort-reliability ratio is rising you’re likely to be heading into trouble. As Leaders have to expend more effort on keeping everyday plates spinning, they become less able to see what’s going on and more dependent on proxies to try to make sense of their school.

Some reliance on proxies is unavoidable. We understand, for instance, that attendance figures stand in for patterns of absence, and assessment data only snapshots what students know and can do. Leading a school necessitates compressing vast quantities of messy information into a form that can be quickly interpreted and acted on.

But when leaders become increasingly distant from the thing being measured, proxies increasingly come to replace what they measure. As long as metrics increases twe don’t have to worry about the welfare of individual children. But metrics are only meaningful when they’re interpreted in context. A principal who’s worked in a school for fifteen years will know their staff well, understand why systems were introduced and have a good grasp of the context behind any blips in the data. But a newly appointed head, given exactly the same information, will have far less of the knowledge needed to make sense of it.

The understandable response is to ask for more information. If data suggests a department might be inconsistent, leaders introduce a new tracker. Because the tracker doesn’t give the full picture, subject leaders are required to add written commentary. Because the commentary varies between departments, a standard template is introduced. Completing the template then needs monitoring, which creates deadlines, reminders and further checking. Before long, staff are making vastly greater efforts for increasingly unreliable outcomes.

This negative feedback loop leads to an overvaluing of proxies. Producing proxies consumes attention; reduced attention makes the underlying work of teaching kids harder; and declining reliability creates pressure for still more evidence.

As we’ve seen, schools don’t make decisions in isolation: each choice alters the conditions in which the next is made, a process known as path dependence.

My favourite example of path dependence is the QWERTY keyboard. No one today would design a keyboard by asking what 19th century typewriter used and yet the solution to very specific mechanical problems has completely shaped the way virtual keyboard on smart phones are laid out. Originally, pressing a key swung a metal arm, called a typebar, towards the paper. If two commonly used typebars (T and H, S and T or A and N, for example) were positioned close together and struck in rapid succession, they could collide and jam together. The logic of the QWERTY keyboard was to prevent jams by separating the position of these letters. Athough typists can learn to touchtype at very high speeds, these arrangement has the unintended consequence of slowing down the maximum speed of typing.

However, once enough typists had learned the layout, manufacturers built machines around it, training courses taught it and employers came to expect it. Even after the mechanical constraints disappeared with the invention of electrical typewriters, changing the keyboard became impossible because everyone had accepted QWERTY as inevitable even though there are more efficient ways to design a keyboard.

The paths we’ve walked down constrain the choices available to us. Policies and interventions don’t exist in a vacuum. There is no Platonic Ideal. Their effects depend on the history of the system into which they are introduced, a phenomenon known as hysteresis, the tendency of a system’s past states to continue shaping its present behaviour.

This helps explain why two schools within the same Multi Academy Trust can adopt identical behaviour policies and end up with very different behaviour. In School A, consistency over years has built habits, expectations and trust. Staff expect colleagues to follow through, students expect rules to mean what they say, and leaders don’t need to make much effort for things to tick along.

The leaders of School B, observing the success of the neighbours may decide to adopt their behaviour system but there’s no guarantee they’ll get the same results. Years of inconsistency have taught students that consequences are negotiable. Staff expect to be undermined by leaders making exceptions, and the ghosts of initiatives past haunt the corridors. Past conditions continue to exert an influence even after those conditions have changed. Introducing a better policy won’t erase the habits, expectations and informal routines produced by years of what came before.

The same hysteretic process applies when two departments adopt the same curriculum. Even though materials are the same, the results they produce are very different because one has deep shared expertise while another is rebuilding after half the team left the previous year. Copying the visible structures of successful schools (a practice institutional theorists refer to as mimetic isomorphism) is so often disappointing for precisely this reason. You can copy a curriculum or behaviour policy, but you can’t copy the accumulated history of implementation, expertise and shared expectations that makes those structures work.

Because exam results are lagging indicators, entropic decline is so easy to miss until it’s too late. Students sitting GCSEs today began secondary school years earlier, so current outcomes still contain traces of previous staffing, curriculum choices and leadership. By the time results fall, a school’s deterioration may have been underway for years.

Finland early PISA success created an international fairy tale about the exceptional quality of Finnish schooling, despite ongoing declines in outcomes. Finnish PISA scores have been in free fall since the mid-2000s, with 2022 results lower than in any previous cycle across maths, reading and science. The OECD now describes the decline as one that “began much earlier.” Although Finnish authorities point to changes in funding, inequality, motivation and the wider social environment, the real problem was the at the conditions responsible for Finland’s success had been systematically dismantled and replaced with approaches that made things worse. But, by only looking at the outcomes of children educated under the outgoing system, other countries were desperate to import the disaster policies Finland was currently enacting7

As shared understanding weakens, schools often compensate by writing more things down, while individual staff create workarounds because formal processes are too clunky. At the same time, slack disappears: timetables tighten, spare capacity is allocated and everyone becomes busier.

Although none of these signs proves a school is in decline, they suggest day-today operations may be too dependent on the past while losing the capacity to absorb what the future may present.

The argument so far has been diagnostic. Good schools become ever more fragile when:

• Experienced staff leave faster than they’re replaced

• Exceptions accumulate

• Initiatives pile up

• Slack is trimmed away

So, what should we do about it? Because maintenance is never finished, there’s no magical solution to organisational entropy. It is a depressing fact of life. But by accepting this fact, schools can take the following steps to slow decay, make it easier to spot and reduce the damage when it sets in.