How do you improve a diverse school system?

A regulator overseeing approximately 200 private schools, across over a dozen curricula and every fee tier, cannot pull one lever and expect it to travel. Notes from the design of a UAE regulator's improvement function.

A diverse school system is easy to describe and hard to improve. In the UAE that diversity is unusually sharp. Two hundred private schools inside a single emirate, working in a dozen curricula, teaching in three or four languages of instruction, charging fees that vary by a factor of ten, and serving families from more than a hundred nationalities. When a regulator inside that system commits to raising quality, the first temptation is to reach for the average. Publish a standard, hold every school to it, measure the mean, and declare improvement over time. The problem with that approach is that a rising mean can conceal an unchanged tail. The schools furthest from the standard are the ones the regulator is most accountable for, and those are the schools an average tells you least about.

What follows draws on a multi-year engagement supporting a UAE regulator to build an education improvement function from mandate to operating rhythm. The regulator already ran inspection, and inspection was surfacing problems clearly. What was missing was the capacity to do something about those problems at scale, in a way that respected how different the schools inside the system actually were.

Improvement is not inspection under a different name.

The first substantive decision was whether to run improvement out of the inspection function or as its own team. There is a case for running it out of inspection: shared framework, shared descriptors, existing relationships with schools. In practice this rarely works, because inspection and improvement ask different things of the same person. Inspectors are trained to observe against a framework and judge. Improvement asks you to sit alongside a school leader over a year, understand why the school is where it is, and help the school pull the levers its history and context actually make available. The skill set overlaps, but the posture required is different.

The regulator we worked with separated the improvement function from inspection early, gave it a director-level owner, and staffed it with school leaders rather than seconded inspectors. Three years on, the payoff was visible in how schools engaged. The improvement team was trusted in a way inspectors were not, because the incentives were different. Improvement staff were measured on whether the school improved, not on whether the visit went well. That distinction mattered enormously to how schools engaged.

A rising mean can conceal an unchanged tail. The schools furthest from the standard are the ones the regulator is most accountable for.

The system has to be segmented before it can be served.

A regulator cannot help approximately 200 schools the same way. Diversity across curriculum, fee tier, language of instruction, and current inspection rating means that the levers that work for a Weak school in a low-fee tier are not the levers that work for a Good school pushing for Very Good. Both schools need help. They do not need the same help, and pretending otherwise makes the offer feel disproportionate in both directions at once.

The improvement function segmented the system in four groups. Schools currently rated Weak or Acceptable in the lowest fee tiers, where the binding constraint is usually teacher supply and instructional coherence. Schools rated Acceptable in mid fee tiers, where the constraint is usually middle leadership and consistency across departments. Schools rated Good pushing for Very Good, where the constraint is usually depth of curriculum and quality of assessment. And schools already rated Outstanding, where the risk is losing momentum through leadership transitions. Each segment got a different offer. The Weak-and-low-fee schools received embedded coaching over a year. The Good-pushing-for-Very-Good schools received peer learning networks and access to subject specialists on demand. The Outstanding schools received the least intensive support and the highest expectations for peer contribution.

In the schools, the segmentation registered as fairness. A Weak school did not feel abandoned. An Outstanding school did not feel patronised. Both accepted that the shape of the help was calibrated to where they were.

Build a bench of practitioners.

Regulators default to expert panels. A subject panel for mathematics, one for English, one for Arabic, and so on. Panels produce guidance documents. What improvement work needs is a group of practitioners who can be embedded in schools for extended periods, running coaching cycles or curriculum reviews with named accountability for outcomes. This is a different procurement instrument, a different quality assurance model, and a heavier professional learning obligation. It is also the arrangement that moves classroom practice, which is where inspection ratings ultimately come from.

The improvement function shifted around half of its external budget from panels to bench arrangements within two years. Coaches deployed in cycles of four to six months. Curriculum leads deployed for a full academic year. Data specialists deployed to build dashboards the school then owned. The remaining budget went to convening. Convenings work once a bench exists. Without a bench they become talk shops.

Publish narratives, not just ratings.

A regulator publishing on improvement almost always publishes movement in inspection ratings. Ratings are legible and, on their own, misleading. Movement in ratings is a lagging indicator that tells you very little about which schools moved in which directions and why. The improvement function moved to publishing an annual narrative alongside the ratings. Which schools moved, what they did, what the improvement function contributed, and what remained. Parents read the narratives. Operators read them. Other schools inside the same segment read them, and for the first time understood what a peer school one tier above them had done to get there.

What we saw regulators get wrong most often.

Three failure modes recur. Improving the average without improving the tail. Building an improvement function inside inspection and watching it get absorbed. Publishing policy as a substitute for building the capacity to enact it. Avoiding those three is most of the work. The rest is craft, patience, and running a segmented offer for long enough that the schools furthest behind can catch up.

Verdant advises regulators, ministries, and school systems on improvement design. If this piece is relevant to work you are considering, we are happy to talk.