By Miss Direction | Policy and Strategy
“Similar schools” sounds like one of those ideas which ought to have existed already.
If you want to understand whether a school is doing well, comparing it with schools serving broadly similar pupils seems rather more useful than comparing it with every school in England and hoping the differences come out in the wash.
The Department for Education and Ofsted have now developed a model which gives mainstream schools a group of 50 statistical neighbours. For secondary schools, the matching takes account of prior attainment, disadvantage, pupil stability, deprivation, SEND, EAL and school size. The comparison has been included in Ofsted’s Inspection Data Summary Reports since September 2026. The methodology is set out here.
It is, on the face of it, a sensible development.
Slightly inconvenient for The Fail Files, really.
But Ofsted has now published some analysis of the measure, and that is where things become more interesting.
So far, so reassuring
Ofsted looked back at schools inspected during 2025/26 and compared the achievement grades they received with what the new similar-schools model now says about their attainment.
At secondary level, 58% of schools graded Strong or Exceptional for achievement were performing above the average for their similar-schools group. Two thirds of schools graded Needs Attention or Urgent Improvement were below it.
For secondary schools with higher proportions of pupils eligible for free school meals, the relationship was stronger again: 68% of those graded Strong or Exceptional were outperforming their statistical neighbours. Ofsted’s analysis is here.
That is encouraging.
There is one fairly important detail, though. Inspectors did not actually have the similar-schools measure during those inspections. It was introduced into the IDSR in September 2026.
The analysis is retrospective.
Ofsted has taken judgements inspectors had already made and compared them with what the new model now says about those schools. Its conclusion is that the findings support the decisions inspectors were already reaching through their consideration of context and other evidence.
They may well do.
What the analysis cannot tell us is how much context affected those decisions.
Schools with stronger attainment are already more likely to receive stronger achievement judgements. Schools performing particularly well against their statistical neighbours are also fairly likely to include schools with stronger attainment.
So some relationship between the two is hardly surprising. It is consistent with inspectors taking context into account. It cannot show us whether contextual information actually caused an inspector to change their view, because the comparison was not available to them at the time.
The more interesting part of Ofsted’s data is therefore not where the two measures agree.
It is where they don’t.
Among secondary schools graded Strong, 8% were performing below their similar-schools average.
Thirty per cent of Expected schools were above it.
Five per cent of those receiving Needs Attention were outperforming their statistical neighbours.
There is no reason to assume those judgements were wrong. Ofsted specifically says the similar-schools measure is not there to determine grades. Inspectors still consider what pupils know, their work over time, what happens in lessons and the wider evidence gathered during inspection. Ofsted’s overview makes that clear.
Which leaves a rather useful question.
What happened in those schools?
If a school was outperforming its statistical neighbours but inspectors still concluded that achievement Needed Attention, what did they find that the comparator could not see?
If another was below its similar-schools average but received Strong, which evidence mattered more?
Those cases could tell us far more about contextual inspection than another chart showing that, broadly speaking, high grades tend to accompany better outcomes.
The disagreements are where the interesting work is.
Similar, within the limits of available mathematics
There is also the small matter of what counts as a similar school.
For secondaries, prior attainment is by far the most influential part of the model, with a weighting of 46%. FSM6 contributes 14%, POLAR4 11%, SEN Support 9% and EHCPs 4%, alongside pupil stability, IDACI, EAL and school size. The technical note gives the full weighting.
There is nothing inherently wrong with that. Any national model has to turn something as complicated as a school community into variables which can actually be measured.
The difficulty is that statistical similarity and lived similarity are not always the same thing.
A useful example appeared in the discussion around Ofsted’s publication. Two schools can have the same average deprivation measure while the populations behind that average look completely different. One might have most pupils clustered fairly close to the mean. Another might have substantial numbers living in very deprived circumstances and another substantial group from considerably more affluent backgrounds.
Same average.
Potentially rather different school.
The technical note acknowledges the wider point. The model cannot explain all the variation in school outcomes, and factors which matter significantly to an individual school may not be sufficiently useful predictors across the whole system to make it into the model.
That is unavoidable.
It does suggest we should be slightly cautious about allowing the words similar schools to acquire more certainty than they deserve.
“Statistically less dissimilar schools” is probably more accurate, although admittedly unlikely to catch on.
There is another choice in the technical note which I find more interesting.
In deciding which variables to include, DfE and Ofsted excluded factors affecting outcomes over which schools have some control. Attendance and behaviour policy are given as examples.
Again, there is logic to this. If the model is supposed to capture the circumstances a school inherits, you do not want to build outcomes the school may itself have influenced into the definition of its context.
The trouble is that schools are not especially good at respecting clean statistical categories.
Attendance can certainly be affected by school practice. It can also be affected by SEND, health, poverty, housing, transport, family circumstances, unmet social-care needs and whether the specialist provision a child needs actually exists.
Often several of those things arrive together.
So deciding that SEND belongs within context while attendance sits outside the model is not purely a mathematical decision. Somewhere underneath it is a judgement about which circumstances a school inherits and which difficulties it can reasonably be expected to overcome.
That may be exactly the right judgement.
But it is still a judgement.
The model has not removed subjectivity from context. Some of the subjectivity happened before the spreadsheet opened.
Then the inspector arrives
Ofsted’s early examples from actual 2026/27 inspections are arguably more useful than the retrospective figures.
In one secondary case study, inspectors looked closely at the experiences and progress of pupils with SEND. They considered whether pupils had the foundational knowledge needed to access the curriculum, checked destinations and gathered first-hand evidence in school. They then used the similar-schools measure to add context to what they had seen. The school had performed significantly better than similar schools over a sustained period and was graded Strong for achievement.
That sounds like the measure being used well.
The data did not hand the school a Strong grade, and context did not become a reason to lower expectations. It helped inspectors judge the significance of what pupils were achieving in the circumstances in which the school was working.
More examples like that would be useful.
Even more useful would be examples where the comparator made inspectors change direction.
Ofsted’s own evaluation suggests there is still some way to go on this. Across schools, 65% agreed that inspectors had sought to understand their context. Among secondary schools, however, only 47% agreed their inspection had been tailored to that context, while 44% disagreed. Ofsted acknowledges that there is more work to do.
That matters.
Because ultimately the problem is not whether an inspection framework contains contextual information. It is whether leaders can have reasonable confidence that different inspection teams, faced with broadly the same evidence, will understand and weigh that context in broadly the same way.
Not identically. Inspection is professional judgement.
But broadly.
The similar-schools measure should help. It gives inspectors a common reference point and may challenge an easy reading of raw attainment data. A school with apparently mediocre results may turn out to be doing unusually well for the pupils it serves. Equally, impressive headline figures may look rather less impressive once genuinely comparable schools enter the picture.
Useful information in both directions.
It just does not settle the judgement.
Ofsted has also announced an independent complaints review pilot for schools inspected during 2026/27. For the first time, the external review will be able to consider not only how a complaint was handled but whether the outcome was fair and reasonable, including concerns about inspection findings. Details of the pilot are here.
That is relevant here because, if professional judgement remains decisive, the system around that judgement matters enormously: training, moderation, quality assurance and a credible route to challenge it.
So perhaps the next useful piece of Ofsted analysis is not another demonstration that stronger grades generally accompany stronger contextualised attainment.
Show us the awkward cases.
Show us the school outperforming its 50 statistical neighbours which nevertheless receives Needs Attention, and explain what the inspectors saw.
Show us the school below its comparator which receives Strong, and explain why.
Show us when contextual data changes the conversation rather than simply agreeing with it.
The similar-schools model is a useful step. Raw attainment has never told the whole story, and giving inspectors a better way to understand the circumstances schools are working in is worthwhile.
But measuring context and applying it consistently are not quite the same job.
At some point the statistics stop, the inspection team looks at everything in front of them, and somebody has to decide what it all means.
Context is finally in the room.
We just don’t yet know how much of a vote it gets.
Miss Direction
Policy and Strategy
Fallowmere Academy for Inclusive Learning
Leave a Reply