How Sisra works

Good practice with school data

ExplanationFor senior leaders and data managers

These principles come from DfE and Ofsted guidance and from research on school data. Use them when you decide what to collect in Sisra Analytics, how often, and who sees the results. Each point gives its source and date.

Start with what each collection is for

Before you add a collection, decide what it is for and who will act on it. The DfE’s 2018 report ‘Making data work’ sets four principles for school data. Its purpose and use should be clear, its precision and limits known, its collection proportionate and its processes reviewed. The government accepted all the report’s recommendations.

If nobody can say what they will do differently because of a collection, stop collecting it.

Collect two or three times a year

‘Making data work’ saw “no reason why a school should have more than two or three attainment data collection points a year”. It found no evidence that six collections a year are worth the teacher time. The Secretary of State’s letter of 17 July 2019 backed two or three a year. It also said the DfE would not use pupil targets or predicted grades to hold schools to account.

Practice has moved that way. Teacher Tapp reported on 27 January 2020 that half-termly or more frequent data drops fell from 58% of teachers in 2018 to 45% in 2020.

In Sisra each collection becomes a report, and you can publish up to two a term. Treat those slots as a ceiling, not a target. Plan each year group’s collections around the decisions they serve.

Decide what each collection holds

Working-at, predicted and target grades answer different questions, so do not mix them in one collection. Give each collection the grade type that matches, name it so staff can tell, and keep to one kind of grade through the year. Data sets, assessment collections and grade types explains where each kind belongs.

‘Making data work’ also warned that asking a teacher for up to 30 data points for 30 pupils “introduces spurious precision”. Ask for grades that teachers can give with confidence. If sub-grades such as 6+ and 6- tell you nothing you would act on, collect whole grades.

Treat predicted grades as estimates

Teachers’ forecasts are often wrong, and mostly too high. Tim Gill and Tom Benton studied 686,596 OCR GCSE entries from June 2014 (Cambridge Assessment, July 2015). Of the forecasts, 43.6% were exactly right and 87.3% were within one grade. Another 42.4% were too high and 14.0% too low. Below grade C, only 20.2% were right and 75.5% were too high.

A level predictions show the same pattern. Lindsey Macmillan and Gill Wyness found that only 16% of students have all three of their best grades predicted correctly (Economics Observatory, 3 September 2020). Three in four, 75%, are over-predicted by at least one grade. ‘Making data work’ added that predicted grades “are rarely connected to processes that help students learn”.

So use predictions to plan support, not to judge teachers, and expect them to be optimistic, especially for lower attainers.

Set targets as ranges

A single target grade suggests more certainty than any forecast has. FFT Education Datalab found that only about 9% of pupils kept to the expected path through KS2, KS3 and KS4 (Treadaway, 5 March 2015). Mike Treadaway recommends target ranges, wider where more is at stake, and no rewards or sanctions tied to expected progress. Stephen Tierney, an executive headteacher, argues for “the expected level of progress as the minimum of a target range” (LeadingLearner, 18 October 2015).

A targets data set holds one grade per qualification, so treat it as the bottom of a range when you talk with pupils and staff. Or hold the range in two data sets, as the demo school does with KS4 Targets and KS4 Targets +1. Flight paths draw a single expected line too, so read them with the same care: the evidence on flight paths explains why.

Read residuals and subject differences with care

Residuals compare subjects within your own school, so they always balance out: if one subject is above its pupils’ average, another must be below. They also reflect how severely each subject is graded. In 2022 results, pupils with similar English and maths grades typically scored about half a grade lower in French, German, Spanish and computer science. They scored half a grade or more higher in art and design (Thomson, FFT Education Datalab, 28 June 2023).

So a languages department with a negative residual every year may be no weaker than any other. Compare a subject’s residual with its own trend, and with its Subject Progress Index (SPI), which compares with other schools. Treat a residual, SPI or Progress 8 score for a class or subject as a reason to look closer and talk to the department. It is not a judgement of its teachers.

Take care with small groups

Percentages swing with every pupil or student in a small group. In a class of 20, one pupil is worth 5 percentage points; in a group of 4, one pupil is worth 25. Give the number with every percentage, such as 3 of 4 pupils (75%), and look at three years rather than one.

Progress figures need the same care. A PolicyBristol briefing found that single-year Progress 8 scores “represent very uncertain predictions as to the future performance of schools”. Its authors recommend multi-year averages (Prior, Jerrim, Thomson and Leckie, November 2021). In Sisra, confidence limits that span 0 mean a score is not significantly different from the national average. Ofsted’s inspection toolkit (September 2026) reminds inspectors that “published data for small cohorts is likely to fluctuate considerably”.

Decide who sees what

Governors and trustees need cohort-level summaries, not every pupil’s grade. ‘Making data work’ advised that governors “should not routinely see data on individual pupils, ‘flight paths’ or other teacher judgement tracking data”. Give governors summaries of validated results, with the number behind each percentage and the school’s own explanation.

Trusts should take data in the form schools already use. ‘Making data work’ said trusts and local authorities should not ask for targets or predictions in a different format, or add collections of their own. The DfE’s 2025 accountability guidance repeats that they should not routinely request such information. Sisra Multi-School builds trust figures from what schools already publish.

Within school, Sisra’s authority groups decide who can see what. The Locked status keeps a report to groups with View Locked Reports while you check it. The DfE’s data protection guidance asks schools to record “who has access to it and how that’s controlled” for each kind of data. That page was updated on 9 July 2026.

The DfE’s secondary accountability guidance (updated 16 October 2025) is direct. “Schools should not try to predict pupil or school level progress scores in advance”, that is, before the provisional data in the autumn. And “Schools should not share individual pupil progress scores with pupils or parents”.

Research backs this. Mark Ledger’s Durham thesis found that leaders with expert knowledge and data “cannot make reliable predictions about schools’ value-added results” (Ledger, 2021). Single-year scores are unstable too, as the PolicyBristol briefing shows.

Sisra can show Progress 8 for a current cohort with an earlier year’s estimates, and Sisra Basic Projections can project headline figures. Treat both as internal and indicative. Say which estimates they use, keep them out of anything for pupils and parents, and use them to choose where to look. Progress 8 for cohorts without KS2 results explains which estimates are in use.

Keep data out of teacher appraisal

‘Making data work’ said that “pay progression should never be dependent on quantitative assessment metrics, such as test outcomes”. It also said teacher judgements should not be used to hold teachers to account. Keep residuals, SPI, Progress 8 forecasts and on track figures out of performance management, and use them to start conversations about pupils’ work.

What Ofsted looks at

Ofsted’s school inspection toolkit (version 2.0, for inspections from 7 September 2026) says: “Inspectors will not review internal data”. Inspectors may discuss with leaders how they use internal assessment data, what it tells them and what they decide as a result. Ofsted does “not need leaders to produce documents specifically for an Ofsted inspection”. Inspectors “do not need information to be presented in any specific format”.

Inspectors start from published data, compared with national averages and similar schools, and they know its limits for small cohorts. So Sisra reports are for your school’s own decisions, which leaders may choose to talk about. There is no need to prepare them for inspection.