Assessment in Sonar

Good assessment practice

ExplanationFor assessment leads, headteachers, senior leaders, governors and trust staff

These principles come from assessment research and from DfE, Standards and Testing Agency (STA) and Ofsted guidance. Use them when you write your assessment policy or decide what to ask staff to record in Sonar Tracker.

Start with the purpose

Before you collect anything, decide what it is for and who will act on it. Evidence Based Education’s four pillars of great assessment put purpose first, then validity, reliability and value. Value means the time an assessment takes should match what it tells you. ‘Making data work’, the DfE’s 2018 report on data and workload, also asks schools to review each collection regularly.

Keep formative assessment in the classroom

Most formative assessment changes what you teach next and never needs recording. The Commission on Assessment without Levels found no intrinsic value in recording it. The Education Endowment Foundation (EEF) recommends a feedback policy that does not over-specify how often feedback is given or what form it takes.

Use Sonar’s formative pages where a record helps your planning, for example to see which objectives a group has not yet achieved. Track a few objectives that matter, such as phonics, number facts and transcription, not every statement for every pupil. In DfE-funded research by the National Foundation for Educational Research (NFER), some teachers called such recording a tick-box exercise (NFER, 2018). Phonics deserves the most detail: the government expects schools to assess every pupil’s progress in it (government response).

Never work out a summative judgement by counting formative codes. How assessment works in Sonar explains why. Nor should STA’s KS2 writing framework become a checklist for years 3 to 6, because STA says its frameworks are not for tracking progress.

Moderate before you compare

Moderate teacher judgements before you compare classes or schools. ‘Making data work’ warned against using teacher assessment to hold teachers to account without moderation across schools.

Within school, look at pupils’ work together, especially for pupils near the line between two judgements. The Triangulation of Data Analysis report puts each pupil’s judgement beside their other evidence.

Across schools, local authorities moderate KS2 writing in at least 25% of schools each year (KS2 teacher assessment guidance). For writing, comparative judgement is another option.

Check judgements against tests

A standardised test taken by every pupil in the same conditions is an independent check on teacher judgement. Becky Allen notes that several pieces of evidence together are more reliable than any one (Allen, 2018). Test scores entered in Sonar appear beside the judgement in Triangulate.

Every score has a margin of error. NFER’s example is a reading score of 108 with a 90% confidence band from 103 to 112 (NFER). When a test and a judgement disagree, discuss it at moderation rather than assume either is wrong. Do not predict KS2 results from termly tests either: a standardised score of 100 is not the KS2 expected standard (Enter test scores).

Use data to ask questions

Use Sonar to choose which pupils and which gaps to talk about, then answer those questions with pupils’ work and conversation. ‘Making data work’ asked for “objective led discussions not algorithms and graphs”.

In a pupil progress meeting, bring:

the pupils who are not on track, or whose judgement has dropped
what each pupil’s gap is, from their work and formative records
what will change in teaching or support

The Pupil Progress Meeting report sets out a class by judgement for one subject.

Keep tracker data out of teacher appraisal. ‘Making data work’ recommended that performance management should not rest on teacher-generated data, and the government accepted all its recommendations. Once judgements carry consequences, they bunch around the threshold and stop being informative. Rob Coe’s example is the cluster of phonics check marks immediately above the threshold (Coe, 2018).

Report to governors at cohort level

Governors need to know how the school is doing, not every pupil’s judgement. Give them summaries and trends for year groups and the whole school, in your usual format, on dates agreed in advance. Show the number of pupils behind every percentage, and leave out names.

The National Governance Association argues that internal data still matters to boards. ‘Making data work’ advised against routinely showing governors tracking data based on teacher judgement. A cohort summary labelled as teacher judgement meets both concerns.

Statutory results are your most comparable data. The Statutory Dashboard shows them against the previous year, and the Governor Data Meeting option in Reports suggests reports for governors. The Minimal Read-only role lets governors look at Sonar with pupil names anonymised (Users and roles).

Take care with small groups

Percentages swing with every pupil. In a class of 30, one pupil is worth over 3 percentage points; in a group of 4, one pupil is worth 25. ‘Making data work’ estimated that a gap between groups in a one-form-entry primary may need to reach about 40 percentage points to be statistically meaningful.

When you report on groups:

give the count with every percentage, such as 3 of 4 pupils (75%)
compare disadvantaged pupils with non-disadvantaged pupils nationally, as Ofsted does
look at trends over three years rather than one

Do not collect evidence for the EYFS profile

The 2021 EYFS reforms set out to cut unnecessary paperwork and tracking. The EYFS statutory framework says practitioners do not have to collect physical evidence of what children know. The EYFS profile handbook says profile judgements need no written or photographic evidence. It also says teachers should not tick off the bullet points in each early learning goal one by one.

Ofsted will not ask for evidence behind profile judgements, and wants staff interacting with children rather than compiling evidence. Practice has changed: in the DfE’s 2023 evaluation, 67% of reception leaders reported spending less time on assessment.

Use observations to share children’s learning with parents. Record one when a child shows something new and significant, not to meet a quota.

Compare like with like across a trust

In-school summative assessment is not designed for comparing schools unless they share a common system (Commission on Assessment without Levels). Before a multi-academy trust (MAT) compares its schools:

agree the same codes, assessment points and definitions in every school
moderate teacher judgements across schools, and never rank schools on unmoderated judgements
use common standardised tests, taken in the same window, for comparisons that matter
show each school’s context and pupil numbers beside every comparison

Take data from schools’ Sonar records through Sonar Multi-School rather than asking for separate returns. Sonar Multi-School can show each school’s full set of judgements only if every school uses the same summative codes. See About Sonar Multi-School.

What Ofsted inspectors look at

Ofsted’s school inspection toolkit, in use from September 2026, says: “Inspectors will not review internal data.”

Inspectors may ask leaders how they use internal assessment and what they have decided as a result. You do not need to produce documents for inspection or present information in any particular format.

Inspectors start from published results, compared with national figures and similar schools, with care over small cohorts. Where there are none, they rely on pupils’ work, lesson visits, talking to pupils and following individual pupils. They hear lower-attaining pupils read and look at how assessment identifies gaps in reading and phonics. In the early years, they compare the proportion of children reaching a good level of development (GLD) with national figures.

Press reports in October 2026 suggest further changes to the inspection framework. Check Ofsted’s current guidance before you rely on this section.