When two assessment results disagree, don't average them or choose the one that feels more convenient. Return to the decision, check whether both results measure the same thing, then compare their provenance, curriculum context, conditions, support, adjustments, missingness and uncertainty.
The assessment lead and teacher should record a provisional interpretation, invite correction and agree what evidence would change it. Disagreement is information about the evidence. It isn't a fixed statement about the pupil.
This interpretation guide is for assessment leads, teachers, SENCos and school leaders in England. It is accurate on 14 August 2026 and doesn't replace the rules for a statutory test, regulated qualification or formal access arrangement.
Begin with the decision the results are meant to support
Two numbers can disagree because they answer different questions. A curriculum-linked check may show whether recently taught material is secure. A standardised measure may sample a broader domain against a reference group. A teacher judgement may draw on performance over time and in several contexts. None becomes the universal truth simply by arriving in a spreadsheet.
Ofqual defines validity through the interpretation and intended use of assessment outcomes. Its reliability guidance also explains that some variability between assessment occasions is inevitable. The first question is therefore: what decision must be made, and which interpretation can this evidence support?
Walk through the disagreement before asking for another test
The details below are invented and combined to show the reasoning. A Year 6 teacher has two mathematics results for one pupil. An end-of-unit fractions task suggests secure understanding. A later standardised mathematics assessment produces a lower result. The planned reporting deadline is close, and the temptation is to declare one result right and the other wrong.
Pass one: state the decision
The immediate decision is whether the next teaching sequence needs more work on fractions. It is not a decision about ability, SEN status, a referral or a permanent group. That boundary matters because the standardised result samples more than fractions, while the classroom task was designed around one recently taught unit.
Pass two: check that the comparison is fair
The assessment lead records the measure, version, date, content sampled, scale and intended use. The teacher adds the task, curriculum point, marking approach and any prompts available. The two results can sit together, but they aren't interchangeable measures of one identical construct.
Ofqual's handbook treats reliability as the consistency expected if an assessment were repeated. Task sampling, marking and occasion can all affect an outcome. A difference should be tested against the technical information supplied with each measure, including any standard error, confidence interval or cautions about comparing forms. Don't invent a threshold when the publisher hasn't supplied one.
Pass three: inspect what happened around each result
Curriculum records show that the pupil had learned the fractions content but missed part of the later ratio sequence. The standardised assessment sampled both areas. The assessment-window record shows the usual low-distraction room was unavailable that morning, and the pupil reports being distracted by several room changes.
Those details don't prove why the lower result occurred. They make the broad claim of a general mathematics decline less secure. The teacher also checks whether either task gave support that changed what was being assessed, whether the responses were independent and whether missing items were treated as wrong answers.
Pass four: keep classroom adjustments and formal arrangements distinct
Reasonable adjustments in teaching and classroom assessment should reduce a relevant barrier so the evidence can represent the intended learning. Formal access arrangements for regulated assessments and qualifications follow their own rules. The current JCQ document applies from 1 September 2025 to 31 August 2026 and requires assessment-specific evidence, including normal way of working where the rule calls for it.
A helpful classroom condition doesn't automatically confer a formal arrangement. Equally, ordinary support shouldn't be withheld until a diagnosis exists. The SENCo and examinations or assessment lead must use the live rule set for the actual assessment, jurisdiction and year.
Pass five: choose the smallest next evidence
The teacher keeps the fractions judgement provisional and avoids creating a fixed group. She plans a short, unfamiliar task that samples fractions and ratio after the missed teaching has been addressed, using the pupil's usual classroom conditions. The class teacher reviews the response with the mathematics lead the following Friday.
The pupil can explain whether the task captured what they knew. A parent, SENCo or other professional can add relevant context or correct the record. The school records what would change the judgement: a repeated pattern across comparable tasks, evidence that the original marking was wrong, or a clear access barrier that made a result uninterpretable.
What two conflicting results cannot prove
Two results cannot by themselves diagnose a need, establish effort, predict a future outcome or justify automatic grouping or referral. A later improvement cannot prove that one lesson, adjustment, intervention or product caused it. Curriculum exposure, attendance, practice effects, maturation, marking, task sampling, conditions and measurement error may all contribute.
At cohort level, show denominators and missing results. Small groups can be unstable and identifiable, so suppress or aggregate where needed and avoid spurious precision. A reporting deadline doesn't make uncertain evidence more certain. It merely gives the uncertainty a diary invitation.
The Standards and Testing Agency's 2026 key stage 2 teacher-assessment guidance is specific to that statutory context, but its evidence principle is instructive: judgements draw on a broad range of evidence over time and across contexts. Use the rule and specification that govern the assessment in front of you.
How Student Radar supports the review
Where the Assess package is in use, Exam & Assessment Windows can keep the window, candidates, arrangements and completion context visible. Assessment Reports can add school, year-group, class and pupil context, including small-group caveats.
These shipped features help staff inspect results beside their context. They don't decide which result is true, infer ability or need, assign a group, calculate pupil risk, or replace the teacher, assessment lead or SENCo responsible for the review.
Use the two-minute decision question before designing the next check, or review the purpose-led assessment calendar before dates harden into habit. To see the connected workflow, request an assessment-focused walkthrough.
Sources and further reading
- Ofqual Handbook: Section D, general requirements for regulated qualifications, Ofqual, published 12 October 2017; updated 4 December 2025.
- Introduction to the concept of reliability, Ofqual, published 16 May 2013.
- Key stage 2 teacher assessment guidance 2026, Standards and Testing Agency, updated 2 March 2026.
- Commission on Assessment Without Levels: final report, Department for Education and Standards and Testing Agency, published 17 September 2015.
- Access Arrangements and Reasonable Adjustments, 2025 to 2026, Joint Council for Qualifications, updated March 2026; valid to 31 August 2026.