Student Radar
All articles
Behaviour articlesBehaviour and inclusion leadership

Did the recording system change, or did behaviour change?

Use an annotated change log and matched windows to review school behaviour data, with recording coverage, exposure, pupil voice and human decisions visible.

Founder, Student Radar

Sources last checked .

  • Behaviour data
  • Recording changes
  • Pastoral review
  • Inclusive practice

Before treating higher or lower incident counts as behaviour change, record what changed in the recording system. Check definitions, thresholds, who records, where and how entries arrive, school exposure and the comparison window. A cleaner reporting process can reveal more events without showing that behaviour has worsened. A quieter dashboard can also conceal missing records.

For behaviour, pastoral and inclusion leaders in England, the first task is to establish whether two sets of records describe comparable observation. Keep an annotated change log beside the figures. Then compare matched windows and inspect the evidence with the people involved. The chart doesn't know that Monday's briefing changed the rules.

Blank layered cards, a small wooden gate and a magnifying glass arranged on a cream table
Editorial illustration: recording definitions and thresholds need inspection before totals can be compared.

Write the change beside the count

One classification change is already dated: DfE's 2026 to 2027 school census data-items guidance says the updated suspension and permanent-exclusion reasons list takes effect in the autumn 2026 term and will be collected from the spring 2027 census. Version the applicable list and effective date when comparing those records. Ordinary behaviour logs do not thereby become census returns, and this does not place the new reasons in the autumn 2026 return.

Start one log entry for each change, with its effective date, source, owner and affected records. Preserve the original record and an auditable correction history. The ICO accuracy principle supports checking personal data, correcting inaccuracies and making the status of an opinion clear. These six annotations make the comparison inspectable:

  1. Definition: what counted as an incident before and afterwards? Keep the actual category wording and version. A new category may split one old total into several parts.
  2. Threshold: what prompted an entry? Recording every reminder after previously recording only removal from class changes coverage.
  3. Recorder: who could and did record? Include staff training, cover arrangements and changes in reporting participation. Avoid turning this into a staff league table.
  4. Entry route: where and how did records arrive? Note a new form, mobile access, import, duplicate handling or delay between event and entry. Compare event dates separately from entry dates.
  5. Exposure: which pupils, lessons, locations and hours were observed? Note attendance, timetable changes, trips, school closures and provision outside the usual classroom.
  6. Window: which dates, days and sessions are included? Mark partial weeks and gaps. Keep the extraction date, filter settings and known exclusions.

Agree the purpose before gathering more detail. Under the ICO data minimisation principle, information should be adequate, relevant and limited to what is necessary. A leadership comparison usually needs aggregate coverage notes; personal accounts belong in the appropriate restricted record. Check access, retention and small-group disclosure risks with the school's data protection lead.

Follow one invented change through two windows

The dates, counts and circumstances below are invented and combined to illustrate a review. They describe no pupil, family, school or Student Radar customer.

A year team sees recorded concerns rise from 18 in one ten-day window to 31 in the next. On the first day of the second window, staff began recording repeated verbal reminders as well as removals. A mobile form also became available to cover staff. The raw difference is thirteen records; its meaning remains unresolved.

Change-log annotation: 14 September 2026; category definition broadened; cover-staff entry route added; pastoral lead owns the definition note; data lead checks entry delays and duplicates. Review on 2 October 2026. Earlier reminders weren't collected consistently, so the two headline totals have different coverage.

The data lead first retains both totals with that warning. They compare the removal category only where the definition and entry route remained stable. They select the same year group, weekdays and lesson periods, record ten school days in each window, and check attendance and lessons actually attended. A cancelled afternoon and a new off-site session remain visible as exposure differences. A rate per attended session may help if the denominator is reliable and its limits are stated; it doesn't repair inconsistent recording.

Next, the team reads a small, authorised sample of source accounts. They check whether several entries refer to one event, whether dates reflect late entry and whether the same category means the same thing to different staff. They don't retrospectively invent unrecorded reminders to make the earlier window look complete. Missing records are not absence of events.

Observable signal
More recorded concerns, concentrated in the newly recorded reminder category.
Additional evidence
Definition versions, entry timestamps, attended sessions, timetable exposure, provision actually delivered and accessible pupil views and parent or carer accounts.
Human question
Does the increase persist in comparable records, and what do people report about the routines and support around those sessions?
Proportionate action
Clarify the category with staff, check duplicate handling and review the accessibility and delivery of an existing transition support arrangement.
Owner and review date
The pastoral lead coordinates the review with the SENCo and data lead on 2 October 2026; each records their action and evidence.
Inference limit
The counts establish a recording difference. They cannot establish a cause, diagnose need, predict behaviour or prove whether the support worked.

Offer pupils a way to contribute that suits their communication needs, including time, a trusted adult or an alternative to a written account. Ask parents or carers about relevant changes without expecting them to explain a dashboard. Keep differing accounts and uncertainty visible. The SEND Code of Practice places participation and the graduated approach within SEND practice; a behaviour count alone cannot determine SEN status.

Two equal paper frames containing wooden lesson blocks connected by thread to a circle of empty seats
Editorial illustration: matched observation windows connect to a review with the people who can explain the context.

Keep safety and individual decisions moving

A data-quality review must never delay a response to a concern. If facts may indicate safeguarding risk, pass them to the DSL or deputy without delay under the school's procedures and Keeping children safe in education 2026, in force from 1 September 2026. Immediate danger requires the emergency or local safeguarding route. Keep safeguarding assessment distinct from a disciplinary decision.

Read the evidence alongside Behaviour in schools and the school's policy. Check individual circumstances, SEND support and reasonable adjustments under the Equality Act advice for schools. A change in recorded frequency must not become a label or an automatic sanction, referral, suspension or permanent exclusion. Any suspension or permanent-exclusion decision must follow the applicable statutory guidance effective from 26 July 2026, with the authorised human decision and required process.

Student Radar Behaviour Analytics provides scope and time-window views for inspecting recorded behaviour. Staff still need to check the definitions, source records, coverage and context outside the total. The view cannot decide whether behaviour changed, establish causation or choose an individual response.

At the next review, carry forward the change log, the matched-window limits and the owner's evidence of what happened. Record what can be compared, what remains uncertain and when it will be checked again. Better recording earns a clearer question before it earns a stronger conclusion.

Sources and further reading

Sources checked on 1 October 2026. This practitioner framework combines the sources below; the change log and example are editorial tools, not a statutory form.

Back to all articles