When a moderation sample raises a concern: decide what else to review

By the EduMark team

If a moderation sample raises a possible inconsistency, start by checking the original judgement against the assessment evidence and the applicable criteria. Then identify which other decisions could have been affected, select further work to test that concern and agree what must happen before results can be released. A concern is a reason to investigate, not proof that every decision is wrong.

Use your awarding organisation’s requirements and your centre’s procedures to decide who may review, change or release decisions. The example below is an internal, pre-release quality check; it is not a substitute for an awarding organisation’s external moderation process.

Start with the decision, not the assessor

Retrieve the learner’s submitted evidence, the relevant assessment brief and criteria, the assessor’s decision and feedback, and any applicable guidance or standardisation materials. Check that these are the versions that applied when the work was assessed.

Ask an appropriately authorised reviewer, preferably someone not involved in the original decision, to identify the precise point of concern. Give the original assessor an opportunity to explain which evidence supported their judgement. Record the evidence and reasons on both sides rather than treating disagreement alone as an error.

If the criteria or guidance remain unclear, escalate the interpretation through your quality lead or awarding organisation before applying a new interpretation to other work.

A fictional example: a recommendation without comparison

In a fictional assessment, learners must evaluate two options and justify a recommendation using supplied evidence. The centre has 24 provisional decisions: assessor A made 14 and assessor B made 10. During an internal sample, a reviewer questions decision A1, where assessor A recorded the criterion as met.

The submission recommends one option and gives a reason, but does not evaluate the second. Assessor A explains that they regarded a reasoned recommendation as sufficient. The reviewer and quality lead check the assessment guidance and agree that, for this fictional task, the criterion also requires evaluation of both options. They record why A1 needs further action under the centre’s rules.

That finding does not establish that the other 23 decisions are wrong. It does suggest a testable concern: was the same interpretation used elsewhere, and did it affect decisions?

Define the group potentially affected

Before selecting more work, identify the boundaries of the concern:

• Criterion and task: which submissions were judged against the same requirement and guidance?
• People: who made the decisions, and did they share the same standardisation or interpretation?
• Time and versions: when was the guidance used, and did it change?
• Decision types: could the issue affect decisions recorded as met, not met, or both?
• Release position: which decisions are still provisional, and who has authority to pause or release them?

In the example, the quality lead considers all 24 decisions potentially relevant because both assessors used the same task. The immediate concern is stronger for A’s 14 decisions, but checking some of B’s work can help test whether the interpretation was shared.

Pending that check, the lead follows the centre’s release procedure and asks the authorised person to hold A’s provisional decisions. They also check whether B’s decisions must be held under the awarding organisation’s rules or while the wider scope remains uncertain. If a mandatory hold or escalation applies, follow it rather than relying on this example.

Expand the sample for a reason

Choose additional work that can test the suspected issue, not just work likely to confirm it. Include decisions by the original assessor and, where relevant, other assessors; consider work close to the criterion boundary as well as stronger work. Record why each item was selected. The awarding organisation may prescribe the sample, its size or the next stage of review.

For the fictional case, the lead agrees this initial extension:

The sample suggests the issue may recur in A’s decisions, but two checked decisions cannot establish that all of B’s are sound. The lead therefore agrees to review A5–A14 against the criterion individually. They also check whether B’s remaining decisions show a reason for wider review and follow any awarding organisation instruction to extend it further.

Ofqual’s research on CASLO qualifications describes risk-based external quality assurance and pre-certification checks in particular qualification settings. These are examples of different arrangements, not a universal sampling formula for every assessment. ([gov.uk](https://www.gov.uk/government/publications/responding-to-criticisms-of-the-caslo-approach-report-d/responding-to-criticisms-of-the-caslo-approach-report-d))

Record the action and the release decision

Keep a concise review record that another authorised person can follow:

• The task, criterion and versions checked; the original concern and supporting evidence.
• The potentially affected population, the sample selected, and why other work was or was not selected.
• Findings for each reviewed decision, any unresolved interpretation and the person responsible for escalation.
• Agreed corrective action, owner and deadline, including any further standardisation or reassessment permitted by the rules.
• The release hold, who can authorise its removal, and the evidence and approvals needed before release.

In the example, the reviewer completes the check of A5–A14 and records each finding. The quality lead agrees any permitted corrections and confirms whether further work is needed. The authorised person then checks that required actions and approvals are complete before deciding on release. If the wider review remains inconclusive, the team escalates rather than treating the absence of a finding as proof that everything is correct.

For regulated qualifications, distinguish your centre’s internal review from the awarding organisation’s role. Ofqual’s handbook describes awarding organisations’ centre assessment standards scrutiny and defines moderation as taking place before final results are issued. The applicable process depends on the qualification and the awarding organisation’s arrangements. ([gov.uk](https://www.gov.uk/guidance/ofqual-handbook/section-h-from-marking-to-issuing-results?utm_source=openai))

Connect the checks to the assessment journey

EduMark Assessment Journeys connects checks, staff, routing and result release. A team can use its review rules and release checks to support an agreed human-led process. Confirm that the configured journey and permissions reflect the centre’s actual hold and approval requirements; the platform does not decide whether the evidence or sampling is sufficient. ([edumark.ai](https://www.edumark.ai/journeys))