ICEQC-R-2005-05 — Strengthening Institutional Self-Review through Verifiable Improvement Evidence

Self-review is a structured enquiry conducted by or under the authority of the institution whose work is examined. Its defining feature is not that every participant belongs to the institution. Its defining feature is that the institution accepts responsibility for the question, evidence, judgement and response.

The process is educational when it helps the institution understand how provision becomes learner experience and outcome. It is administrative when it records whether a procedure occurred. Both can be legitimate, but they answer different questions. A register showing that programme committees met can confirm governance activity; it cannot establish that a weakness in assessment was identified or corrected.

An **institutional account** establishes context: programmes, learners, staff, resources, responsibilities and material change. An **evaluative finding** compares observed evidence with an explicit criterion. **Improvement evidence** follows a decision through implementation to an operating result. **Assurance evidence** permits scrutiny of the process and conclusion.

Confusion among the four creates characteristic errors. A detailed account is described as evaluation although it contains no judgement. A finding is treated as improvement although no action has operated. A favourable indicator is treated as assurance although its definition, coverage or selection cannot be examined. Separation allows the same document to serve several purposes without allowing one to stand in for another.

Source: ICEQC institutional self-review framework.

Improvement and accountability are not opposing purposes. An institution requires protected space to identify weakness and test change; learners and the public require truthful information about material conditions and unresolved risk. A system that punishes every recorded weakness encourages concealment. A system that keeps every weakness private removes accountability and may leave affected learners without remedy.

The balance depends on consequence. Early exploration of a teaching method can remain within professional enquiry. An unsafe facility, discriminatory barrier, unreliable award decision or prolonged absence of required teaching has a direct public-interest dimension and requires competent action. The review protocol identifies which matters may be addressed internally, which require notification and which require immediate protection.

The Dakar Framework connects quality and measurable learning outcomes with participation, national action, accountability and attention to learners in difficult circumstances. It does not prescribe one institutional self-review model. It establishes the public purpose against which an internal process is judged.[REF-01]

The European school recommendation gives explicit standing to self-evaluation as a means of school learning and improvement, balanced by an outside view, methodological support and participation. The instrument respects national responsibility for education-system organisation and acknowledges differences of context. It therefore supports principles rather than mechanical transfer of one national procedure.[REF-03]

The Bergen Communiqué addresses higher education within a defined regional reform process. Its call for systematic internal arrangements directly connected with external quality review, and for greater student involvement, reflects a contemporaneous expectation that institutional autonomy and public responsibility operate together. The commitments are not attributed beyond participating systems or converted into a global mandate.[REF-04]

An institution is rarely homogeneous. Programmes, sites, grades, languages, modes of study and learner groups can experience different conditions. An institution-wide average can therefore conceal the unit where improvement is required.

The unit of analysis follows the question. It may be a programme, course, class, service, cohort, award decision, facility or the institution as a whole. Findings are aggregated only when definitions and relevant conditions permit. A concern found in one programme does not support an adverse conclusion about all provision; a favourable institutional average does not resolve the programme concern.

Self-review identifies action within and beyond institutional control. Timetable organisation, local academic practice and access to existing resources may be institutionally controlled. Staffing establishment, public finance, national curriculum, major construction or statutory award rules may rest elsewhere.

An institution does not fail merely because it cannot unilaterally resolve an external constraint. It does, however, remain responsible for accurate identification, interim protection within its authority, documented referral and follow-up. The external authority's lack of response remains part of the improvement record.

Themes such as teaching quality, student support, governance or resources are useful for organising a review programme but are not questions. They do not identify the population, expected condition, evidence or decision. They tend to produce broad descriptions and lists of activities.

A reviewable question contains an observable subject and a reason for enquiry. It can be answered within a defined period and leads to a decision. The institution may begin with a broad concern, but it narrows that concern before collecting evidence.

The question statement records: the educational provision or service; intended population; period; criterion; observed concern; decision that the review will inform; and material exclusions. It avoids language that assumes the conclusion.

“Why are students dissatisfied with feedback?” assumes both dissatisfaction and a single causal explanation. A neutral question is: “What proportion of assessed work in the first semester received feedback within the published period, what did the feedback enable students to do, and which conditions explain material variation among programmes?” This formulation permits confirmation, qualification or rejection of the original concern.

Criteria can be binding, authorised or developmental. Binding criteria derive from applicable law, rights, safety duties or formally governing rules. Authorised criteria derive from approved curriculum, programme specifications, published student commitments and adopted institutional policy. Developmental criteria express an improvement objective beyond the current minimum.

The status is stated because consequences differ. Failure against a legal access duty is not managed in the same way as incomplete progress toward an aspirational objective. A developmental ambition cannot lower a binding obligation, and a local policy cannot be treated as law.

Terms such as adequate, timely, effective, appropriate and sufficient require operational interpretation. The review identifies the object, population, period and educational consequence. “Timely feedback” may mean feedback before the next related task, within a published number of days or at another point justified by programme design.

Operational interpretation is not an opportunity to choose a convenient threshold after results are known. The institution records the source and rationale before analysis. Where professional judgement remains necessary, more than one informed reviewer examines boundary cases and records disagreement.

An institution may face tension between curriculum breadth and available time, rapid return of assessment and depth of marking, access and finite specialist capacity, or confidentiality and participation. The self-review does not conceal the conflict in a single score.

The review identifies which expectations are mandatory, which harms are foreseeable, what options exist and who has authority to decide. It states the trade-off and any group bearing the cost. Where no lawful and educationally acceptable option is available within institutional resources, the evidence supports referral rather than a fictitious internal solution.

Source: ICEQC question-design examples. Examples are illustrative and do not establish sector requirements.

Before evidence collection, the institution asks whether the question has a legitimate criterion, identifiable population, feasible evidence, competent decision owner and realistic review period. If any element is absent, the question is revised or referred.

A question is not rejected merely because evidence is imperfect. The institution distinguishes what can be established now, what requires new evidence and what remains beyond reasonable inference. The reviewability test prevents extensive collection for a decision that no one is authorised or prepared to make.

Evidence selection begins with the proposition to be tested. Starting from available records encourages the institution to report what is easy to count rather than what the question requires. A claim-to-evidence map lists each proposition, the most direct source, corroborating source, expected limitation and decision consequence.

For an access question, the proposition may be that every compulsory class was reachable by the learner group. Room inventories are insufficient. The map requires individual timetables, route and facility conditions, actual changes, and safe learner confirmation. Each source contributes a different part of the proposition.

Policies, programme specifications, course outlines, minutes and plans establish intended arrangements and authorised decisions. They have high relevance to what ought to occur, but limited relevance to what occurred.

Version and effective date matter. A policy approved after the reviewed event cannot be used as the criterion unless it restates an earlier applicable requirement. Undated documents are treated cautiously. Minutes record the matters and decisions reported to a meeting; silence in minutes does not prove that a problem did not exist.

Registers, timetables, staffing records, case files, financial transactions and assessment records provide systematic evidence where definitions and custody are stable. They can support full-population analysis and reduce recall error.

Administrative records are created for operational purposes. Their categories may not match the review question; missing entries may mean no event, no record or no requirement to record. The institution documents how duplicates, late entries, withdrawals, transfers and changes of status are treated.

Observation provides direct evidence of a condition or practice during the observed period. It is well suited to accessibility, facility function, teaching process and use of resources. It remains sensitive to timing, observer judgement and changed behaviour under observation.

Work samples can establish the task set, learner response, feedback and progression of work. They do not represent all learning unless selected through an appropriate rule. The sample includes ordinary and incomplete work, not only exemplary products. Names and sensitive information are protected.

Surveys estimate reported experience or opinion among respondents. Their validity depends on question wording, eligible population, response and mode. Satisfaction can identify an issue but is not a direct measure of curriculum, safety or learning.

Interviews permit explanation and exploration of cause. They are not converted into frequencies unless the selection and method support that use. Power relations affect candour: learners may not criticise teaching in the presence of teachers, and staff may not identify governance concerns to a direct manager. Safe arrangements and accurate reporting of non-participation are therefore part of the evidence design.

Assessment evidence requires a clear construct, common administration where comparison is intended, scoring quality and a defined population. Examination pass rates can change because entry rules, absences, assessment difficulty or marking change. Progression and completion are also affected by definitions and the treatment of transfers and repeaters.

International comparative evidence can provide system context but does not assign causal responsibility to one institution. The population sampled, age or grade basis, assessment domain and uncertainty remain essential to interpretation.[REF-09] [REF-10]

Budget allocation, commitment, expenditure and asset receipt are separate stages. Financial evidence can establish lawful and actual use of funds; it cannot alone establish that the educational service operated.

Self-review links expenditure to the specified action, physical or service receipt and intended population. Donated resources, staff time, maintenance and household costs are identified where material. A low recorded expenditure may reflect transferred burden rather than efficiency.

Corroboration is strongest where sources have different error structures. A timetable and teacher interview may both derive from the same planned schedule; an observation adds evidence of delivery. A committee report summarising a student survey is not independent of the survey.

Institutional evidence does not become invalid because staff collected it. Independence is one quality among several. Directness, coverage, timing and transparency may make an internal record stronger than a brief outside observation. Consequence determines whether an additional independent check is necessary.

Source: ICEQC claim-to-evidence map.

Evidence is sufficient when it is adequate for the specified claim and decision, not when every possible source has been collected. The judgement considers directness, coverage, consistency, timing, independence, missingness and consequence of error.

The institution records one of four positions: sufficient to decide; sufficient for a provisional decision with follow-up; insufficient and requiring specified evidence; or incapable of resolution through the proposed method. This prevents both unlimited collection and unsupported closure.

Participation serves two purposes. It provides evidence about educational experience that administrative records cannot show, and it gives affected people a legitimate place in institutional improvement. These purposes overlap but remain distinct. A participant can describe experience without holding decision authority; a governing body can hold authority without possessing direct experience.

The European school recommendation identifies teachers, pupils, management, parents and experts as relevant participants in external and self-evaluation. The right of children to express views in matters affecting them adds a rights basis, with weight appropriate to age and maturity.[REF-03] [REF-06]

Participation begins with the population affected by the review question. Broad invitations commonly reach confident and available respondents while missing those facing language, disability, distance, cost, schedule or power barriers.

The institution maps groups, identifies the consequence of non-participation and chooses feasible routes. It reports invitations, participation and known exclusions. It does not claim representativeness merely because many responses were received.

Learners can provide direct evidence about access, teaching experience, feedback, workload, support and safety. Questions are limited to matters they can reasonably observe. Learners are not asked to determine professional competence, legal compliance or the cause of an outcome.

Participation arrangements protect against retaliation and unnecessary identification. Group discussions are not suitable for every concern. Where a disclosure indicates harm, the institution follows the applicable protection route rather than retaining it solely as research material.

Teachers and other staff hold essential evidence about curriculum, workload, resources, implementation and learner needs. Self-review can support professional learning when staff can identify problems without every statement becoming an individual performance allegation.

This protection is not immunity for misconduct or neglect. It is a separation of purposes. Evidence gathered for institutional improvement enters an individual process only under lawful authority, fair procedure and verified facts. The 1966 Recommendation concerning the Status of Teachers provides an important contemporaneous basis for consultation, professional responsibility and fair treatment.[REF-11]

Families and communities can contribute evidence on access, cost, language, attendance and institutional communication. Their views are not presumed uniform. Participation arrangements recognise that some adults may not speak for all learners and that family preference cannot justify discrimination or an educationally inadequate condition.

External expertise is used where the institution lacks technical competence or where an outside challenge strengthens credibility. The expert's question, evidence access, method and interest are recorded. Reputation alone does not replace a transparent basis for judgement.

Disagreement can concern fact, criterion, interpretation or action. The review separates these forms. A disputed attendance count is resolved through source reconciliation; disagreement about whether the count is acceptable requires a criterion; disagreement about cause requires further analysis; disagreement about action requires a decision under the applicable authority.

Minority evidence is retained where it identifies a group-specific condition or plausible risk. Consensus language is not used to erase a material dissent. The final judgement states the principal alternative explanation and why it was accepted, rejected or left unresolved.

Personal information is collected only where necessary for the question and is accessed according to role. Review records separate identifiers from analytical data where feasible. Publication uses aggregation, suppression and careful narrative description.

Confidentiality has limits where immediate protection or law requires action. Participants receive an accurate explanation of those limits. An institution does not promise secrecy it cannot maintain, and it does not publish detail merely to demonstrate transparency.

Source: ICEQC participation and protection framework.

Participation has a cost in learner time, teacher time, administration and possible exposure. Repeated consultations on the same unresolved issue can reduce trust. The institution reuses valid evidence where definitions and consent permit, coordinates requests and explains what decision followed.

The depth of participation is proportionate to consequence and diversity. A minor timetable adjustment may require evidence from the affected class. A change to programme structure requires broader participation and an account of groups not reached. Proportionality reduces unnecessary burden without allowing leaders to consult only convenient voices.

An institution possesses detailed knowledge, continuing access and the ability to connect records across time. These are substantial advantages. The same institution also has interests in how findings affect reputation, finance, staff and learner confidence. A credible method retains the advantages while controlling the risk that evidence is selected or interpreted to protect the institution.

Integrity does not depend on a claim of neutrality. It depends on procedures that make the judgement open to correction: fixed questions, documented selection, preservation of exceptions, declared interests, separation of roles where feasible, and a recorded response to challenge.

Exemplary lessons, successful graduates and completed actions can illustrate practice but cannot estimate its prevalence. The review population includes ordinary, weak, missing and incomplete cases according to a rule established before results are known.

Purposive selection remains legitimate for some enquiries. A review of a rare access failure properly selects the affected cases. Its conclusion concerns those cases and the mechanism, not the institution-wide rate. The selection method and intended inference are stated together.

Evidence from a single week can be affected by examinations, holidays, induction, staff absence or preparation for review. A period is selected because it represents the proposition or because its exceptional status is the subject of enquiry. It is not selected because records are most complete or results most favourable.

When the timing cannot be changed, the institution limits the conclusion. An observation during one intensive teaching week can establish practice during that week; it cannot establish routine use across the year.

Missing evidence is classified as no event, no record, item not applicable, refusal, loss, inaccessible record or unknown. These states are not combined. A blank assessment-return date may mean that work was not returned, that the date was not recorded or that the learner did not submit. Each interpretation leads to a different finding.

The review reports the number eligible, the number observed and the reason for material loss. A favourable rate among complete records is not applied to missing cases without a stated and defensible assumption. Where the result is sensitive, plausible bounds are presented.

Review teams can favour evidence that supports an initial concern or a preferred action. The protocol requires at least one credible alternative explanation and identifies the evidence that would distinguish it. A falling completion rate, for example, may reflect weaker teaching, changing entry, altered award rules, more complete recording of withdrawals or an external shock.

The purpose is not to list every imaginable cause. It is to test explanations that are plausible, material and associated with different decisions. Where the evidence cannot distinguish them, the finding remains unresolved and the next enquiry is specified.

Leadership judgement has evidentiary value but can dominate an internal process. Junior staff, learners or distributed sites may withhold contrary information. Meeting agreement is therefore not treated as confirmation that no contrary evidence exists.

The institution uses routes appropriate to its context: independent facilitation, anonymous written evidence, separate discussions, rotation of review leadership or direct access to records. The safeguards are evaluated by whether material dissent became visible, not by whether participants reported satisfaction with the meeting.

An institution may improve the indicator without improving the educational condition. A response-time target can encourage premature closure; an attendance target can change recording; a pass-rate target can alter assessment or entry. The review identifies how the measure could be influenced and uses a balancing measure where risk is material.

For feedback timeliness, a balancing measure may examine usefulness and learner opportunity to apply feedback. For complaint closure, it may examine recurrence and unresolved remedy. For completion, it may examine progression rules, withdrawals and learning evidence.

The person responsible for an activity can supply essential evidence and explanation but need not make the sole final judgement. Where capacity permits, the review separates evidence custody, analysis, judgement and approval. Where it does not, another informed person examines selection and reasoning.

Role separation is proportional. A low-risk departmental enquiry may use peer challenge. A finding affecting learner rights, award integrity, safety or substantial public resources requires stronger independence and competent authority.

Source: ICEQC internal enquiry integrity register.

The institution provides a route for factual correction before a finding is finalised. Corrections identify the original record, new evidence, authorising person and effect on the judgement. An inconvenient correction is not rejected because the report is advanced.

After publication, material errors are corrected openly and promptly. The correction states what changed and whether decisions require reconsideration. The general principles of professional method, accountability for sources, correction and confidentiality in official statistics provide a relevant integrity model, while institutional self-review remains distinct from official statistics.[REF-05]

Evidence records have an identified owner, date, version and retention period. Working copies are distinguished from authoritative records. Changes preserve a trace sufficient to explain the final figure or judgement.

Custody protects both integrity and confidentiality. Broad access is not synonymous with transparency. Reviewers receive the least personal detail needed, while the method and aggregate basis remain available for scrutiny.

The review first states what the evidence shows. It then compares that finding with the criterion. A rating, where used, follows these steps and does not replace them.

The finding identifies population, period, measure, variation, missingness and material exception. “Feedback is effective” is not a finding. “In the sample of 120 first-semester assignments, 83 were returned before the next related task; 19 were returned later, 8 had no recorded return date and 10 were not submitted” is a finding capable of examination.

A judgement statement contains four elements: the criterion; the evidenced condition; the degree and distribution of difference; and the consequence for action. It separates verified fact from interpretation.

Where the criterion is met for most but not all learners, the judgement does not collapse the result into a majority pass. It identifies whether the exceptions are random, concentrated or connected with a protected or disadvantaged group. A small number can be material where the right or safety consequence is serious.

Confidence concerns the evidence supporting the judgement, not the favourability of the condition. A strong adverse finding can have high confidence; a favourable finding can have low confidence because coverage is poor.

The review uses plain positions: well supported; supported with material limitation; provisional; or unresolved. These positions are accompanied by the reason. A numerical confidence score is avoided because its precision and weighting would be difficult to justify across dissimilar questions.

Materiality considers educational consequence, legal or protective significance, number and identity of learners, duration, recurrence and effect on awards or progression. It is not determined only by institutional size.

One incorrect award decision can be material to the learner and to award integrity. A minor delay affecting many records may be operationally significant without undermining educational validity. The judgement explains which dimension makes the finding material.

Institution-wide averages are disaggregated where the question and lawful records permit. Relevant dimensions may include programme, level, site, mode, sex, disability, language, age or other national categories. The institution does not collect sensitive characteristics without purpose and safeguards.

Differences are interpreted with counts and denominators. A rate based on five cases is not compared rhetorically with a rate based on five hundred. Statistical uncertainty is considered where sampling is used; verified individual harm remains actionable even when a group rate is unstable.

Context explains conditions that affect the interpretation or feasibility of action. It does not erase the criterion. Rapid enrolment growth, remote location or constrained finance may explain a staffing or facility gap; the gap and affected learners remain visible.

The judgement distinguishes institutional control, shared control and external control. It identifies action within authority and the evidence required for referral. This avoids both unfair attribution and institutional avoidance.

Self-review often needs a causal explanation to choose action. It distinguishes immediate mechanism from wider contributing conditions. A delay in feedback may arise immediately from the sequence and volume of assessment; staff capacity, course design and approval rules may contribute.

Cause is not assigned from sequence alone. The review tests whether the proposed mechanism was present, whether it preceded the condition, whether the pattern differs where the mechanism differs and whether plausible alternatives remain. The resulting wording reflects the strength of that test.

An institution may use categories to summarise findings, but the categories have definitions, evidence thresholds and exception rules. A colour or number cannot be the only recorded result. It is linked to the full judgement and action.

Aggregation across domains is approached cautiously. Safety, access, curriculum, teaching and outcomes are not fully substitutable. A high score in one cannot compensate for a serious failure in another. Where an overall category is required, non-compensable conditions and uncertainty are explicit.

Source: ICEQC evaluative judgement record.

Panel members examine the evidence independently before collective discussion where feasible. This reduces early conformity. The panel considers the proposed judgement, strongest contrary evidence, material missingness and effect of changing a key assumption.

The final record gives reasons. Where members cannot agree on a material question, the dissent and its evidence are preserved. A majority decision can authorise action, but it does not transform unresolved fact into consensus.

The unit whose work was reviewed receives the draft factual record and can provide correction, contextual evidence and an alternative interpretation. The response has a fixed period and cannot remove a supported adverse finding by negotiation.

The final report records material changes arising from response. It distinguishes correction of fact from acceptance of an explanation or proposed action. This makes the review fair without making the subject of review the sole judge of the evidence.

An evaluative finding creates a decision requirement, not an automatic intervention. The institution first determines whether the condition requires immediate protection, correction, further enquiry, broader system action or no change. The decision is proportionate to materiality and confidence.

A weakly supported low-risk concern may justify a small test. A high-confidence safety or award-integrity finding requires immediate competent action. Uncertainty is not treated uniformly.

The institution states why the selected action is expected to change the finding. The proposition links action, mechanism, operating condition and intended result. Assumptions and external dependencies appear beside it.

For delayed feedback, reducing the number of simultaneous assessment deadlines may redistribute marking time and permit return before the next task. The proposition depends on actual workload, assessment sequence, staff capacity and no compensating increase elsewhere. A generic workshop does not address the mechanism unless evidence identifies knowledge or practice as the constraint.

The review considers more than one feasible response where the decision is material. Options are compared on expected educational effect, time, cost, reach, recurrent burden, institutional capability, risk and reversibility.

The lowest immediate cost is not always the most efficient response. A temporary arrangement can protect current learners but create recurring cost; a durable redesign can require more initial effort. The decision records why the selected option is preferable and why plausible alternatives were not selected.

Every action has one accountable owner and named contributing roles. Accountability is assigned to the role with authority to make the required decision, not to the person most available to complete the form.

Where the institution depends on a ministry, governing body, employer, landlord or public service, the plan separates internal action, formal request and follow-up. It does not record the external decision as an institutional task already completed.

The plan records finance, staff time, facilities, technical support, information and authority. It also identifies which existing activity will be reduced or delayed. An action is not costless because it uses current staff.

Teacher and learner time receive particular protection. Meetings, surveys and development activity can displace instruction or study. The plan schedules them so that the improvement process does not create a larger educational loss.

The plan distinguishes authorised, initiated, completed and operating states. A revised policy is authorised when validly approved, initiated when relevant people and systems begin to use it, completed when specified transition tasks finish, and operating when the intended service functions.

Stage definitions prevent a meeting, purchase or publication from being reported as the final result. They also locate delay. An action may be approved on time but fail at staffing, communication, supply or actual use.

Reach identifies which intended units or people received the action. An institution-wide announcement is not evidence that learners requiring a service received it. Training offered is not training completed; training completed is not supported practice.

The plan specifies the intended population and monitors differential reach. A convenient pilot can exclude the programme where the original concern was greatest. If that is necessary for safety or feasibility, the limitation and later stage are explicit.

Implementation commonly requires adjustment. The record distinguishes minor operational adaptation from a change to the core mechanism or population. Minor adaptation can preserve the proposition; material change requires a revised proposition and interpretation.

The institution does not report fidelity to a plan that it did not implement. Nor does it treat every adjustment as failure. The question is whether the critical components operated and whether the changed action still addresses the finding.

The plan considers safety, rights, privacy, academic or curricular integrity, staff conditions, financial control and burden transferred to learners or families. A favourable intended result cannot justify an unlawful or harmful means.

Risk controls are specific. “Monitor closely” is not a safeguard unless responsibility, evidence, timing and response threshold are defined. Temporary measures have expiry and review conditions so that they do not become an unexamined permanent standard.

Source: ICEQC improvement action charter.

Before implementation, the institution states what result would justify continuation, adaptation, wider use or cessation. Rules include both service effect and safeguards. A result is not accepted if it is achieved through exclusion, weakened assessment or an unrecorded transfer of burden.

Decision rules can include a quantitative threshold where justified, but they also address distribution and critical exceptions. A target of 90 per cent timely feedback does not permit the remaining 10 per cent to be consistently concentrated in one programme or learner group.

The register links every material finding to an action, referral, accepted risk or reasoned decision that no action is required. It records status and elapsed time. Findings do not disappear when a report cycle ends.

Repeated actions for the same unresolved cause trigger escalation. The register distinguishes recurrence from a new case and permits leaders to see where local plans repeatedly fail because an institutional or system condition remains unchanged.

Follow-up begins with the original criterion, population and evidence. It does not replace them with the activity most readily reported. If the finding concerned learner access, follow-up measures access; expenditure and construction are supporting evidence.

Where the original evidence was weak, the institution improves the measurement but records the resulting comparability limit. A more complete follow-up can reveal previously unrecorded cases and should not be presented automatically as deterioration.

Verification confirms whether critical action components occurred, for whom, when and under what adaptation. It uses the action charter rather than a general progress narrative.

Completion evidence is proportionate to consequence. A low-risk scheduling change may be confirmed through timetable and sampled delivery. Structural safety, assessment validity or substantial expenditure requires competent evidence and stronger separation of roles.

An operating result is the first verified educational or institutional condition expected after implementation: a service accessible, a required component delivered, feedback returned in time, a case decided through the proper route or a teaching practice used under the specified conditions.

The result is defined closely enough to the action to support a management decision. It is distinct from longer outcomes. Restored access may later contribute to participation and learning; those effects require additional evidence and time.

Baseline and follow-up use the same definition, population and period, or the difference is adjusted and disclosed. Programme restructuring, changed assessment, new record systems, altered enrolment and staff turnover can create an apparent change.

Where a common subset is used, the review states that it represents continuing cases rather than the whole current institution. Where no valid comparison exists, the report presents current operating status and implementation, not a fabricated trend.

The institution examines whether the operating change was followed by the intended educational outcome. A before-and-after change is descriptive. Stronger attribution requires evidence that the action preceded the change, the mechanism operated, competing explanations were examined and an appropriate comparison exists where feasible.

Urgent or universal service action may not permit an untreated comparison. This does not prevent action. It limits the causal claim. The institution can report that the condition improved following implementation and explain why contribution is plausible without claiming exclusive cause.

Follow-up asks what else changed. Faster feedback may reduce depth; expanded access may increase staff workload beyond safe or sustainable levels; a new complaint route may expose confidential information; a revised timetable may disadvantage part-time learners.

Unintended effect evidence uses groups and measures relevant to the risk. Absence of complaints is not proof of absence where the action itself affects the ability to complain.

An operating result can depend on temporary finance, exceptional staff effort or a single individual. Sustainability review identifies recurrent cost, role ownership, maintenance, staff capability and the conditions needed after the initial period.

The review does not require certainty about indefinite continuation. It distinguishes stable operation, time-limited operation with an authorised continuation plan, and an at-risk result without secured means.

Continuation is justified where the action operated, the intended result is evident and risks remain acceptable. Adaptation is justified where the mechanism is plausible but reach, fidelity or context limited the result. Expansion requires evidence that capacity and relevant conditions exist in additional settings. Cessation is justified where the mechanism fails, harm is material or resources can be better directed.

An inconclusive result does not default to continuation. The institution records the value and burden of further evidence, the consequence of delay and any interim measure.

Source: ICEQC improvement verification framework.

A finding closes when the required operating condition is verified, the case is invalid or duplicated with reason, or a competent authority formally accepts a residual condition under a lawful and evidence-based decision. Referral, approval, expenditure or a stated intention does not close the finding.

Closure records the evidence date and any continuing monitoring. If the condition recurs, the institution reopens or links the case so that repeated temporary correction is not mistaken for durable improvement.

Individual reviews are periodically analysed for recurring causes, delayed decisions, failed actions and uneven service. This cross-case analysis can identify a governance, resource or information problem not visible in one review.

Aggregation does not erase open cases. A system lesson and an individual remedy proceed together. The institution records which changes have been made to review design itself, including removal of unused indicators, strengthened participation or revised decision authority.

An outside view tests whether the institution's reasoning can be understood and whether its evidence supports the judgement. It can identify omitted populations, inconsistent criteria, favourable selection and conclusions that exceed the method. It can also confirm that an institution has reported a serious condition accurately and taken proportionate action.

The 2001 European recommendation describes external evaluation as a source of methodological support and outside perspective for continuous improvement, rather than a process confined to administrative checking. This distinction protects both institutional learning and public accountability.[REF-03]

The outside reviewer examines the review question, criterion, population, source selection, missingness, analysis, judgement, improvement response and verification. The task is not necessarily to repeat the whole review. It identifies high-consequence or uncertain links and tests them directly.

The reviewer does not replace the institution's question with an unrelated preference after evidence collection. If a material omitted issue emerges, it is recorded as a separate finding requiring its own criterion and evidence.

Access is sufficient to test material claims while respecting confidentiality. The institution provides source registers, selection rules and de-identified records where possible. Personal files are not disclosed merely for convenience.

Where law or protection limits access, the reviewer records what could and could not be examined and how this affects confidence. A confidentiality restriction is not described as full verification.

A visit can establish current conditions and allow direct discussion, but it remains a limited period. The institution does not stage exceptional activity as routine evidence. The reviewer samples against the published question and follows material exceptions.

Interviews test understanding, participation and implementation. Agreement among selected interviewees does not establish prevalence. The visit record distinguishes direct observation, participant report and reviewer interpretation.

External challenge is recorded as a question or proposition with an evidence basis. The institution has an opportunity to correct fact, provide further evidence and explain context. The reviewer then states whether the challenge is resolved, modifies the judgement or remains as a difference.

The process is not a negotiation over the favourability of language. Changes follow evidence or corrected interpretation. Material unresolved disagreement remains visible to the competent decision-maker.

External scrutiny can rely on institutional evidence where source custody, selection, definition and limitations are sound. Re-collecting every item can waste resources and weaken institutional responsibility. Reliance is greater where the institution has retained contrary cases, corrected errors and verified earlier actions.

Reliance is reduced where the review population cannot be reconstructed, material records are missing, sources derive from one interested account, or repeated internal findings closed without operating evidence. The response is targeted verification, not automatic rejection of all internal work.

The depth of external review follows consequence, novelty, evidentiary weakness and institutional history. A routine low-risk improvement can be tested through records and samples. Safety, learner rights, award integrity, substantial public expenditure or a contested outcome requires more direct and technically competent scrutiny.

Proportionality also limits burden. Institutions do not repeatedly supply the same verified evidence to several processes without a clear purpose. An external request identifies the decision it supports and the period for which existing evidence remains valid.

An external reviewer reports within the evidence examined. A brief review of one programme cannot support an institution-wide conclusion without a valid sampling basis. Absence of a discovered problem does not prove that no problem exists.

The final record distinguishes confirmation of the self-review method, confirmation of a particular finding and direct external finding. These have different scopes and are not merged into a general endorsement.

Source: ICEQC external perspective record.

The external report informs the competent institutional or public authority. It does not absolve that authority from making and recording the decision. Findings are prioritised by educational consequence and evidentiary confidence, not rhetorical strength.

The institution incorporates accepted corrections and actions into its improvement register. Rejected recommendations include reasons. Silence or general acknowledgement is not an institutional response.

Public reporting enables learners, families, staff, governing bodies and public authorities to understand what the institution examined, what it found and what remains to be done. Its purpose is not institutional promotion or publication of every working record.

The account is truthful, proportionate and connected to remedy. Material risks and unresolved failures are not concealed within an average or extensive descriptive text. Confidentiality is maintained without using privacy as a reason to omit aggregate conditions.

The public report distinguishes recorded fact, institutional calculation, evaluative judgement and future commitment. “Ninety of 110 sampled items met the published return period” is a recorded result. “Timeliness is inconsistent across programmes” is a judgement. “The assessment schedule will be revised by September” is a commitment.

These statements occupy separate fields or sentences. A commitment is not written in a form that implies completion, and an interpretation is not presented as directly observed fact.

For each material review, the public account states the question, criterion, population and period; principal evidence; finding and distribution; limitation; decision and responsible authority; and current implementation or verification status.

Technical detail is proportionate. A short method note can support a simple full-population count. A sampled outcome comparison requires the sampling, response, uncertainty and assessment basis. Unsupported precision is not a sign of rigour.

An institution that reports weakness credibly does not thereby establish general failure. It demonstrates that the condition was detected. The relevant accountability questions are whether the finding is accurate, learners are protected, responsibility is assigned and improvement is verified.

Conversely, the absence of adverse findings is not evidence of high quality when the review questions, selection or reporting rules make weakness unlikely to appear. The public report includes material missingness and open findings.

Institutions differ in programme, population, mission, resources and context. Public comparison requires common definitions and an appropriate basis. Raw outcomes can penalise institutions serving learners with greater prior disadvantage or encourage selection.

Comparison can still be useful when it identifies an enquiry rather than a verdict. The report presents counts, definitions, reference periods and uncertainty. It does not rank institutions on an internally constructed measure that was not designed for comparison.

Small groups may experience material exclusion while their public statistics are unstable or identifying. The authority acts on verified need but limits disclosure. Cells are combined or suppressed only where necessary, and totals are designed so that subtraction does not recover protected values.

Narrative examples are also reviewed for identification. Removing a name may be insufficient where programme, event and personal circumstance point to one individual.

The public account identifies the review period and date. Material correction states the original error, corrected result and effect on judgement or action. The earlier record is not silently replaced where it has informed a public decision.

Updates distinguish improved status from changed definition. An institution cannot show progress by removing difficult cases from the denominator or changing a threshold without restating the earlier result under the new basis where feasible.

Actions are labelled authorised, initiated, completed or operating. Delayed and at-risk actions remain visible with reason and next decision date. A referral to another authority is reported as referral, not remedy.

Where no valid action is available within current resources, the institution states the unresolved condition, interim protection and requested decision. Restraint is preferable to a nominal action that cannot change the service.

Source: ICEQC public self-review account.

An annual synthesis groups reviews by educational function, reports open and closed material findings, identifies recurrence and states changes made to the self-review method. It does not reproduce every local record or convert diverse findings into one quality score.

The synthesis includes the areas not reviewed and the basis for selection. This prevents an annual report from implying comprehensive coverage where attention was limited to a small number of favourable topics.

The common disciplines are question, criterion, evidence, participation, judgement, action and verification. Their application differs because authority, learner age, programme structure and educational purpose differ.

The report does not assume that a school self-review model can be transferred unchanged to a university, or that an academic programme review can govern a primary classroom. Sector application begins by identifying the competent body and the educational service.

School review commonly examines curriculum delivery, instructional time, teaching conditions, learner attendance, inclusion, safety, support and learning evidence. Children require age-appropriate participation and protection. Families and public authorities may hold responsibilities not present in the same form in adult education.

The school is often dependent on district or national systems for staffing, curriculum, finance and facilities. Self-review therefore maps responsibility and keeps external constraints visible. The European recommendation's balanced treatment of self-evaluation, outside perspective and participation provides a relevant regional example, not a universal procedure.[REF-03]

Vocational provision adds workplace learning, occupational equipment, employer participation and transition to work or further learning. Review questions identify whether practical experience matches the approved programme and whether learners can complete it safely and equitably.

Employment outcomes can inform review but are affected by labour-market conditions, region, occupation and learner characteristics. An institution does not infer teaching quality from placement rates alone. Employer opinion likewise remains one source rather than the criterion for all educational purpose.

Higher education self-review commonly operates through programme, department and institutional governance. Academic judgement, research, curriculum, assessment, student support and public responsibility interact. The institution must preserve the distinction between academic peer judgement and administrative evidence while requiring each to show its basis.

The Bergen Communiqué calls for systematic internal mechanisms connected to external review and emphasises institutional participation, student involvement and public responsibility. Application remains governed by national arrangements and institutional authority.[REF-04]

One institutional policy can operate differently across sites. Self-review samples or covers sites according to risk, size and difference, and does not infer consistent operation from the central office.

Remote sites may have weaker records and higher observation cost. This does not justify their systematic exclusion. The method uses proportionate local verification and reports access limitations.

Small institutions face difficulty separating reviewer and activity-owner roles, and public reporting can identify individuals through small numbers. Peer challenge, governing-body review or a competent external person can strengthen independence.

The method remains proportionate. A small institution need not reproduce the committee structure of a large one. It must still maintain traceable evidence, fair participation, reasoned judgement and verified action.

Large institutions risk fragmentation and aggregation. Local units may use different definitions, while central reports obscure variation. A controlled core of definitions and review records is therefore necessary, with room for questions specific to programmes.

Central synthesis tests comparability before aggregation. It identifies units not reporting and unresolved cross-institutional causes. Local responsibility does not permit the governing institution to disregard repeated system patterns.

Expansion changes learner composition, staffing, facilities, programme load and administrative capacity. Historical rates may not describe the current population. Self-review uses cohort and capacity evidence and records structural breaks.

The Dakar and Millennium commitments give expansion a strong public purpose, but access without adequate educational conditions is insufficient. Review connects enrolment growth with instruction, support, completion and distribution.[REF-01] [REF-12]

Where records, staffing or safe access are limited, the institution prioritises essential service, protection and a small number of decision-relevant measures. It does not claim comprehensive review.

Evidence can be provisional while action remains necessary. The record states uncertainty, interim measure and follow-up. A reduced method is credible when its scope is explicit; it is not credible when missing evidence is converted into a favourable institutional account.

Source: ICEQC sector application framework.

Institutions can learn from one another about evidence custody, participation, action tracking and verification. Transfer begins with the function of the method, not its label. A useful instrument is adapted to the receiving sector's authority, learners and educational purpose.

Claims that a practice is “good” require evidence of the condition it improved, the setting and resources. Dissemination includes limitations and failed applications, not only a favourable example.

The governing body or equivalent authority approves the purpose, receives material findings, assures that competent decisions are made and monitors unresolved risk. It does not rewrite evidence to protect reputation or intervene in individual academic judgement without authority.

Governance receives a concise account of method, material findings, open actions, overdue cases and limitations. It asks whether the institution can detect and correct weakness, not whether every indicator is favourable.

Leadership allocates time, access, analytical capability and protection for candour. It ensures that findings reach the role able to act and that cross-unit problems do not remain within isolated plans.

Leaders disclose their interest where they own the activity reviewed. Their response is evidence and decision input, not the sole conclusion. Retaliation against a good-faith participant undermines the reliability of future reviews and requires direct response.

Review leaders require competence in question design, evidence selection, basic statistical interpretation, facilitation, confidentiality and improvement planning. Subject expertise is added according to the question.

Training is demonstrated through the quality of review work, not attendance alone. Institutions examine inter-reviewer agreement, correction patterns, missed risks and whether findings led to verifiable decisions.

The schedule is risk- and purpose-based. It includes recurring core matters, emerging concerns, follow-up and a balanced coverage of educational functions. It is not limited to areas likely to report success.

Frequency follows the rate at which evidence and risk can change. Safety or access issues require immediate review; curriculum and outcome questions may follow programme cycles; stable low-risk processes may be sampled less frequently.

Self-review uses valid records already needed for teaching, support and governance. It adds evidence only where the decision requires it. This reduces parallel reporting and improves the incentive to maintain operational records.

Integration does not mean that administrative records determine every question. Learner experience, observation and work samples may remain necessary. The institution identifies where ordinary systems do not observe the educational condition.

Review requires protected staff and learner time, access to records, analytical support and authority to report findings. An unfunded expectation can produce superficial reports and divert teaching time.

Independence is protected through terms of reference, access, recorded challenge and direct reporting of material concerns. It is not established merely by placing review in a separate office if leadership can suppress evidence without record.

The institution examines the performance of self-review itself: questions completed; population coverage; corrections; material findings; time to decision; actions operating; recurrence; participant burden; and unresolved external dependencies.

A high number of reports is not necessarily favourable. A low number of weaknesses can reflect effective provision or weak detection. Interpretation combines quantity with evidence quality, scope and subsequent results.

Intervention is required where self-review repeatedly omits the same disadvantaged group, closes findings without operating evidence, changes criteria after results, suppresses material disagreement, or produces plans unrelated to findings.

Other warning signs include extensive description with no evaluative judgement, identical findings across dissimilar units, unexplained absence of missing data and a high rate of administrative completion with recurring service failure.

Source: ICEQC institutional self-review capability framework.

Confidence in institutional self-review accumulates when the institution identifies material weaknesses, preserves contrary evidence, corrects error, acts within authority, refers what it cannot control and verifies results. It is reduced by claims of comprehensiveness unsupported by coverage or by favourable conclusions unsupported by direct evidence.

No single review proves sustained capability. Consistency across questions and time matters. Evidence of learning includes changed instruments, improved definitions, shorter unresolved delays and abandonment of actions shown not to work.

Public authorities can strengthen self-review by clarifying purpose, reducing duplicate reporting, protecting participation, providing methodological support and ensuring that institutions have a response route for conditions beyond their authority. They can also require public reporting proportionate to risk and safeguard personal information.

Authorities should avoid incentives that equate the absence of weakness with success. A mature system expects institutions to find and correct problems. External scrutiny is directed to the quality of that process and the material educational result.

Institutional self-review is credible when it produces a bounded judgement that another informed reader can trace, an improvement decision connected to the finding and later evidence that the intended condition changed. Its authority comes from method, candour and action, not from institutional assertion.

The institution remains the principal actor because educational improvement depends on people who organise and deliver the work. Public responsibility requires that ownership be accompanied by participation, outside challenge and visible treatment of material risk. The balance is not achieved through a uniform form. It is achieved through clear questions, legitimate criteria, verifiable evidence and decisions that remain open until the educational service operates.

An institution establishing or revising self-review first identifies material open questions and applicable criteria. It creates a claim-to-evidence map, protects participation, records judgement and connects each material finding to an accountable decision.

It then tests a limited number of reviews through follow-up before expanding the schedule. The test of readiness is not whether templates are complete. It is whether the institution can detect an unfavourable condition, report it accurately, take proportionate action and show whether the condition improved.

The first cycle should also establish a dated baseline for unresolved findings, assign authority for decisions that cross departmental boundaries and identify which evidence can be published without exposing learners or staff. These arrangements make later comparison possible and prevent the review calendar from displacing urgent correction. Where evidence is incomplete, the institution records the limitation, the means of obtaining stronger evidence and the date on which the judgement will be reconsidered.

The initiating concern is recorded in the words in which it first reached the institution, together with source and date. It may arise from routine evidence, learner or staff report, a governing question, external observation, changed provision or follow-up to an earlier finding. Recording the original concern permits later examination of whether the review merely confirmed its starting assumption.

The initiating record does not become a finding. Where immediate safety, rights or award integrity is at risk, protection and competent notification proceed before the full review. The need to act does not remove the need to establish facts; it changes their order.

The review leader restates the concern as a neutral, answerable question. The question does not assume failure, cause or preferred remedy. It identifies the service, population, period and decision.

For example, “Poor teaching is causing students to leave the evening programme” contains two untested propositions. The neutral review asks: “How do continuation and withdrawal differ between evening and daytime entrants over the last three cohorts, what reasons are recorded or reported, and which institutionally controlled conditions are associated with the difference?” The first stage is descriptive; causal examination follows only where evidence permits.

The protocol defines who or what is eligible. Learners may be counted by person, enrolment, course registration or assessment attempt. Staff may be counted by person, post or teaching assignment. Facilities may be counted by room or scheduled use. The selected unit follows the proposition.

Entry, exit, transfer, deferral and repeat status are specified. A review of first-year continuation cannot silently exclude learners who withdrew before the census date if early withdrawal is part of the concern. Duplicate exposure is identified where one learner contributes several course records.

The period covers the operation relevant to the question. It records calendar dates, academic term or cohort as applicable. Where a seasonal or cyclical effect is plausible, the design includes the relevant cycle or limits its conclusion.

The evidence cut-off for the review is stated. Events after that point may inform follow-up but do not alter the original finding without a clearly dated update.

Each criterion entry records the exact provision or intended condition, source, adoption or effective date, authority, population, operational interpretation and consequence of non-fulfilment. The entry distinguishes binding, authorised and developmental criteria.

Where a source uses qualitative language, the operational meaning is approved before outcome analysis. The institution retains the source text and identifies any local interpretation. It does not present the local interpretation as if it were the wording of the external instrument.

Source: ICEQC criterion register.

Exclusions identify matters deliberately outside scope and why. An assessment review may exclude the validity of the curriculum if that requires different expertise, but it cannot imply that curriculum validity was confirmed. Exclusions are examined for whether they remove the group or condition most likely to show weakness.

An exclusion caused by lack of access is reported as a limitation rather than a design choice. The review records who can authorise access and whether further enquiry is required.

The protocol names the role that will receive the judgement and the decisions within its authority. It also identifies required referral. A review without a decision owner can generate evidence without remedy.

Where the owner is responsible for the activity reviewed, an additional challenge or approval route is assigned according to consequence. The role arrangement is disclosed in the review record.

For each proposition, the plan identifies direct source, supporting source, selection, collection period, responsible person, confidentiality requirement and expected limitation. It includes evidence capable of contradicting the initiating concern.

The plan is revised only for a recorded reason. If an unavailable source changes the inferential strength, the judgement is narrowed. The institution does not substitute a convenient survey for direct evidence and retain the original claim.

The plan states the counts, rates, comparisons or qualitative analysis to be used. Denominators and missing categories are specified. Where judgement coding is required, the protocol defines categories and comparison of reviewers.

Exploratory analysis can identify an unanticipated pattern. It is labelled exploratory, and any new criterion or threshold is not represented as pre-specified. A material new question receives its own review record.

The question protocol is approved by a role with appropriate authority before substantial evidence analysis. Changes record date, reason, approving role and effect on interpretation. Minor wording clarification is distinguished from a change to population, criterion or decision.

The final report includes the operative question and discloses material changes. This prevents a review from moving toward an easier question after encountering adverse evidence.

The evidence register identifies every source used for a material finding. It enables another informed reviewer to locate the source, understand its purpose and determine which analytical result depends on it. It does not contain personal detail unnecessary for control.

Each source receives a stable identifier, title, owner, creation purpose, reference period, coverage, version, custody location, access restriction and linked proposition. Derived files identify the source records and transformation.

Sources are classified as governing document, operational record, observation, participant evidence, work sample, assessment, financial record, contextual statistic or technical opinion. Classification assists interpretation but does not assign quality by itself.

An operational record can be comprehensive and weakly defined. A small observation can be direct and unrepresentative. The register separately assesses relevance, directness, coverage, timing and independence.

Where sampling or full-population analysis is intended, the institution constructs a list of eligible units. The frame date, source and known omissions are recorded. Duplicate programmes, inactive learners, closed sites and ambiguous status are resolved or retained as explicit categories.

Comparison between frame totals and authoritative institutional totals can reveal omissions. Agreement does not prove completeness if both derive from the same deficient source. Relevant local registers or service lists provide an additional check.

A census includes every eligible unit and is suitable where each case requires decision or the population is manageable. A probability sample supports estimation where every unit has a known selection chance. A purposive sample examines particular conditions or variation but does not estimate prevalence without additional basis.

Selection can combine methods. All high-risk cases may be included while routine cases are sampled. The resulting analysis reports the high-risk cases separately or applies appropriate design treatment. It does not present the combined unweighted sample as representative of the institution.

Sample size depends on population, variation, precision, subgroup needs and clustering. The protocol avoids a universal minimum. A small qualitative enquiry can be sufficient to identify a mechanism; it cannot establish an institutional rate. A rare high-consequence failure may require complete case review even where prevalence is low.

The institution does not use a conventional number to create an appearance of statistical adequacy. It explains the precision or inferential limitation that follows from the actual design.

Every selected unit has a disposition: observed, ineligible, duplicate, refused, unavailable, lost, not reached or pending. Item non-response is separate from whole-case non-response.

Follow-up effort is consistent enough to avoid concentrating completion among convenient respondents. Where replacement is permitted, the rule is fixed before collection and the original non-response remains visible. Replacing a remote or dissatisfied participant with an accessible one changes the evidence and cannot erase the first case.

Records are reconciled using identifiers, dates and event sequence. Totals are checked across sources that serve different purposes: enrolment and finance, timetable and attendance, assessment and award, referral and service.

Differences are not averaged away. The institution determines whether they reflect timing, definition, duplication, delayed entry or error. Unresolved differences remain in the limitation and can themselves identify a control weakness.

Source: ICEQC evidence register.

Qualitative material is coded against the review question, with allowance for unanticipated material themes. The code definition includes inclusion, exclusion and boundary examples. More than one reviewer examines a subset where judgement is material.

Frequency of a theme is reported only when selection and coding support a count. A vivid account is not described as common without evidence. A rare account can nevertheless require action where it identifies verified harm.

The work-sample list records course or class, task, learner status, selection method, submission status and protection. Samples include missing and incomplete work according to the population rule. Exemplary display material is not substituted for the selected sample.

Reviewers assess only the proposition defined. A sample used to examine feedback timing does not automatically support a judgement about the academic standard of the task.

Observation records date, duration, setting, observer role, activity, participants, applicable schedule and any departure from ordinary conditions. Criteria distinguish fact from professional judgement.

Where observation can affect behaviour, the review notes announcement and observer presence. Repetition across relevant periods is used where the condition varies. A missed visit is not coded as negative or favourable observation.

The survey register records eligible population, invitation method, response, item wording, mode, period and exclusions. Reported percentages use the correct denominator and show counts. “No opinion” and skipped items are not combined with favourable responses.

Open invitations are described as responses received, not representative estimates. Where some groups have lower response, the interpretation considers whether their experience is likely to differ.

Every derived rate specifies numerator, denominator, exclusions and unit. Where a learner can appear several times, the report states whether the measure concerns learners or events. Rounding follows calculation and does not cause totals to be forced artificially to 100.

Changes in definition or source create a break. Results on either side may be reported but are not described as trend without a valid bridge.

The analyst signs the evidence register as complete for the stated propositions. A separate reviewer confirms that each material table or judgement traces to registered evidence and that material exceptions have not been omitted.

Sign-off is not a declaration that the evidence is perfect. It confirms that known limitations and corrections are visible and that the evidence supports no broader claim than recorded.

The cases show how evidence is converted into a bounded institutional judgement. They are hypothetical and do not establish required thresholds. Each case uses a different evidence problem so that the method is not reduced to one repeated decision form.

An institution has published that marked work will normally be returned within 15 working days. A review examines 160 assignments from four programmes during one semester. Twelve were not submitted and therefore did not enter the return-time denominator. Of 148 submitted items, 104 were returned within 15 working days, 28 later, and 16 had no reliable return date.

The observed timely rate among all submitted items is at least 70.3 per cent if every missing date was late and at most 81.1 per cent if every missing date was timely. Among records with known dates, the rate is 78.8 per cent. The report does not select the latter as the institutional rate without the missing-date qualification.

Programme results range from 58 to 94 per cent among known dates. Assessment sequencing shows that three programmes schedule major tasks in the same fortnight. Learner work indicates that feedback returned after the next related task could not inform that task. The judgement states that the published timing condition was not reliably demonstrated and that the educational consequence is concentrated in programmes with clustered deadlines. It does not conclude that staff commitment or feedback quality is generally poor.

The institution lists 22 learners with recorded mobility requirements. Timetable and route inspection show that 20 can reach all compulsory classes. One learner has an authorised room change for two classes but the changed room remains inaccessible through the only open entrance. One learner's record states “arrangement agreed” without a timetable or route.

The institution-wide access rate of 20 out of 22 does not satisfy the criterion that compulsory teaching be accessible to each affected learner. The two cases require immediate verification and remedy. Small numbers limit public detail but not institutional action.

The finding also identifies a control weakness: accommodation is recorded as agreed before its operation is checked. The improvement response therefore addresses both the two individual routes and the verification point in the accommodation process.

Administrative data show an 18 per cent first-year withdrawal rate in the evening programme and 9 per cent in the daytime programme. Entry qualifications, age, employment and missing exit reasons differ between groups. A survey reaches 42 per cent of evening withdrawers, who frequently report timetable and transport difficulties.

The evidence supports a material difference in recorded withdrawal and identifies plausible access constraints among respondents. It does not establish that timetable and transport caused the whole rate difference, because respondent selection and learner characteristics may contribute.

The decision is a targeted timetable and access enquiry, with a small reversible schedule adjustment in one intake and direct follow-up. The institution does not delay all action until perfect causal evidence is available, nor present the pilot as a proven retention intervention.

A programme specification requires three supervised practical experiences before award. Records for 96 final-year learners show 91 with three verified experiences, 3 with two and 2 with ambiguous records. Award decisions have already been entered for all 96.

The review does not report 94.8 per cent completion as generally satisfactory. The criterion applies to each award. The five cases require immediate record and provision review under the competent academic authority. Ambiguous evidence is not coded as completion.

The system finding concerns reconciliation between practical completion and award approval. Action includes correction or provision for affected learners and a control that prevents final approval before the required evidence is verified.

A support service reports that 92 per cent of referrals were “closed” within ten days. Examination of 120 cases shows that closure includes appointment offered, learner contacted, referral transferred and service completed. Only 54 cases record that the defined support took place; 19 are transfers without confirmation; 31 are offers with no recorded attendance; and 16 lack a clear outcome.

The published indicator measures administrative handling, not service received. The finding is not that only 54 learners benefited, because non-attendance and missing records require further classification. It is that the institution cannot establish reach from the closure measure.

Improvement first revises case states and confirms responsibility after transfer. A later review can assess response time and reach separately.

Programme completion rises from 72 to 81 per cent between two cohorts. During the same period, entry selection changes, the definition of completion excludes deferred learners, and a low-completion pathway closes. Recalculation under the earlier definition and common programme scope produces 75 per cent for the later cohort.

The evidence supports improvement of three percentage points on the more comparable basis, subject to cohort difference. It does not support the published nine-point improvement as a like-for-like change. The institution corrects the public comparison and marks the definition break.

Twenty teachers receive support focused on structured learner questioning. Observation is completed for 15 before and after the support; three have only follow-up evidence and two are unavailable. Eleven of the comparable 15 show the defined practice at follow-up, compared with four at baseline.

The result supports increased observed use among the 15 teachers with paired observations. It does not establish an institution-wide prevalence or a learning effect. Observation timing, selection and possible observer response remain limitations.

The institution continues targeted support, observes the missing and follow-up-only cases and examines learner participation before considering broader expansion.

Source: ICEQC hypothetical judgement casebook.

The cases show why denominators, criteria and event states matter. A favourable majority cannot compensate for an individual minimum obligation; an administrative status may not correspond to service; and a trend can be created by definition change.

They also show that limited evidence can still support proportionate action. The response is to align the claim and decision with the evidence, not to withhold every decision until a comprehensive evaluation is possible.

The instrument begins with the final judgement, confidence and affected population. It then records the condition to change and the immediate mechanism. It does not restate the broad review theme.

The action owner confirms authority. Where action crosses units, each contribution and the final accountable role are identified. An unresolved authority question is referred before deadlines are assigned.

At least two feasible options are considered for material actions unless immediate protection leaves only one safe course. The appraisal records expected operating result, time, full resource demand, reach, recurrent implication, risk and reversibility.

The option not selected remains in the record with reason. This allows later review to determine whether the decision was reasonable on the evidence then available, even if the result is unfavourable.

Milestones are observable states rather than verbs such as “progress” or “develop.” They state the item, population, date and evidence. “Assessment schedule approved for all first-year programmes and reflected in the operative timetables” is verifiable. “Improve assessment coordination” is not.

Operating status is always included. Completion of preparation is insufficient where learners have not received the intended service.

The log records planned date, actual date, responsible role, evidence, reach, adaptation, unresolved issue and next decision. Entries are made close to the event and can be corrected with a trace.

The log is not a narrative of effort. It permits analysis of where the action stopped and whether the implemented form retained the intended mechanism.

Source: ICEQC implementation log.

The baseline sheet records measure, definition, population, source, period, value, missingness and limitation. It also records whether evidence was reconstructed after the action began. A reconstructed baseline remains usable for some decisions but does not carry the same confidence as a pre-action measure.

Where the action begins urgently before a full baseline, the institution records the safest available pre-action evidence and establishes current status as soon as possible. It does not delay protection solely to improve evaluation design.

Reach uses the eligible population as denominator and distinguishes offered, received, completed and using states where relevant. The appropriate state depends on the proposition.

Reasons for non-reach are classified. The institution examines whether access, schedule, language, disability, cost or selection created uneven participation. It does not attribute every non-participation to learner choice without evidence.

The verifier examines the service under ordinary conditions and within a period suited to the mechanism. A new timetable is checked during operation; a revised assessment sequence is checked across the relevant task cycle; a support process is checked from referral to service.

The source and verifier are proportionate to consequence. The person who completed a low-risk action may record operation, subject to sample review. A high-consequence result requires competent and sufficiently independent confirmation.

The outcome sheet preserves the original measure and adds any balancing measures. It shows baseline and follow-up counts, denominators, missingness, definition change and relevant concurrent events.

Interpretation uses the weakest necessary causal wording. “Improved after implementation” is not replaced by “was caused by” unless the design warrants that statement.

The sustainability record covers recurrent finance, workload, technical competence, maintenance, governance ownership and dependence on a particular person. It identifies the period for which continuation is reasonably secured.

An action supported by temporary funds is not necessarily rejected. It is classified accurately, with a decision date before resources end and protection for affected learners.

Closure identifies the original finding, action, operating evidence, distribution, unresolved risk, responsible approval and date. It records whether longer outcome monitoring continues. “No further action” requires a reason.

Closure can be reversed if evidence proves incorrect or the condition recurs. The register links recurrence to the earlier case, preventing repeated short-term correction from appearing as separate success.

Institutional reports show counts with percentages unless disclosure risk prevents it. Counts reveal the scale and help readers identify results based on small populations. The eligible denominator is defined before calculation.

If 36 of 40 respondents report access to a service, the respondent rate is 90 per cent. If 60 learners were eligible, the evidence establishes confirmed access for 60 per cent of the eligible population and a 90 per cent favourable response among respondents. Neither figure alone describes the unknown 20.

One person can generate several events or registrations. A feedback review may report assignments returned, learners receiving at least one timely return, or courses meeting the return condition. These measures answer different questions.

The unit appears in the title, numerator and denominator. Event rates are not described as the proportion of learners affected. Where clustering by learner or course matters, the analysis acknowledges that events are not independent observations.

A rate rising from 50 to 60 per cent increases by 10 percentage points and by 20 per cent relative to the original rate. The report names the calculation. It does not use “10 per cent improvement” where the intended meaning is 10 percentage points.

Relative change can appear large when the baseline is small. Both baseline and follow-up values remain visible.

Cohort entry is fixed according to the review question. Continuation may mean re-enrolment in the next period; completion may require fulfilment within a stated duration. Transfers, deferrals, deaths and unknown destinations are shown separately where material.

Apparent completion can rise when slow completers are removed or the observation window changes. Every cohort table states entry basis, outcome window and treatment of status changes.

Cross-sectional evidence compares different populations at one or more points. Longitudinal evidence follows the same people or units. A repeated cross-section can describe institutional change but may reflect changing composition. A panel can describe change among continuing cases but may lose those most affected by failure.

The institution selects the design suited to the decision and, where useful, reports both. It does not describe a panel result as if it covered all current learners.

Simple bounds show whether missingness could alter a conclusion. If 70 of 90 observed cases meet a condition and 10 are unknown, the all-case rate lies between 70 and 80 per cent. A threshold of 75 per cent cannot be declared met or unmet without an assumption or further evidence.

Bounds can be wide and uninformative; that is itself evidence that record quality limits the decision. A modelled adjustment is used only where its assumptions are stated and defensible.

An average response time can be affected by a few very long cases. The median describes the middle case but can conceal the same tail. Institutional review therefore reports an appropriate distribution: median, range or intervals, and cases beyond a service threshold.

Where the obligation concerns every urgent case, a favourable average cannot compensate for one dangerous delay. Distribution is interpreted against the criterion.

Small-group rates can change sharply with one case. The report uses counts, avoids rhetorical comparison and considers multi-period aggregation only where the service and population remain comparable. Aggregation must not delay remedy for verified individuals.

Public disclosure is limited where a small cell could identify a person. Internal decision-makers receive the evidence needed under appropriate access.

A sample estimate differs from the full population because of selection. Where a probability sample is used, the report provides an appropriate measure of sampling uncertainty and reflects stratification or clustering. The calculation does not address non-response, measurement error or bias in the frame.

Where a non-probability sample is used, a conventional margin of error is not attached. The report describes selection and limits the inference.

A statistically detectable difference may be too small to alter an educational decision, while a large educational difference in a small population may not meet a conventional significance rule. The institution considers effect, consequence, confidence and population together.

Formal tests are used when their assumptions and decision value are clear. They do not replace an account of counts, definitions and distribution.

A benchmark is used only where the definition, population and period are comparable and where the benchmark has a legitimate relation to the review criterion. An external average is not automatically a minimum standard or a target.

Internal comparison over time can be more useful but remains sensitive to changed programmes, assessment or intake. The report identifies every material break.

Combining indicators requires explicit weights and rules for missingness and compensation. Because access, safety, academic integrity and outcomes are not interchangeable, this report favours component findings.

If an institution uses a composite for screening, it retains all components and defines conditions that cannot be compensated. The composite does not become the sole basis for closure or public ranking.

Source: ICEQC quantitative interpretation controls.

A service handles 200 valid cases. It resolves 120 within 5 days, 50 in 6–10 days, 20 in 11–20 days and 10 after 20 days. The proportion within 10 days is 170/200, or 85 per cent. The proportion beyond 20 days is 5 per cent.

If all 10 cases beyond 20 days are low-risk, the tail may indicate capacity or process inefficiency. If two are urgent protection cases, the same aggregate has a different consequence. Severity and timing must therefore be analysed together.

A programme reports 80 completions among 100 entrants in the later cohort and 63 among 90 in the earlier cohort. The completion rates are 80 and 70 per cent. Further review shows that 10 entrants in the later cohort transferred from a closed programme after completing half the curriculum; 9 completed.

The published all-entrant rate remains a valid description if the entry rule includes transfers, but it is not directly comparable with the earlier cohort without examining entry stage. Excluding the transfers solely because they improve the result would also be invalid. The report presents the structural change and, where useful, a common-entry subgroup.

Every draft finding is tested through five questions: What is the strongest contrary evidence? Which eligible cases are absent? What other explanation would lead to a different action? Which assumption most affects the result? What harm could follow if the judgement is wrong?

Answers are recorded beside the judgement. “None known” is acceptable only after the relevant evidence and participants have been considered.

The subject unit receives the source extract and proposed fact. It can identify an incorrect record, date, population or calculation. Correction requires evidence and preserves the original analytical trace.

Disagreement with the implication is not classified as factual correction. It enters interpretation or action response.

An alternative interpretation states the same evidence and proposes a different meaning. The review panel evaluates which interpretation better accounts for the pattern and limitations. It may retain both where evidence cannot resolve the difference.

The final record avoids artificial certainty. An unresolved alternative can change confidence, action or the need for further evidence.

Material dissent records the proposition, evidence, author or represented role, and consequence for judgement. Personal information is protected where disclosure creates risk. Anonymous assertion without evidence can still identify an enquiry but is not treated as established fact.

Dissent is not measured only by number of supporters. A single technically supported objection can invalidate a calculation or reveal a serious case.

Reviewers disclose responsibility for the activity, financial or professional interest, close reporting relationship and prior advocacy of an option. The review authority decides whether declaration, limited participation or replacement is necessary.

An interest does not automatically disqualify a knowledgeable participant. It changes the role and need for challenge.

When new evidence materially changes a finding, the institution reassesses every dependent action and public statement. It records whether an action remains justified on other grounds. Correction is not delayed to protect consistency with an earlier report.

Source: ICEQC challenge disposition protocol.

A challenge closes with a reasoned disposition and notification to the relevant participant where appropriate. Repeated challenges on the same unresolved weakness are linked and escalated.

The institution monitors whether people can use the correction route without adverse treatment. A formal route with no credible use is not assumed effective.

The summary states the question, population and principal finding in direct language. It avoids adjectives of institutional excellence and does not begin with activity undertaken.

Where the finding is provisional, that status appears in the first account rather than a distant note.

The evidence note gives sources, period, coverage, selection and principal limitation. It is concise but sufficient to prevent an internal sample or respondent result from appearing comprehensive.

Material definition changes appear beside the affected result.

The public status uses the controlled states: finding confirmed; decision authorised; action initiated; specified tasks completed; operating result verified; or closed with reason. Dates and responsible authority accompany material open actions.

An institution may group minor actions, but it does not group a high-consequence overdue case into a favourable aggregate.

The outcome statement identifies baseline, follow-up and population. It uses causal wording only where design supports it. Where no comparable baseline exists, current status is reported.

The statement includes distribution where a group remained underserved.

The report states the limitation most likely to change interpretation. It also identifies unresolved conditions outside institutional authority and the action requested from the competent body.

The phrase “work is ongoing” is insufficient without the service affected, responsible role and next decision date.

Source: ICEQC hypothetical public reporting model.

The annual synthesis reports material questions reviewed, coverage by educational function, open findings at start, new findings, operating closures, other reasoned closures, overdue actions, recurrence and material limitations. It states areas not reviewed.

These fields are interpreted together. More findings can reflect better detection; fewer closures can reflect stronger operating evidence. The synthesis therefore avoids an unsupported performance score.

A structured enquiry owned by the institution into a defined aspect of its educational work, producing evidence, judgement and a response. The term does not imply that institutional evidence is sufficient for every external or public decision.

A bounded question identifying provision, population, period and decision. A theme or domain is not a review question.

The applicable or adopted condition against which evidence is judged. Its source, authority and operational meaning are stated.

A dated description of provision, population, resources, responsibilities and context. It is not an evaluative judgement by itself.

An evidence-based statement comparing an observed condition with a criterion for a defined population and period.

Evidence that a response was implemented, reached its intended population and changed the defined operating condition or outcome.

Evidence that permits another informed person to understand and test the basis, method, limitations and consistency of a judgement.

Evidence observing the condition stated in the proposition. Directness depends on the claim: an invoice is direct evidence of a transaction and indirect evidence of learner access.

Support from an additional source or method with sufficiently different errors. Repeated summary of one source is not corroboration.

All persons, events, units or records to which the question and intended conclusion apply.

The operational list from which cases are selected or coverage is assessed. A frame can omit eligible units or include ineligible ones.

The absence of a case, response or item from the evidence. Its reason and possible relation to the finding are examined.

The significance of a finding considering educational consequence, rights or safety, number and identity of people affected, duration, recurrence and integrity of provision or awards.

The degree to which the available evidence supports the stated judgement. It is separate from whether the judgement is favourable or adverse.

The involvement of affected and responsible persons in supplying evidence, interpretation or decision, with roles and safeguards made clear.

Evidence inconsistent with the proposed finding or preferred explanation. It is examined and retained where material.

The stated relationship between an action, its mechanism, the operating condition expected to change and the intended result.

The eligible people or units that actually received the action or service. Offer, receipt, completion and use are distinguished where material.

The extent to which critical components of the specified action operated. Fidelity does not require every incidental detail to remain unchanged.

The first verified educational or institutional service condition expected from the action. It is distinct from administrative completion and from a longer outcome.

An informed view from outside the activity or institution that tests criteria, evidence, reasoning or result. It does not imply comprehensive independent re-performance.

A judgement sufficient for a limited and reviewable decision but subject to a stated uncertainty and further evidence.

A discontinuity in definition, population, source, assessment or provision that limits comparison over time.

A measure selected to detect a material unintended effect or indicator displacement associated with an improvement action.

A reasoned decision that the operating result is verified, the case is invalid or duplicated, or residual status has been lawfully accepted by competent authority. Referral or planned action is not closure.

A proportionate report of question, evidence, finding, judgement, action, verification and limitation. It protects individuals while preserving material institutional accountability.

An institution cannot examine every educational activity at the same depth each year. It therefore maintains a review programme that combines recurring essential questions, follow-up to open findings, emerging evidence and periodic examination of less visible functions. Selection is recorded so that unreviewed areas do not disappear from governance attention.

The programme is not a list of reports due. It is an account of which educational risks and improvement questions require institutional judgement, when evidence will be sufficiently mature and who can act on the result.

Selection draws on learner access and outcomes, complaints and appeals, staff evidence, material change, new or expanding programmes, previous review limitations, unresolved actions, resource shifts and external policy or legal change. It also includes a rotation across curriculum, teaching, assessment, support, governance and educational conditions.

An indicator can initiate a question but does not predetermine the finding. A falling rate, rising complaint count or unusual variation is screened for definition, coverage and consequence before a full review is commissioned.

Some questions recur because the consequence of failure is high or the condition changes frequently. They can concern safe and non-discriminatory access, delivery of required curriculum, integrity of assessment and awards, protection, and operation of essential learner services.

Recurrence does not require identical annual collection. A stable control can be tested through a risk-based sample, while known weak or changed areas receive complete or more frequent review. The institution explains why the current coverage is sufficient.

A new programme, site, timetable, assessment regime, information system, partnership or rapid increase in enrolment can change assumptions on which earlier evidence depended. The programme schedule therefore includes review before implementation where risk can be prevented and after operation where actual effect must be observed.

The review period is long enough for the relevant mechanism to operate but not so long that a serious condition persists without detection. Interim evidence and safeguards are used where mature outcome evidence will take time.

Open material findings take priority over repeated diagnosis of the same condition. The programme identifies the date at which operating evidence is expected and the authority responsible for delay.

A delayed action can require a revised decision rather than another progress report. Where the original criterion remains unmet, the institution considers interim protection, additional authority or cessation of an ineffective course.

Units with better records, active leadership or easy access can receive more review attention. The programme tests for under-coverage of remote sites, small programmes, part-time or evening provision, minority languages and groups less able to raise concerns.

Absence of complaints is not treated as evidence of low risk where the route is inaccessible or confidence is weak. Selection includes direct enquiry into conditions that ordinary reporting may not reveal.

The programme estimates time required from learners, teachers, administrators, analysts and decision-makers. Reviews sharing a population or source are coordinated, subject to confidentiality and purpose. Unused evidence requirements are removed.

Burden is assessed against decision value. A repeated survey that produces no changed decision is revised or ceased. A high-consequence review is not omitted solely because it is difficult, but its method is made as direct and proportionate as possible.

The schedule matches questions to available expertise. Statistical comparison, assessment validity, facility safety, accessibility and confidential protection evidence may require different competence. The institution obtains support or narrows the question where competence is unavailable.

Capability planning includes time for evidence preparation, challenge and follow-up, not only collection. Commissioning many reviews without capacity to decide or verify creates an appearance of activity and an expanding unresolved stock.

Source: ICEQC institutional review programme selection record.

The programme sequences diagnosis, action and verification. It avoids commissioning a broad new review while the evidence required to determine the previous action is due. Where related questions can be resolved together, their criteria and populations remain distinguishable.

Sequencing also protects the academic year. Collection and observation occur when the service operates normally, unless disruption itself is the subject. Decisions are timed early enough to affect the next relevant cycle.

The competent governing authority approves the programme, material changes and deferred high-consequence questions. Approval includes a statement that resources and decision routes are sufficient for the planned work.

Leadership can add an urgent review, but it records what other work is displaced. Removal of a review after adverse preliminary evidence requires explicit reason and governing visibility.

Monitoring shows questions scheduled, initiated, decided and verified; populations covered; open material findings; overdue decisions; participant burden; and correction. It distinguishes review completion from improvement verification.

An annual count of completed reviews is interpreted with scope and consequence. One rigorous review that restores a material service can have greater public value than many descriptive reports.

The institution periodically asks whether the review programme detects material conditions in time, represents its diverse provision, supports decisions and reduces recurrence. It examines missed cases, corrections after external challenge, unused findings and actions that never reached operation.

The result can change selection, authority, evidence definitions and capability. Programme evaluation is not presented as proof that all educational provision is sound. It concerns the institution's capacity to enquire and respond.

The minimum annual statement identifies the educational functions and populations reviewed, the basis for selection, material areas not reviewed, open findings carried forward, operating results verified and significant limitations in evidence or capacity.

This statement allows the public and governing authority to understand coverage without publication of sensitive case detail. It also prevents a favourable annual narrative from implying that silence equals confirmed quality.

A credible programme gives priority to consequence, unfinished remedy and under-observed populations. It reserves sufficient capacity for verification. It does not use fixed rotation to postpone an urgent condition or repeated data collection to avoid a difficult decision.

The proper institutional measure is whether review attention reaches the educational conditions most in need of judgement and whether those judgements produce verified change. Scheduling discipline is therefore part of self-review integrity, not a separate administrative exercise.

Evidence available at programme-planning stage will often be incomplete. The institution records whether the uncertainty concerns the existence, scale, cause or consequence of a condition. It then selects the least burdensome enquiry capable of resolving the uncertainty that matters to the decision. A brief screening may establish whether a full review is warranted; it is not published as the final evaluation.

Uncertainty is treated differently where delay can cause harm. A credible indication of unsafe access, discriminatory exclusion or invalid award practice leads to precaution and competent examination even when prevalence is unknown. By contrast, an unverified difference in a low-consequence indicator may first require confirmation of definition and source. This distinction prevents the review programme from giving equal urgency to every unexplained variation while also preventing incomplete records from postponing protective action.

The selection record states what evidence would cause the institution to increase, reduce or end the enquiry. A review that continues despite repeated evidence that its question cannot inform a decision is closed or redesigned. A review that uncovers a wider material condition is expanded through an approved change, with the original and enlarged populations kept distinct.