A school can meet a target and move away from its purpose at the same time. It can raise attendance while increasing the presence of students who are unwell, reduce behaviour reports by moving conflict into unseen places, or improve an average because those facing the greatest barriers no longer sit the test. The figure is not necessarily lying; it is answering a question that is too narrow with considerable precision. Better measurement begins when the institution stops asking one number to tell the whole story and creates a conversation among outcomes, processes, distribution, and experience.
The problem with the single figure is not only statistical. It also organises attention. What enters an indicator receives time, money, and recognition; what remains outside can appear secondary even when it sustains learning. A school watching only pass rates may overlook task quality, support received, or emotional cost. One watching only satisfaction may confuse comfort with learning. One counting activities may reward movement without effect. A learning school asks what each measure reveals, what it conceals, and what behaviour it might invite from people who know they will be judged through it.
A figure is a window, not the building#
Every indicator compresses a complex reality to make it visible. That operation is useful: it helps identify trends, compare periods, and decide where investigation is needed. Error begins when the representation replaces what it represents. A reading rate summarises different performances; an average delay combines quick journeys with extreme waits; a participation percentage may count a signature without checking understanding or influence. A measure offers one partial window. Looking through several windows does not remove uncertainty, but it reduces the risk of mistaking a convenient view for the entire building.
Every indicator compresses a complex reality to make it visible. That operation is useful: it helps identify trends, compare periods, and decide where investigation is needed.
The first discipline is to write the question before selecting the data. If the question asks whether a new intervention improves access to reading, reading speed provides information, but so do engagement with texts, comprehension, enjoyment, independence, and the opportunities missed during support. If the question asks whether a rule improves safety, recorded incidents must sit beside help requests, avoided places, belonging, and differences among groups. The set follows from the purpose rather than from whatever happens to be available on a digital dashboard.
It is also important to distinguish a signal, a target, and a threshold. A signal indicates something requiring attention; a target states the desired direction; a threshold activates a response. Confusing them creates damaging automatic reactions. Rising absence can be a signal for conversation, not proof of disengagement or an immediate reason for punishment. Falling grades may require examination of assessment rather than lower expectations. By naming the function of every measure, the institution prevents exploratory information from quietly becoming a judgement about a person.
A small dashboard that requires thought#
A useful frame brings together four perspectives. The first examines outcomes: what changed in learning, access, wellbeing, or relationships. The second follows processes: who received the action, at what intensity, within what time, and with what fidelity. The third considers equity: how benefits, burdens, and errors are distributed across groups and journeys. The fourth gathers experience: what students, families, and professionals believe the measure meant. No column automatically compensates for failure in another. Together they support a more truthful explanation and a more cautious decision.
The dashboard should be deliberately small. Eight to twelve signals usually produce more thought than fifty traffic lights. Each one needs a definition, source, frequency, owner, permitted breakdowns, and limits on interpretation. An uncomfortable sentence should be added: “this evidence does not show that…”. That warning protects against easy inference. Attendance does not prove engagement; access to a platform does not prove comprehension; a survey completed by few people does not represent the community; a temporal association does not prove that an initiative caused the change.
Combining sources is sometimes called triangulation, but its value does not lie in accumulation. Two administrative records generated by the same form are not independent perspectives. Diversity appears when a quantitative trend is tested against work samples, observation, short interviews, and user journeys. Agreement among sources increases confidence. Disagreement should not be averaged away to erase tension; the team investigates which population, moment, or mechanism explains it. A discrepancy is potential knowledge, not dirt to be cleaned from a report.
Measuring for learning without manufacturing defensive behaviour#
Measures change the system they observe. When a punctuality target becomes a public judgement, some teams may record generously, close the gate earlier, or exclude complex cases. When the number of incidents defines success, willingness to document them may fall. When a test carries prestige, the curriculum narrows towards tested material. Every indicator therefore needs an incentive review: what a reasonable person might do to improve the figure without improving reality, and who would bear the cost of that adaptation.
Separating learning from accountability reduces this risk. Teams need protected spaces where an adverse signal can be treated as early information rather than automatic blame. Protection does not mean impunity. It means distinguishing honest error, negligence, and manipulation, with proportionate consequences for each. A learning meeting first asks which mechanism produced the outcome and what support or redesign is appropriate. Oversight then retains responsibility for action, documentation, and repair. People can tell the truth without the institution abandoning its duty to respond.
Technical quality is also ethical. Data must be reliable enough for the decision they support, yet every decision does not require laboratory precision. A reversible trial may use small samples and frequent review; a decision that restricts rights needs stronger evidence, broad participation, and analysis of adverse effects. The school applies proportionality: the greater the reach, duration, irreversibility, and possible harm, the greater the demand before acting. Urgency does not abolish the rule; it requires safeguards and an early review.
Three cases in which the average stopped being enough#
Case 1: reading improvement that concealed withdrawal#
A primary school introduces twenty minutes of adaptive digital reading each day. After eight weeks, the dashboard shows more correct words per minute and leaders celebrate progress. A teacher notices, however, that some students choose very easy texts to protect their scores and that two learners with visual difficulties participate less. Interviews reveal that several pupils read faster but talk less about meaning. The average answers a question about fluency; it does not yet answer questions about comprehension, access, or reading identity.
The team retains the fluency measure and adds a monthly comprehension sample, variety of texts selected, exit rates, accessibility, and reading conversations. It breaks down patterns without publishing individual rankings and reviews six contrasting learning journeys. The platform becomes one tool among others; shared reading, accessible print, and peer recommendation return. After another cycle, speed grows slightly less, while comprehension and participation are distributed more evenly. The school records that broad, modest improvement serves its purpose better than a large average built on defensive strategies.
Case 2: fewer absences and more fragile presence#
A secondary school launches automated messages and monthly recognition to raise attendance. Its rate improves by three points and appears to validate the programme. The health office reports more students arriving unwell; some families say they interpreted messages as threats; tutors notice students arriving without materials after nights spent caring for relatives. Physical presence rose, but readiness to learn did not always rise with it. The average also concealed that persistent absence barely changed among young people facing unreliable transport.
The school replaces the general reward with differentiated support. It retains attendance and punctuality but adds staying through the day, participation, reported reasons, travel times, support referrals, and feelings of welcome. A weekly review identifies repairable barriers without turning family circumstances into permanent labels. Routes are coordinated, flexible catch-up is offered, and a human call precedes an automated message. The outcome is no longer the pursuit of countable presence: the school seeks presence that is safe, feasible, and educational, and can explain which response works for each barrier.
Case 3: administrative speed that transferred the work#
A school digitises permissions and measures the time between sending and signature. The median falls from five days to a few hours. The project is presented as a saving, although the office receives more calls and some families complete forms repeatedly because of language or password errors. The indicator counted the segment visible to the system, not the wait before it or the labour transferred to households and staff. People who failed to complete the journey disappeared from the calculation, so the figure improved precisely by excluding the hardest experiences.
The team maps the complete journey from receipt to resolution. It measures completion, attempts, help requests, comprehension, support minutes, and differences by channel. Paper, telephone, and face-to-face routes remain equally valid; language is simplified; accessibility and export are required from the supplier. The median becomes slightly worse when every case is included, but abandonment and repetition fall. The school learns that efficiency is not making the segment it controls move quickly. It is reducing the total effort required for every person to reach the service.
Who interprets, who may challenge, and who decides#
Data do not enter a meeting with meaning of their own. Situated people interpret them. The process therefore needs different voices and clear functions. The programme manager explains implementation; nearby professionals contribute context; students and families describe effects; someone with analytical competence examines quality and bias; and the accountable authority decides and explains. Participation does not mean voting on every detail. It does mean being able to alter the understanding of the problem and receiving a response to what has been contributed.
Disaggregation demands care. Comparing groups can expose inequality, but it can also harden categories, reveal small communities, or nourish low expectations. The school agrees which comparisons are legitimate, applies minimum sizes, restricts access, and avoids publishing tables that could identify a person. When a pattern emerges, the first question concerns conditions: which barrier, treatment, or resource differs. An association is not turned into a group deficit. Equity data redistribute institutional responsibility; they do not manufacture identity-based diagnoses.
The right to challenge includes correcting records, providing context, and seeking review of automated or standardised decisions. A family should know which information is used and why; a student should find an adult able to explain a consequence. Objections are documented as part of the evidence. If they repeatedly come from one group, the pattern indicates a design problem even when each complaint seems individually resolvable. The institution learns when contradiction can travel from the point of experience to the level where rules and resources are assigned.
Review rhythms and useful memory#
Not every measure should be watched at the same rhythm. Signals of harm, access, and safety require short attention; learning outcomes need enough time to become meaningful; distributional effects require comparison across periods. A staggered calendar might include weekly checks of critical incidents, monthly conversations about processes, termly outcome reviews, and annual evaluation of the dashboard itself. This cadence prevents both reacting to every fluctuation and waiting until year end to discover accumulated harm.
Every review ends with one of five decisions: continue, adapt, extend, pause, or stop. The selected option includes its reasoning, uncertainty, accountable person, date, and the signal that would justify reconsideration. A short record preserves what the school expected, what happened, which sources it consulted, and what it learned. Personal data are not retained indefinitely, and memory is not confused with surveillance. What should travel is decision logic: measure complete journeys, look for displacement, or keep an alternative route before digitising.
The dashboard must also be allowed to disappear. Indicators create work and tend to outlive the problem that produced them. An annual review removes measures that no longer guide decisions, duplicate sources, or create more burden than knowledge. Others move to sampling. The closing question is specific: what different decision did the school make because it had this information? If nobody can answer, the measure deserves redesign or retirement. Freeing attention is a condition for seeing what still matters.
Four institutional references for building judgement#
The OECD framework on schools as learning organisations connects shared vision, professional learning, collaboration, inquiry, knowledge systems, connection with the environment, and leadership. It helps a school examine whether a local observation can travel into an institutional decision and whether leaders are willing to examine their own assumptions.
The Education Endowment Foundation’s implementation guidance organises change around exploring, preparing, delivering, and sustaining. Its contribution prevents launch from being confused with effect: intelligent fidelity, adaptation to context, and preparation of conditions all become objects of observation. The European Commission’s SELFIE tool offers an example of school reflection that gathers the perspectives of leaders, teachers, and students on technology use rather than reducing dialogue to a decorative score.
The UNESCO Global Education Monitoring Report on technology in education presses institutions to ask about appropriateness, equity, scalability, and sustainability before assigning educational value to a technical solution. The four sources do not form a recipe or replace local voice. They provide questions for examining purpose, implementation, experience, distribution, and capacity for revision. Their responsible use enriches judgement instead of importing indicators detached from context.
Measuring better in order to decide differently#
Before approving an initiative, a learning school can write a short card: purpose, hypothesis, four perspectives, possible adverse effects, groups likely to be hidden, pause thresholds, and review date. It then selects the minimum data that can reveal both progress and harm. This order matters. When data are chosen first, purpose bends towards availability; when purpose leads, the institution can combine existing records with listening, observation, and proportionate samples.
At the review meeting, an average opens the conversation and never closes it. The team looks for distribution, discordant cases, and total cost; it asks which competing explanation also fits the evidence and which missing information would change the decision. Affected voices do not decorate the presentation: they enter before the conclusion. Leaders then communicate what is retained, what changes, and why. Such traceability shows that measurement is not surveillance from above but responsibility for the consequences of a collective choice.
Stopping reliance on a single figure does not mean abandoning clarity. It means constructing a more demanding clarity that can hold tensions without hiding them. A school can recognise that an intervention improves an outcome and worsens an experience, benefits a majority while burdening one group, or works for a term but cannot be sustained. Naming those relationships restores room to adapt and repair. Good measurement does not offer instant victory; it enables the institution to see better, explain honestly, and change its decisions before an indicator becomes destiny.
