The report card shouldn’t be the first warning.

The exam is over. The marks are in. A parent asks, “Could we have seen this coming?”

Every principal knows that conversation. Teachers may already have noticed a child hesitating in class, struggling with recent work, or quietly losing confidence. Yet those observations can be difficult to bring together before the next result arrives.

What schools need is a useful signal while there is still time to have a conversation, review the work, and offer support.

Which students deserve a closer look before the next exam?

That question is behind EDII’s student score forecasts.

We started by listening to schools

We spoke with schools to understand how they read academic progress and where an earlier signal could help their teachers. Those conversations helped us focus the work on a practical question: what would make a forecast useful in a real discussion about a child?

Turning that question into something usable took time. We had to understand the records behind the percentages: which assessment a mark belonged to, how different papers were scored, what missing entries meant, and whether earlier records really belonged to the same student and academic context.

That groundwork matters. A number can look convincing on a screen while resting on a misunderstood record. We wanted the evidence, its limits, and the teacher’s next step to be understandable together.

A clearer starting point for teachers

For supported classes and subjects, EDII brings a forecast into the student profile: an estimated exam percentage, a visible range, and context about the academic evidence behind it.

A teacher can use that view to decide whose recent work to review, which parent conversation to bring forward, and where a little more attention might be useful. The teacher brings the classroom knowledge; the forecast gives that conversation another starting point.

How the signal becomes useful

1. Start with relevant academic history. Earlier, comparable results provide a starting point. A useful forecast depends on the quality and relevance of those records.
2. Show the estimate with its uncertainty. EDII displays a percentage and a range, together with evidence context. When the available evidence does not support a forecast, an estimate can be withheld.
3. Let the teacher choose the next step. Review the student’s work, speak with the child, and choose a manageable support plan. Classroom observations and family conversations help the teacher understand what a number alone cannot explain.

The technical work behind a useful percentage

A result needs context before it can become useful evidence. Here are some of the distinctions we worked through:

• Keep the assessment attached to the mark. A score of 40 out of 50 is 80%; 40 out of 80 is 50%. The maximum, subject, class, and assessment context travel with the result. Even equal percentages do not automatically make two papers comparable.
• Match records carefully. School, student, subject, and academic session matter. A repeated identifier or an unverified link between years must not quietly combine different histories.
• Preserve the meaning of missing results. A genuine zero, an absence, a medical status, and an unavailable mark are different things. Treating them as the same number can create a misleading academic story.
• Separate earlier evidence from the outcome being checked. In modelling, a “label” is the actual result a forecast is checked against. That later outcome belongs on the evaluation side; it must not leak into the information used to make an earlier forecast.

Coded labels, with the child’s identity kept out of the story

When we explain the work publicly, we show aggregate findings and fictional labels such as Learner 001 or School A. No real student names or private school identities appear in the examples here.

Inside the workflow, identifiers help match the correct records; they are not treated as indicators of a child’s ability. Coded references support that separation, but a code alone is not proof that an underlying record is fully anonymous.

Here is what a forecast means in practice

Consider this fictional example. A teacher sees an estimated result of 58%, alongside a range of 38% to 78%.

The wide range matters. It tells the teacher to treat the estimate as uncertain, then investigate: Is recent classwork telling the same story? Is the child making progress? What would be worth reviewing together?

Even the same latest mark can sit inside very different classroom stories:

Fictional classroom data, shown to illustrate teacher interpretation. These rows are not model predictions and do not explain EDII’s calculation.

A forecast can prompt that closer look. The teacher decides what it means for the child.

The evidence, in plain numbers

In the Class 4 Mathematics historical check, forecast ranges were compared with 98 final-exam results. Of those, 85 fell inside the ranges and 13 fell outside.

That is a measure of range coverage. It does not mean the exact predicted score was correct 86.7% of the time, and it does not guarantee a future result. The ranges were broad, averaging about 40 percentage points.

What a principal can do with an earlier view

• Make academic reviews more focused. Bring the forecast, recent work, and the teacher’s observations into the same conversation. Agree on which students need attention first.
• Give parent conversations a practical next step. Share what the teacher has observed and the support being considered. Keep the estimate in context, with its uncertainty visible.
• Give support a review date. Choose a manageable action, then check classwork or the next assessment for progress. The forecast starts a conversation; follow-through gives it educational value.

Every child deserves room to surprise us. A forecast should help a school offer attention, encouragement, and challenge without turning an estimate into a label.

The opportunity is the time between seeing a signal and sitting the exam. That is where teachers can act.