The character score

Character score turns signals the system already holds — attendance, results, improvement, conduct, participation, assessments — into one weighted number per person, per period.

The signals were always there. What this adds is not data, it is defensibility: a score that decides an award *will* be contested, and the only answer that ends the conversation is a table of the events behind it.

The What counts tab holds one scheme per person type. Each row is a parameter with a weight, a rubric, and a minimum.

Every metric says *why it is worth measuring*, next to its weight, because a weight nobody understands is a weight nobody sets deliberately.

The rubric is absolute, not a ranking. Full marks at the target, nothing at the floor, straight line between — so "95% attendance is full marks" is a policy you set, and this term's score is comparable to last term's. Ranking within the year group instead would mean somebody always loses even when everybody did well, and it would destroy the improvement metric the whole design rests on.

Direction lives on the metric, not on your scheme. Lateness is lower-is-better wherever it appears, so it cannot be configured backwards by accident.

Publishing bumps the version. Scores already finalised keep the version they were computed under, so re-weighting mid-year can never rewrite a number somebody was already given.

Two schemes ship as drafts with opinionated weights. Review and publish them — a blank weighting screen is where a feature like this dies.

A metric with fewer events behind it than its minimum is left out and its weight shared among the rest. It is never scored zero.

That single rule does a lot of work:

• A day scholar has no hostel record. Not applicable is not the same as bad, and they are neither rewarded nor penalised for a category that cannot apply to them.
• A teacher with eight students cannot be compared with one who has two hundred on the same eight-student measure.
• A metric that is declared but not built yet — the AI conduct signal from ilmora-connect — sits in a published scheme scoring nothing, so the day it ships nothing needs re-versioning.

Without the redistribution, leaving a metric out would quietly become a penalty, which is the exact thing leaving it out was meant to avoid.

Ranking computes a whole cohort for a period. Equal scores share a rank — two people on 84% are joint second and the next is fourth, because inventing an order between them is a decision the data does not support.

Click any row for the breakdown. That screen *is* the appeal process:

Anything excluded says so, in words, with the reason.

Finalising freezes it. Recomputing afterwards produces a new score rather than editing the old one, so nothing already awarded can move.

An award can only be made from a finalised score — an award from a draft is an award from a number that can still change. The person, period and term come from the score itself, so an award can never name something the score does not. It is not announced until you announce it.

• Their own score — everybody, on their own portal, and only once finalised. A number that decides an award and which its subject cannot see is a file, not a score. Drafts never appear, so nobody watches their own figure move while the office is still working.
• The ranking and the working — only scores.view.
• Building, computing, finalising, awarding — only scores.manage.

Ranking, one person's breakdown, the scheme audit — which publishes the weights and the rubric so the ranking can be argued with on its own terms rather than believed — and the awards register, which counts awards decided but never announced.

Ask Ilmora reaches these by name: *"best student"*, *"how is it calculated"*, *"who won"*.

Fee timeliness. It measures the parents' cash flow, not the child, and it stays a finance report.

Raw marks, for staff. Raw marks mostly measure the intake, so scoring them rewards whoever was handed the top set. Staff are measured on the improvement of their *own* students against their own prior baseline.

Planned absence. Only unplanned absence counts. A system that scores somebody down for using leave they were granted has turned the entitlement into a trap, and people notice within one term.