Methodology

Eligibility is a hypothesis. An approval is a promise. Only delivery is a fact. Only facts enter the headline figure.

This is enforced by the structure of the database rather than by good intentions, because whichever number gets optimised is the number that improves. If matched value were the metric, the rational move would be to loosen matching and add programmes, which actively harms students by burying them in applications they cannot finish. If verified accessed value is the metric, the rational move is to fix the step where students drop out.

None of the following ever contributes a cent to accessed value:

• the theoretical or maximum value of a programme
• opportunities discovered, or programmes matched
• the advertised maximum of a variable award
• eligibility of any kind, including confirmed eligibility
• started applications
• submitted applications
• approvals that have not yet delivered anything
• the unrealised remainder of a recurring benefit
• anything a student reported with nothing to support it

Each of these has its own counter with its own name. They are reported separately, they are never summed together, and none of them is ever called impact.

An outcome is accessed only when a student actually:

• received money, or
• received a credit against an education cost, or
• used a fee waiver on a transaction they completed, or
• received a qualifying product or service, or
• had an eligible education expense waived or removed

Approved but not yet delivered is tracked separately and is worth zero in the financial headline.

The headline counts tiers A, B and C. Tier D is stored, is shown to the student as their own record, and appears on the impact page under a separate label. It never enters the headline.

• A the provider or institution states it directly to SCRN
• B official documentary evidence supplied by the student, such as an award letter or a disbursement statement
• C an approved institutional partner attests on the student's behalf
• D the student said so, with nothing supporting it

Every tier B and tier C item is reviewed by a person before any ledger entry exists. A reviewer may not review evidence their own account attested to, a kind of evidence has a ceiling it cannot be graded above, and once value has been booked against a piece of evidence its tier is frozen. Those four rules are enforced inside the database, so they hold however the review is triggered.

Value accrues per realised period. An annual maximum is never booked on the first payment. A transport benefit stated as 1,200 dollars a year that has delivered three monthly credits of 100 dollars counts 300 dollars. The remaining 900 is not an asset. It is a hypothesis about the future, and hypotheses do not appear in the ledger.

A waiver counts when the fee was actually waived on a transaction the student actually completed, at the amount the provider publishes. Five application fees of 80 dollars actually waived counts 400 dollars. A waiver granted and never used counts zero.

A device, a bus pass or a meal programme counts at the value the provider states. If the provider states no value, the outcome is recorded and counted in the outcome totals, and contributes zero to the financial figure. SCRN does not appraise donated goods. Estimating a market value for something a provider did not price is how honest numbers become dishonest ones.

The ledger is append only. Updates and deletions are refused by the database and the permissions to perform them are revoked, so application code cannot opt out.

A correction is never an edit. Correcting 500 to 350 is a reversal of minus 500 pointing at the original entry, plus a correction of plus 350. Every published figure is a sum over an append only table, so any number can be recomputed as of any past instant, and the correction history is itself publishable. The count of reversals is on the impact page.

The same student obtaining the same benefit for the same period cannot be counted twice. A database trigger holds a lock on the combination of student, programme and benefit period and refuses a second live entry.

A limit worth stating plainly: this cannot catch one person holding two accounts. Near duplicate outcomes are flagged for a human, and identity resolution is deferred rather than solved.

Every figure on the impact page is computed by a database function at the moment the page is requested. There is no cached total, no stored counter and no number written into the code. If the database cannot be read, the page says so rather than showing a stale figure.