Marking queue

Loading…

Oldest first. Gates are checked before any mark is given. A submission failing either is returned, not marked low.

Ref	Learner / school	Project	Tier	Age	State

The pill shows the data-driven gate pre-check; the recorded gate check on the marking screen is the decision of record.

The submission

Gates — record the check first

Gate 1 · Ten logged relationships

Gate 2 · Dated baseline before the change

Note (required when either gate fails)

AI proposal — proposes only; you certify

1 · The artefact — 6 marks

Is the one change traceable to the costed problem, and was the same quantity measured the same way before and after, so a stranger could repeat both halves?

Watch: assessors reward a big saving. A hand-drawn table showing a fall of four litres, measured the same way twice, outscores a designed poster claiming forty litres from a before-figure nobody can date. Mark the comparison, not the size of the number.

2 · The process and its failures — 6 marks

Does the record show a change actually run, that slipped somewhere, and was measured honestly anyway — rather than a plan written up after the result was known?

Watch: a learner whose change did nothing, who logged every one of the twenty-one days and names what they would test next, takes six here. A clean win with three unexplained gaps in the log takes two.

3 · The defence — 8 marks

The standing question, every time: "would this have happened anyway without your change, and how do you know?" Restating the result is not a defence. Naming a confounder they logged, and saying what it would have done to the number, is.

Decision

—out of 20

Feedback — the artefact

Feedback — the process and its failures

Feedback — the defence

What to do next

The decision is computed by the server from the published rules: certified at 14+ with defence ≥ 3, otherwise not-yet with feedback. AI proposes; you certify. Every override is recorded, and a sample is re-marked blind for moderation.

Certified

Certificate issued.

Verification ID