What a Ref[In] Report tells a supervisor

A supervisor with six theses on their desk does not have a spare week per bibliography. A REFerence [IN]tegrity report exists for exactly that desk: one page that says how a document's citations and its reference list actually relate — what matches, what's missing, what's never used — before anyone opens chapter one.

This post walks through the report we produce — formally the Academic Reference Integrity Report, the Ref[In] Report for short — the way a supervisor reads it: the score first, the two tables second, and the boundaries of what it claims third. The third part matters most, so we'll say it here too: it is a quality instrument, not a misconduct detector, and it is built to keep those two ideas apart.

What the report actually measures

Everything in the report rests on one mechanical exercise: read every in-text citation, read every reference-list entry, and pair them. Done completely — with every line accounted for — that pairing exposes the two defects that matter most to a reader:

• Missing references — a source is cited in the text, but no entry for it exists in the list. A reader who wants the source cannot get there. This is the serious one: it breaks the contract a citation makes.
• Orphaned entries — an entry sits in the list, but nothing in the text ever cites it. Usually the residue of drafting: a source that was cut, a list that was padded, an import that never got used. Less serious, still worth clearing — an examiner notices a list that doesn't match its text.

The report presents each as a table you can act on: the citation as written, where it appears, and what the checker looked for. Beyond the pairing, a currency panel summarises how old the cited sources run, and an activity summary records what was checked and how — so the report is legible as evidence, not just as a verdict.

How to read the score

The headline number — the Ref[In] Score — compresses that pairing into a single figure of bibliography health. You don't need the formula to read it well; you need to know three things about how it behaves.

1. Only real defects move it. The score is computed from the missing and orphaned tables — the two things above that a writer should actually fix.
2. Missing weighs heavier than orphaned. A citation pointing at nothing harms a reader more than an entry nobody cites, and the score reflects that asymmetry rather than counting all defects alike.
3. Open questions never lower it. Where the checker could not decide — an ambiguous pairing it flags for a human — the ambiguity is listed for review but costs the writer nothing. A tool that docked points for its own uncertainty would be grading its confidence, not the bibliography.

The score is designed to be argued with: every point it loses is a row in a table you can check by hand.

1  the score          bibliography health at a glance
2  missing table      citations with no list entry   ← fix first
3  orphan table       entries the text never cites
4  currency panel     how recent the sources run
5  activity summary   what was checked, and how

layout is illustrative; the reading order is the point

Reading one as a supervisor

The practical routine is short. Glance at the score to calibrate attention. Go straight to the missing table — those rows are the feedback that saves a student real pain later, because missing citations are what examiners chase to the library and fail to find. Skim the orphan table for scale rather than detail: three orphans are drafting residue, thirty are a literature-review habit worth a conversation. Then look at the currency panel once — if a thesis on a fast-moving field leans on decade-old sources, that is a supervision point no citation table will surface on its own.

Timing matters as much as reading. The report earns most of its value when it arrives before the supervision meeting rather than during it: the student runs the draft, fixes the mechanical rows themselves — a missing entry is not a conversation, it's a chore — and brings the residue that actually needs judgement. Meetings stop being proofreading sessions; the ten minutes once spent discovering that chapter 5's references were never entered get spent on whether the argument's evidence base is the right one. The instrument's job is to remove the findable from the meeting so the discussable can have it.

What does "good" look like? Not zero. Real, honest bibliographies almost always carry a stray orphan or one unresolved citation — a healthy report usually has something in it. The signal a supervisor is reading is not perfection; it is whether the defects are few, explained, and shrinking between drafts.

What the report is not

Boundaries, stated plainly, because a document like this gets misused when they are implied:

• It is not a plagiarism report. It never compares a student's prose to anyone else's. It compares a text to its own reference list.
• It is not a misconduct finding. A missing reference is a defect, not an act. The report deliberately has no vocabulary for intent — a formatting habit and an invented source produce the same row, and distinguishing them is a human's job, made easier because the row exists.
• It is not a grade. A brilliant argument can sit on a messy bibliography and a hollow one on a spotless list. The report measures the scaffolding, so the reader can spend their judgement on the building.

This framing is also why the report works at cohort scale. A department that runs drafts through the same instrument gets a consistent, anonymisable picture of citation hygiene across a programme — where the common defects cluster, whether they shrink after supervision — without any individual ever being accused of anything by a machine. Quality, not misconduct, is the entire posture.

Why it carries a name

The report and score are named — Ref[In], for REFerence [IN]tegrity — because an unnamed number is unaccountable. Naming the standard pins what the figure means across documents, drafts and years: a score on one thesis is comparable to a score on another because both were produced by the same published discipline — complete pairing, conservation of every line, defects counted only when a human could verify them from the table. When a bibliography has been refreshed — say after updating outdated references before submission — re-running the report shows the improvement in the same currency.

Questions supervisors actually ask us

Can a student game the score?

The crude route — deleting citations to empty the missing table — doesn't work quietly, because every deletion is a recorded movement in the run's own accounting, and a bibliography that shrank between drafts is more conspicuous than one with two honest defects. The productive way to raise the score is indistinguishable from doing the work: add the missing entries, resolve the orphans. An instrument you can only game by genuinely improving the document is gamed in the right direction.

Is the score comparable across disciplines?

The pairing mechanics don't care whether the thesis is nursing or numerical analysis — a citation pointing at nothing is the same defect everywhere, which is what makes the number comparable. What differs by field is the currency panel: reference age reads differently in machine learning than in classics, so age is presented as context alongside the score rather than folded into it.

What does re-running do to the record?

Each run reports the draft it was given; a re-run after fixes is a new snapshot, not an overwrite of history. The useful supervision artefact is the pair: the before, the after, and the delta — evidence of process, which for most committees matters more than either snapshot alone.

Can we run it across a cohort without singling anyone out?

That is the designed use at programme level: aggregated, anonymised distributions — how many drafts sit in each band, which defect classes dominate — with individual reports staying between student and supervisor. The programme learns where to aim teaching; nobody's name is attached to a number. Opt-in, aggregate-first, and framed as writing support: those aren't courtesies bolted on, they're the conditions under which measurement helps instead of chills.

The two-minute takeaway

• Score first — calibration, not verdict.
• Missing table — the rows worth a student's immediate attention.
• Orphan table — scale tells you if it's residue or habit.
• Currency panel — one glance, one possible conversation.
• And nothing in it accuses anyone of anything. That part stays human.