One orphan, one missing: a healthy report

Most of what we publish in this section is pathology — the mangled names, the phantom years, the checkers that cried wolf. This case file is the opposite: one unremarkable report from an ordinary late-stage draft. 213 citations paired cleanly with 95 reference entries. One entry nothing cites. One citation nothing matches. That's it. And that — not zero — is what healthy looks like.

It's worth reading a healthy report slowly once, because most people only ever read their reports in a panic. Knowing what ordinary findings look like, and how fast they resolve, is what makes the genuinely bad report recognisable when it arrives.

The whole report, in two rows

MISSING FROM THE LIST                                        1
  (Duarte, 2018)        ch. 2, p. 34    nothing in the list matches
  diagnosis: footnote added in the final week; entry never made

ORPHANED IN THE LIST                                         1
  Sekimoto, R. (2016)   list entry 41   never cited in the text
  diagnosis: the paragraph citing it was cut in draft four

everything else         213 citations ↔ 95 entries      paired

names and counts invented; the shape is a typical late-stage draft

Before the two rows, note what the report is quietly saying with everything it doesn't list: 213 citations each found their entry, including the ones that needed work to pair — the hyphenated surname cited unhyphenated, the organisation cited by acronym against its full-name entry, the shared-author year that had to borrow its byline from the citation before it. A clean pairing is not 213 easy matches; it is a few dozen easy ones and a long tail of constructions that would each read as a defect under a naive comparison. The absence of noise is the tool's work, not the document's simplicity — which is exactly why the two rows that remain deserve to be taken seriously: they survived a matcher that was genuinely trying.

Two findings, and neither is mysterious. Both have the same root cause: a document is a moving thing and its reference list is a trailing indicator. Text gets added after the list was "done"; text gets cut and the list keeps the evidence. A bibliography drifts out of sync with its text at exactly the speed the text changes, which in the last month of a thesis is fast.

Reading the missing row

The missing citation — (Duarte, 2018), chapter 2, page 34 — carries its own investigation kit: where it appears and what was searched for. Click through, read the sentence, and the story is obvious: a footnote bolted on in the final week, citing a source the writer knows well and never entered. The fix is a two-minute entry addition, and it matters more than its size suggests, because a missing reference is the defect a reader actually hits — an examiner who wants the source and cannot get there.

What you should not do with a missing row is the quiet alternative: delete the citation to silence the finding. That converts a bookkeeping gap into a claim with no support at all — and since every deletion is a recorded movement on the ledger, it doesn't even hide. The honest fixes are: add the entry, or decide the sentence doesn't need the claim. Silencing the messenger isn't on the list.

Reading the orphan row

The orphan — Sekimoto, R. (2016), entry 41 — is gentler. Nothing in the text points at it, and the draft history explains why: the paragraph that cited it died in revision four, and the entry stayed behind. Orphans are drafting residue more often than they are anything else, which is why they weigh less than missing citations everywhere we count. The triage is a three-way choice: cite it (if the source still earns a place in the argument), cut it (if it was only ever scaffolding), or — where your department permits — move it to a further-reading section. Any of the three is fine. Leaving it unexamined is the only wrong answer, because thirty of these tell a different story than one.

One orphan is residue. Thirty orphans are a habit — a list padded faster than it was read.

Why zero is the wrong target

A report with literally nothing in it, on a document of real size, is possible — and usually means one of two things: a writer who spent hours polishing bookkeeping that a tool could have reconciled in minutes, or a tool that wasn't looking very hard. Findings in the single digits, each explained by the draft's own history and each resolvable in minutes, are the signature of a bibliography that is being used — cited from, revised against, trimmed. The Ref[In] Score is built around the same idea: it moves on the defects worth fixing, it counts a citation pointing at nothing more heavily than an entry nobody reads, and it never docks you for the questions the checker itself couldn't resolve. Health is defects that are few, explained, and shrinking between drafts — the exact read a supervisor takes from the same page.

Health is a trajectory, not a snapshot

The single most useful thing you can do with a report like this one costs nothing: run it again at the next draft and compare. A bibliography's health is best read as a slope. Draft three had six missing and four orphans; draft four has one of each; the writer is winning, and — just as important — the same findings are not reappearing. A finding that comes back after being fixed is a process leak: an old version of the list pasted over a newer one, a reference manager re-exporting the entry you deleted, a collaborator's copy overwriting yours. One recurring row is worth more attention than five fresh ones, because fresh findings are the draft moving and recurring findings are the workflow fighting you. This is also the honest way to use the Ref[In] Score between drafts: not as a grade to maximise on one evening, but as a number whose direction tells you whether the last month of writing left the bibliography better or worse than it found it.

Three unhealthy shapes, for contrast

Knowing the healthy report is half the skill; the other half is recognising the three shapes that mean something structural, not incidental:

• Orphans in the dozens. Thirty entries nothing cites is not residue — it is a list that grew by import rather than by reading. The fix is not thirty triage decisions at the end; it is a habit change in the middle: entries join the list when they're cited, not when they're downloaded.
• Missing citations clustered in one chapter. A dozen missing rows, all from chapter 5, almost always means one thing: the chapter was written (or heavily rewritten) after the reference list was "finished". The cluster is good news in disguise — one working session with the list open closes all of it.
• Mirrored misses at scale. Fifty missing citations and fifty orphans, with similar names on both sides, is the signature of a systematic pairing failure rather than fifty real defects — a name-variant epidemic, a style mismatch, or a checker reading the document under the wrong grammar. The fix is one diagnosis, not a hundred edits — and a report whose rows carry evidence lets you make that diagnosis in minutes.

The common thread: the distribution of findings is itself a finding. Two rows scattered across a document are noise from a living draft. The same count concentrated, mirrored, or multiplied tells you where the process — not the bibliography — needs the attention.

The five-minute close-out

1. Open the missing row's context; add the entry or fix the sentence. Never delete the citation to mute the row.
2. Triage each orphan: cite, cut, or further-reading.
3. Re-run. Watch the two rows go to zero — or explain themselves.
4. Keep the report with the draft it describes. Next revision, the delta is your early warning.

One caution to carry out of an otherwise happy story: a report like this says your citations and your list agree with each other. It says nothing about whether the sources themselves are current — a perfectly paired bibliography can still lean on a decade-old evidence base, which is a different audit on a different axis: how old is too old for a reference list.