Audit of an Empty File: When Nine Athletics Dimensions Return Nothing but N/A

A file landed on my desk on a Monday morning. It had nine sections — event and performance analysis, athlete condition, competition structure and qualification, domestic landscape, rules and anti-doping, training system, risk matrix, public narrative and expectation, and industry transmission. Every cell in every section carried the same entry: N/A. Title blank. Source blank. Article type blank. One-sentence summary blank. Information points empty. The total number of populated fields in the entire document was one — 'Domain Label: athletics'.

The thread started with a number that did not exist.

In eleven years I have opened hundreds of reports carrying wrong figures. This is the first one that carried no figures at all. And an empty report has to be handled nothing like a wrong one — a wrong figure can be corrected; a blank cell cannot.

Context: how an empty file gets built

Every text-deconstruction pipeline runs in layers. First the title, source and publication date are extracted. Then entities — athletes, coaches, competitions, federations. Then information points: marks, times, venues, wind readings. Analysis comes last. If the first layer returns empty, every layer beneath it returns empty as a matter of course. That is what happened here. Nine sections of N/A are not a property of an athletics article; they are a report on the pipeline.

In April 2026 I scraped five years of AFI results. What I found then was the same disease running the other way. The 2026 Federation Cup men's 400m final carried three timing entries that sat 0.3 to 0.5 seconds away from the photo finish. The report ran as 'The 0.3-Second Lie' and PTI cited it. What I learned that week is still the first line of my checklist: an error never announces itself as an error. It arrives wearing the costume of a number, and the costume is the whole defence.

The empty file is the second organ of the same illness. Once the title-and-source layer fractures, the eight sections below it fill with N/A — and N/A looks tidy. Dirty data makes noise. Empty data stays quiet. We keep mistaking silence for cleanliness.

Core: the audit of nine cells

Audit 1: the implausibility test. Any published athletics item carries at least one of four things — an athlete name, an event name, a mark, or a venue. All four vanishing together is abnormal. My finding is that this failure originated in the extraction stage, not in an empty source document. My confidence on that is high. The alternative — that the original piece genuinely contained no athlete — is possible, and my confidence there is only medium. The distinction matters, because in the first case the fault is the pipeline's and in the second it is the source's, and the two have different remedies.

Audit 2: the single populated cell. 'Domain Label: athletics' is the only entry, and I will not accept it unchecked either. The question is direct: was that label extracted from text, or assigned by a classifier? If assigned, it is an inference, not a fact. Pasting an athletics label onto a document containing zero athletics entities is not precision; it is the performance of precision. I have seen such labels substitute for analysis often — as if a race's story could be written from its tag instead of its splits.

And if the label itself is uncertain, the branch question stays open. Athletics contains five separate worlds — track, field, combined events, race walking, road racing. Each carries its own injury map. Sprinting means hamstring and Achilles; hurdling means hip and knee; jumping means ankle and lumbar; throwing means shoulder and elbow; distance means stress fracture and plantar. Not one of those maps can be applied here, because we do not know which event was at issue.

Audit 3: the empty age curve. I have the curve ready — sprint peak 24 to 29, distance 26 to 31, throws 28 to 33, marathon extending past 35. All of it sits unused, because there is no athlete name and no date of birth. The PB-progression anomaly test cannot run either — the test that asks whether a single year's improvement exceeds roughly three times the athlete's historical annual average. There is no progression series, so there is no spike to locate inside it. The injury-history 'glass body' flag — two or more consecutive seasons lost — is equally unassignable, because no medical record was supplied.

Audit 4: the dual qualification door. Entry to the Olympics and World Championships runs through two channels: a direct qualifying standard, or World Ranking points. This file has no achieved mark, no ranking position, no window date. National selection architecture is thornier still: in some countries a single trials race decides everything, and even a reigning world record holder can be eliminated by it. Here the athlete's federation is unknown, so pronouncing on selection architecture would be dishonest.

Audit 5: 'not assessed' versus 'assessed as low risk'. The most dangerous line in the whole document is the one recording anti-doping risk as N/A. Do not read it as low risk. Absence of evidence is not evidence of compliance. That maxim is the foundation of my work. Picture a database holding 1,200 athletes, where one athlete's injury-history cell was left blank. Three years later that blank cell becomes a selection error, then a funding error, and then a story saying the athlete never fulfilled his potential. The athlete takes the blame for a blank cell.

Audit 6: the three-source rule, and why it cannot apply here. At the Paris 2026 men's javelin final I watched the live feed duplicate Neeraj Chopra's third attempt as 89.45m twice, corrupting the graphic. I cross-checked Omega timing, the AFI manual log and my own arithmetic, confirmed 89.45m, and instructed the broadcast to correct it. In the same final, Arshad Nadeem's 92.97m went into the record as an Olympic record. After that night I wrote the 'three-source verification rule' for my company: primary sensor, official log, manual backup.

Now apply that rule to this file. No primary sensor. No official log. No manual backup. A triangle needs three points and there is not one point here. That is why this file can only be fixed by re-ingestion, never by verification. Any correction generated from inside this file would be an estimate — and I do not call estimates data.

Audit 7: the rules layer, which demands everything and supplies nothing. Had an athlete existed in this file, I would first check whether the mark was record-legal. World Athletics rules are plain: records require a tailwind of no more than +2.0 metres per second. Altitude above roughly 1,000m assists sprints and jumps and penalises endurance races. Road shoes are capped at a 40mm sole. A single false start has meant disqualification since 2026. Top athletes must file whereabouts; three missed tests in twelve months constitute a violation. Samples are stored for ten years, which means a medal can be reclaimed long after the ceremony.

The whole checklist is ready. There is no way to apply it, because there is no event, no athlete, no mark, no date. The rules test that would apply to any story cannot apply to this one, because the story did not survive.

Audit 8: the economy nobody counts. Indian athletics has no league. Its professional structure sits elsewhere — sports-quota jobs at Railways, Services and the public-sector undertakings, TOPS funding, state cash awards. All of it is conditional on a recorded performance. For an athlete whose record sits blank in a database, that structure closes its door, whether or not he was standing on the track.

And coverage? I treat it as a measurable variable. The number of athletics bylines in an Olympic year against the same count in a non-Games year tells you how seriously the pipeline is being watched. Three seasons of events, one question: who is actually doing the work? In a non-Games year bylines thin out, archives go unmaintained, and the extraction layer fails in silence because nobody is looking. I do not have that byline count either. That is the real point — the thing I want to measure is itself unavailable.

Contrarian: do not read an empty file as low risk

Two reactions are natural, and both are wrong.

The first is to treat the empty file as safe. When nine cells say N/A, a reader drifts toward the conclusion that no risk was found, therefore no risk exists. This is the same error I met in timing data — a blank cell is not 'zero', it is 'unknown', and reading unknown as zero is the most expensive mistake in the trade. It happens quietly, which is why it is so rarely caught.

The second is to fill the blanks with names. The industry does this. One agency copy becomes five bylines, and nobody feels the need to re-check the primary source. A dedicated athletics beat barely survives a non-Games year, no archive is kept, and institutional memory contracts to the next story.

I will also admit my own habit is at risk here. My signature is counter-intuitive discovery — finding the contradiction everyone else walked past. An empty file is bait for that habit. The temptation is to pull a speculative discrepancy into a narrative, because 'everything checked out' is not an exciting piece to file. I will not do it. The honest finding is that the audit failed, and a failed audit is still a finding, entitled to one honest paragraph and no more.

The fault belongs to no individual. Understand first why the error survived: agency filing at the title layer, no dedicated beat, and an archive nobody owns. Errors survive inside those conditions. Calling a previous chronicler negligent, without naming the conditions, is easy and wrong. I lost a freelance gig once for refusing to publish a preview without my template. The same system that made me rigid makes that reporter hurried.

Takeaway: one date before the next story

Do not forward the file that came back empty. Re-run the extraction stage against the original document, and audit the extraction step, not only the source. Backfill the publication date, because without a date there is no heat cycle and no peaking window. I rebuilt the whole audit from scratch, then checked it twice — and both times the result held: one populated cell, eight empty.

Athletics' commercial exposure rests entirely on the Olympic cycle. Which means the value of any single news item depends on where it sits inside that cycle. This file does not record that position either. If we cannot place a story in time, how exactly do we expect to find the athlete inside it?