A Null Record Is Still a Finding: Auditing a Failed Stage-1 Football Extraction

The record reached my desk with nothing in it to fill. The domain field carried a single word — "football". Title: not available. Source: not available. Date: not available. One-sentence summary: blank. Author stance: not applicable. Purpose: not applicable. Information points: zero. Entities: zero. I have watched matches and taken notes for fourteen years, and in 2026 I coded all 52 matches of the FIFA U-17 World Cup in India for fouls, cards and added time. Even then there was something to count, even when the count was thin. Here the count is zero. The first instinct is to treat this as a delivery fault and ask for the material again. In audit language, this record is a completed entry whose result is null, and filling a blank field with a guess is the most expensive mistake available. When a form arrives fully filled and contains only a category, the question stops being what the article said. It becomes which gate failed to close. Stage-1 extraction is the intake form of the pipeline — the place where raw material is turned into claims that someone downstream can test. The 12-point incident-report template I designed during the 2026-21 ISL bio-bubble, later adopted by eight clubs, rested on one demand: every disputed call must carry the IFAB law number. A form with the law number left blank is not a lighter form. It is an unprocessable one. The same rule applies to an extraction record. A valid Stage-1 record carries a title, a source, a date, an article type, a one-sentence summary, author stance, purpose, five to ten information points each with its own source field, the entities involved, and any time-sensitive claims. Each of these is a separate grip. Leave any of them empty and every check behind it empties too. Here all five checkpoints were found open, and that is the first formal finding of this report. Gate one: no core viewpoint exists. The summary is blank, stance is not applicable, purpose is not applicable. This does not mean the article was balanced. "Stance not applicable" and "neutral" are two different entries, and merging them into one box is precisely the habit the audit exists to prevent. Where there is no summary there is no thesis; where there is no thesis, reconstructing bias or intent becomes speculation. The sample is zero, so the causal sentence stays at zero as well. Gate two: zero information points means no claim to test. Football is a domain where claims are usually countable — penalties, added minutes, box entries, cards, assignments. Had those points existed, my next task would have been to split them a second time: home against away, referee against referee, score state against score state. When there is nothing to split, calling a thirty-thirty gap a pattern is only spending words. A record with not one fact, quote, figure or event cannot yield entities either, which means any football topic at all — match report, transfer, tactics, governance, injury — would be a guess. Gate three: source quality cannot be weighed. The source field is not applicable, and no information point carries a source field of its own. Reliability, independence, recency and evidentiary strength all lose their footing. In August 2026, when FIFA suspended the AIFF over third-party interference, I wrote a 1,200-word explainer within the week and then audited 54 transfers in the 2026-23 ISL window against the salary cap; three clubs had registered players under force-majeure clauses without written AIFF approval. Every line in that report carried a document and a date, which is why the AIFF circular of 12 September 2026 followed. Without source fields no such outcome is reachable. Gate four: time sensitivity is unassessed. There are no dates, events or temporal claims, so even the relevance window cannot be fixed. In the regular season this gate matters most — a claim built on table position or form goes stale inside seven days. A claim without a date cannot be checked, only repeated. Gate five: the article type is unclassified. News, analysis, opinion, feature, rumour and commentary are held to different evidentiary standards. A thin factual base can be a feature of the genre in an opinion column; the same thinness in a news report is a defect. With the genre undecided, even that judgement cannot be made. This is where the easiest and least honest path opens. 1,439 words is a number, not evidence. The desk asks for copy, and the fastest route looks like inventing a match, a disputed penalty and a stoppage-time gap. Such a piece would not read badly. It would be fluent, confident and entirely unverifiable, which is the most dangerous combination on offer. In 2026 a fan-site editor asked me to drop the tables from my U-17 report and use initials because readers would feel "bias". I refused both and published the full report with a methodology appendix. At the time it felt like fighting an institution. It was one principle: filling a blank transfers a debt to the reader. The second-order damage is larger. A record filled by invention contaminates the chain behind it. When downstream verification fails, the failure will be blamed on the data rather than on the hand that patched the gap with imagination. A null record sets a boundary on the loss; a fabricated one removes it. The null does buy something, and it is not small. It is a failure signature. The record carried a domain label but no body, which narrows the field: the classifier worked, the body fetch did not. That is testable rather than speculative. The trouble starts with a schema that permits "not applicable" entries without an accompanying incompleteness flag — the system will keep producing silent nulls and nobody will count them. The rule on paper against the rule in force: an empty field in a schema looks harmless until its enforcement log exists. In this case the log does not exist. Every rule leaves a trail; my job is to timestamp it. There are five trails here and all five point to the same place. The question at the morning meeting should not be what the article said. It should be which gate was left open. On the day every desk tracks its null-return rate as seriously as it counts published words, an empty form will itself become a report.