When a “Tennis” File Turns Out to Be Aviation News: Notes from a Source-Verification Case

The file landed on my desk at 6:40 in the morning, tagged “tennis”, carrying 26 information points and the usual request: build a deep analysis by the end of the day. I read all 26 points in the order my ritual demands, from the first line to the last, skipping none. No player. No set. No surface, no ranking, no serve. Instead: Adrian Hamilton Manns, chief executive of Oman's low-cost carrier Salam Air, nominated to become the next chief executive of Pakistan International Airlines after the flag carrier passed into private ownership under a consortium led by Arif Habib. Two unnamed sources. An appointment not yet publicly announced. A target date at the end of October, subject to security clearances. What stopped me was a single line of dates. Information point ten places the Arif Habib consortium takeover on 29 June 2026, a date sitting ahead of any plausible publication moment. The same file carries a second discrepancy: the title calls the subject “Adrian Hamilton”, the body calls him “Adrian Hamilton Manns”. Two small errors, very small, exactly the kind of errors I would have waved through a decade ago because they do not disturb the narrative. That same kind of error once forced me to write a letter of apology. A modern sports desk runs on pre-labelled data streams. A machine ingests the text, a machine classifies it by topic, and the writer at the end of the chain receives it with a domain label attached. That label decides which analytical frame gets applied: the tennis frame for tennis, the basketball frame for basketball, the aviation frame for aviation. When the label is wrong, every frame downstream is wrong with it, and the writer at the end of the chain is the last person who can still catch it. Across eleven years watching this industry, four of them at a Manchester newsroom, I have seen that chain work flawlessly thousands of times. It has failed once, and that single failure was enough to produce a complete analysis of something that does not exist. In 2026, as a second-year student, I reported that the referee had booked defender Trent Alexander-Arnold in the 23rd minute of the derby between the University of Manchester and the University of Liverpool. The card belonged to a team-mate. My editor reprimanded me, I wrote a letter of apology, and for the six weeks that followed I sat logging 189 card incidents from the 2026 World Cup to build myself a reference dataset. Since then I run three checks before publishing: the player's name, the minute of the event, the type of card. My first mistake was never the red card shown to the wrong man. It was the belief that I would never show one. My ritual has three layers: trace the provenance of the data, place it in its historical context, then measure its deviation from a statistical norm. Applying the first layer to this file returns an immediate result: two anonymous sources, no publication date, no author name, no source link. A record like that cannot be graded for reliability, because there is nothing to grade. On a normal sports story this layer costs me fifteen minutes. Here it cost three and returned a void. The middle layer, historical context, also returns empty, for a different reason. What is the historical context of a tennis piece? The tournament system: Grand Slam, Masters 1000, ATP 500, ATP 250, Challenger; the rules on mandatory entry, on protected ranking, on medical timeouts, on off-court coaching. This file touches none of it. The only historically structured material here is an executive career sequence: Salam Air, Fly Arystan, flyafrica.com, IndiGo, South African Airways, Pacific Blue, Mandala Airlines, Air New Zealand. A résumé, not a performance dataset. I can read it like a results table, but every comparison I draw from it would be invented. The final layer is where I am usually most patient, and also where I most easily fool myself. Standard deviation requires a standard. To say that a player holds 68 per cent of first-serve points means something only if I know the average on that surface, for that gender, in that season, against that opponent. On this file the average does not exist, so the deviation does not exist. The check returns a null, and the null is the most honest answer I have. If I skipped all three layers and wrote on regardless, what would happen? I know precisely, because I have seen such drafts in circulation. A three-thousand-word analysis with a statistics table, trend charts and a contrarian section, all built on a figure who has never walked onto any court. The table would look immaculate: first-serve percentage, return points won, break-point conversion, ranking-points defence pressure. Nobody would check it, because nobody would have a source to check it against. Readers would believe it, because professional formatting manufactures its own authority. Of the 26 information points, exactly one strategic tension is worth analysing, and it belongs to aviation. Manns describes himself on his professional profile as an airline turnaround specialist. Almost his entire career sits with low-cost carriers: Salam Air, Fly Arystan, flyafrica.com. PIA is a full-service carrier, historically a flag carrier tied to state ownership. The gap between a low-cost model and a full-service network model is a genuine integration risk, documented in airline management literature. I could write three pages on it. But if I bolt it onto a tennis frame, I would have to call it a style mismatch and compare it to a player moving from hard court to clay. The analogy sounds elegant. It is also entirely meaningless. All nine analytical dimensions on this file return null, and I want to say something about the null. In my trade, insufficient information is a conclusion, ranking alongside every other conclusion. It does not mean laziness or ignorance. It is the output of a verification process that was carried out in full and failed. The difficulty is that a null produces no copy, no traffic, no post. So it tends to get filled in with something that merely sounds plausible. Here I have to state plainly what I tell colleagues whenever technology becomes the argument. VAR is not wrong. The people operating VAR are wrong. And that is precisely where my work begins. The classification system behind this file behaves the same way: it holds no malice, it does not decide to mislabel. It does exactly what it was built to do, on an input that somebody placed in the wrong drawer. I have no intention of suing the system. My job is to point at the link in the chain that let an aviation file fall into the tennis tray. There is a deeper counter-intuitive layer, and this is the part I want sports readers to weigh. The largest failure here is not the wrong label. It is that no gate anywhere asked a simple question: does this text mention the sport it has been labelled with? One question, one count, three seconds. The chain has been optimised for label compliance, for completing the form correctly, and has not been optimised for checking content. That is the failure mode I encounter most often after eleven years in this industry: a system performing perfectly on a false assumption. The counter-argument deserves a hearing. Sometimes a wrong label exposes a real story. PIA's privatisation and the appointment of a turnaround executive is genuinely newsworthy in aviation, and done properly it would become a serious piece on integration risk. In sport, though, a wrong label exposes nothing. It only issues a licence to fabricate. There is no real story behind a serve statistics table for someone who has never held a racket. And if mislabelling happens at scale, which is entirely possible with automated tagging, then an entire batch of records from the same ingestion run may be carrying the identical fault. I log every card, every minute of stoppage time. Because a wrong number repeated three times becomes a fact in the end-of-season report. I know that from my own record: in 2026, as a first-year sport science student, I volunteered as a data analysis assistant for FC United of Manchester. In the Northern Premier League fixture against Radcliffe Borough I found two penalty-area fouls the referee had missed and the official statistics system had not recorded. I spent three days reviewing the full footage, counting every contact, building a comparison table against the match report. Three days for two fouls. But those two fouls were the only thing still standing when the official report was challenged. My proposal for the sports content chain sounds banal, and I accept that. Before any analytical frame is applied to a text, a minimum gate must run: count the entities belonging to the labelled domain. For tennis, that means player names, tournament names, surface names, round names. If the count returns zero, the record goes back to source instead of moving forward. The gate costs almost nothing. The cost of skipping it is what I have just described: a complete analysis of something that does not exist, written by someone convinced they are doing the job properly. For readers, the test is similar and annoyingly simple. A decent tennis analysis names the opponent, names the round, names the set score. If a piece is dense with tactical vocabulary yet never anchors itself to a specific match, the odds are it was built from a source with no match in it. When data contradicts the eye, trust the data, but never skip the check on where it came from. And when the data is entirely absent, trust the absence. A tournament is a system. Every referee decision is a variable. My work is simply the check. This week the check returned the finding that there is no tournament in the file I was handed. What I do next is obvious: return the record to its proper drawer, log the case, and wait for the real document. If the chain learns to say insufficient information without flinching, the quality of sports journalism as a whole improves within a single season. That is the change I want to see, and it starts with a three-second gate.