A “Tennis” Label on Car Sales: The Data Lesson Vietnamese Tennis Has Not Learned

Open the file and the first line reads “Domain Label: tennis”. Below it sit forty-five information points. Not a single player. Not a single tournament. No set, no ranking, no final. All forty-five points describe car sales in Pakistan: Suzuki, Toyota, Honda, Changan, BYD, MG, Kia, Hyundai; the PAMA manufacturers' association; general sales tax; internal combustion engines competing with electric and hybrid vehicles.

I sat still for three minutes. Not out of surprise — forty-four years in this industry teach you that bad data always hides in the corners nobody inspects. I sat still for another reason: if an editorial system can slap a “tennis” label on car sales, what label is Vietnamese tennis putting on its own numbers?

What matters is that nothing in the pipeline stopped the error. A wrong label travelled the full length of the process and was caught only at the last step, when a human read it and noticed there was no tennis player anywhere in the document.

The car industry has something Vietnamese tennis does not

The Pakistan Automotive Manufacturers Association publishes monthly sales figures. One source, one definition, one publication rhythm, used by everyone. That is why a financial journalist can place two data series side by side — rising sales, missing policy — and publish an argument that can be checked the following month. If the number is wrong, the next release exposes it.

Vietnamese tennis has no equivalent structure. Ask how many people play tennis in the country and you get at least four answers depending on who is asked. Clubs count enrolled students. Court owners count peak-hour bookings. The federation counts players registered for national events. Grassroots organisers count participating teams. Four definitions, four sample sizes, four time windows. Nobody is lying. Nobody can use anyone else's number.

The practical consequence: every debate about whether Vietnamese tennis is rising or falling cannot be settled with data, so it is settled with sentiment. Sentiment always leans toward whoever speaks loudest, or whoever has the most money in the room.

I followed the ATP Challenger series in Ho Chi Minh City from 2026 to 2026, both from the stands and from the balance sheet. On the sporting side, it put Vietnamese tennis on the ATP points map. On the operational side, we never maintained a consistent attendance series long enough to prove that value to the next season's sponsor. The event ran, the event stopped, the data stopped with it. Once the data stops, nobody can prove what ever worked.

In 2026, Ly Hoang Nam and Sumit Nagal won the Wimbledon boys' doubles title — a verifiable fact, with a date and a source. But if someone asks me how many Vietnamese children enrolled in tennis programmes in the following twelve months, and how many were still playing three years later, I have no answer I could defend in front of a demanding sponsor. A powerful media moment, however strong, cannot replace a continuous measurement series.

Three layers of a mislabelling

The error in that file has three layers, and all three have a twin inside tennis.

The first layer is entity extraction. The system reads a document, assigns a topic label, but finds no player, tournament or tennis organisation. That absence should have been an alarm. It was treated as a gap to skip over. In Vietnamese tennis, this layer corresponds to counting players: a child training twice a week at a district club and a ranked professional with ATP points are entirely different entities, yet in many industry reports both land in the same box called “participation”.

The second layer is label-to-content consistency. The label says tennis, the content says cars. They diverge by one hundred percent, and no step compares them. In tennis, this is where sponsorship reports appear with a “reach” metric computed from total post impressions — with no distinction between real people and bot accounts, between someone watching three seconds and someone watching three minutes. The metric is technically correct. It simply does not measure what the sponsor believes it measures.

The third layer is downstream contamination. Once a wrong metric enters the summary table, it stops being a data row and becomes a premise for next season's budget allocation. And it stays there a long time, because nobody has an incentive to audit a number that flatters them.

These three layers explain something I see repeatedly at clubs and academies in Vietnam: plenty of spreadsheets, very few decisions reversed because of them. Data exists for reporting, separated from data used for deciding.

A metric without a definition is just an opinion written in numbers.

Two times my own model failed

In 2026 I consulted for Becamex Binh Duong while the club was losing the communications battle to bigger rivals. Instead of buying advertising, I collected six months of social engagement data on twenty-seven players. One nineteen-year-old forward showed 340 percent engagement growth in nine matches, 4.2 times the squad average. We shifted budget toward personal branding for young players, pairing it with behind-the-scenes content and livestreams. Fourth-quarter merchandise revenue rose 28 percent.

The interesting part is not the 340 percent. The interesting part is that six months of data on twenty-seven subjects, with a single definition of “engagement”, was enough to reallocate a budget. Small sample. Clear definition. Closed time window. That was sufficient.

In 2026 I built a sponsorship-effectiveness model for five Vietnamese brands around the World Cup, drawing on data from sixty-four matches. The model projected 2.1 million impressions for a beer brand. Actual: 780,000. I spent two weeks re-auditing the data and found the missing variable — time zones, and the Vietnamese habit of watching live football late at night. The model was arithmetically sound and behaviourally wrong.

A wrong forecast is not a failure; it is free data for the next calculation.

But turning a mistake into data requires a condition: you must record the original forecast, with its timestamp and assumptions, before you know the outcome. Otherwise memory quietly rewrites the past for you, and you learn the wrong lesson.

Since then, every report I write for clubs in the region carries a closing section called “limits of the analysis” — the variables outside anyone's control that could bend the conclusion. That section is rarely read closely. It is also the only part of the report I would defend in court.

When the stadium closes, data becomes an asset

In 2026, with Covid-19 suspending every competition, Becamex Binh Duong lost all ticket revenue — an estimated 12 billion dong in four months. Management proposed cutting all communications spending. I objected, on a simple argument: this was the only window in which we could convert fans from rented data into owned data.

Using the dataset built since 2026, we segmented 18,000 loyal supporters and designed a membership package at 99,000 dong per month with exclusive content: online press conferences, Zoom interviews, behind-the-scenes material. Six months later: 4,200 members, 415 million dong in revenue, enough to keep the youth team's operating fund alive.

Four thousand two hundred is not a big number. Its structure is what matters: 4,200 direct relationships, with names, with payment histories, contactable next season without depending on any third-party algorithm.

Vietnamese tennis sits at exactly this point. Academies and clubs already hold direct relationships with parents and students — the most reliable payers in the entire ecosystem. Most of that still lives in notebooks, group chats and coaches' memories, never converted into a data asset that can be taken into a sponsorship negotiation.

The counterintuitive point: this industry lacks refusal, not data

In seven years of consulting, I have never met a Vietnamese sports organisation short of spreadsheets. I have met many with ten dashboards and not one decision reversed by them.

The conventional view says the problem is missing data, so the solution is collecting more. I think it is the reverse. The competitive advantage of a tennis organisation over the next five years will lie in its capacity to refuse: refuse to put undefined metrics into KPIs, refuse to sign sponsorship reports built on impressions, refuse to publish a figure it cannot reproduce next cycle.

New media does not kill brands; it exposes brands without substance.

The same logic applies to data. New platforms have not made sports data better — they have made meaningless metrics easier to produce, and easier to catch.

One uncomfortable consequence follows: if Vietnamese tennis builds a serious data standard, several currently celebrated programmes will look smaller than they do today. That is precisely why data standards tend to be postponed by the people with the largest interest in keeping things blurry.

And here I have to warn myself. I am a foreigner working in Vietnam, and I have more than once applied a Western measurement frame to a market with different consumer behaviour. The 2026 World Cup miss is the evidence. Any data standard for Vietnamese tennis must be built from domestic behavioural data — which hours parents bring children to practice, which months courts fill, which grassroots events draw genuine spectators — not from an imported sample.

What to do, measured in weeks

A domain-consistency gate, at its simplest, needs three lines of logic: if the topic label is tennis and no player, tournament or tennis organisation entity is extracted, block the record and route it to a human reviewer. The cost is close to zero. It would have stopped a forty-five-point file from entering the system.

Vietnamese tennis needs a sector-level version: one definition of “player”, one definition of “spectator”, and a publication rhythm regular enough that the series can be challenged. It does not need to be perfect. It needs to be consistent.

So when someone tells you Vietnamese tennis is growing at a double-digit rate every year, the first question should be: growing on what, measured with which instrument, over which window, and who last audited it?

If you cannot answer that in thirty seconds, you do not have data. You have a belief set in a nice font.