Business reporting runs on single figures. One conversion rate, one churn number, one average order value. They are easy to put on a slide and easy to repeat in a meeting, and that is the problem. A figure that everybody quotes stops getting questioned.

The cleanest way to see what goes wrong is to look at a dataset where nothing is at stake for anyone reading. So here is one from professional tennis, where the records are complete and public.

A statistic that is true and useless

Across 116,382 completed professional singles matches, the player who loses the opening set goes on to win 16.52% of the time. About one in six. Commentators quote it constantly and it is accurate.

Now split those same matches by the score of that first set:

– lost it 7-6: wins 21.92%, from 14,265 matches

– lost it 6-4: 19.41%, from 25,546

– lost it 7-5: 18.89%, from 10,252

– lost it 6-3: 17.39%, from 25,970

– lost it 6-2: 13.30%, from 19,739

– lost it 6-1: 11.06%, from 14,269

– lost it 6-0: 7.18%, from 5,833

Three times the difference between the top and the bottom. The best comparison sits in the middle of that list. Two groups are nearly identical in size, 14,265 and 14,269 matches, and one recovers twice as often as the other.

Notice also that the ordering never breaks. Every time the first-set margin widens, the recovery rate falls, through all seven scorelines with no exception. That consistency is what separates a real effect from noise in a sample.

What that looks like in a company

The structure repeats everywhere. A 3% site-wide conversion rate can be one channel at 9% and another at 0.4%. A 22-day average collection period can be most invoices paid in a week and a handful dragging past ninety days. A headline customer satisfaction score can be two segments pulling in opposite directions and cancelling out.

In every case the average is arithmetically correct and strategically empty. It tells you where the middle is. It does not tell you what to do, because the action always lives in one of the groups and not in the blend.

LOCAL NEWS: The 100 Best Doctors in Arizona for 2026

INDUSTRY INSIGHTS: Want more news like this? Get our free newsletter here

Two questions that find the hidden split

The first: does the pattern hold its order? Break the metric down by one sensible dimension and read the groups in sequence. If each step moves the same way, the dimension is doing real work. In the tennis data the rate falls at every one of seven steps.

The second: are the groups comparable in size? Two buckets of 14,265 and 14,269 can be compared directly, because size cannot explain the gap between them. Segments of 400 and 40,000 cannot, and the small one will swing for reasons unrelated to the thing you are measuring.

The baseline is not fixed either

The same records carry a second warning. There is often no single normal value at all. Across 2,239,440 service games, players hold serve 74.43% of the time on grass. On hard courts it is 69.95% and on clay 65.98%. A hold rate that looks ordinary in one setting is strong in another.

By tour the gap is wider again. Men’s professional matches hold 78.62% across 230,715 games and women’s 64.36% across 266,503. Both reproduce the rates published elsewhere, from a method tuned to neither, which is the test of whether a measurement is trustworthy.

The practical bit

These splits could be examined for one reason. The source records were kept at the level the events happened, match by match and point by point. Nobody decided in advance that first-set margin would be the useful dimension. It turned out to be, and the detail was still there.

That is the lesson for anyone designing reporting. Aggregate as late as you can. Keep the dimension you are sure you will never need. The question that makes a number useful is usually one you have not thought of yet. Pre-aggregated data answers only what somebody already anticipated.

The figures above come from the Live Tennis API. It serves live scores, fixtures, players and match events as JSON, with completed results and point-by-point history. They are published openly at blog.livetennisapi.com/studies.json under CC BY 4.0, free to reuse with attribution. Every number here can be checked. Method: completed best-of-three professional singles matches, 1 January 2023 to 14 September 2026. Retirements and walkovers excluded. Counted from the