Evidence persuades when it compares
💎CORE ARGUMENT
A fact on its own rarely moves anyone. Many people died near the pump is a tragedy. People supplied by one water company died at many times the rate of their neighbours supplied by the other is an argument. What changed is not the amount of data. The second sentence holds two groups side by side, alike in everything except the thing in question. Evidence persuades when it compares.
John Snow's map of the Broad Street pump is the famous picture, but the stronger work is his comparison of two water companies serving the same London streets. Houses next door to each other, the same air, the same poverty, different pipes: one company drew from the river below the city's sewers, the other from above them. Chance had run an experiment and he read the result. Florence Nightingale made the same move with army deaths. She set those from wounds beside those from preventable disease, and the picture carried its own conclusion.
A comparison answers the question a sceptic is about to ask, which is as opposed to what? A lone number invites every alternative explanation. A fair comparison has dealt with most of them before anyone speaks.
At work this has changed how I open. I do not show what a feature did. I show customers with it beside customers without it, or the same dispatcher before and after, and I say first what makes the two groups alike. When the scorecard prototype confused everyone in November, part of the trouble was that a weighted score is a number with nothing beside it.
The weakness is in the word fair. Two groups that differ in more than one way make a comparison that persuades just as well and is wrong. My own course project had exactly that fault until Hana caught the confounded variable.
How this note changed
- 7 June 2025: First version, the week of the final exam. Snow and Nightingale only.
- 16 November 2025: Added the scorecard review: a weighted score has nothing beside it.
- 23 November 2025: Added my own confounded comparison, with thanks to Hana.
🧱BUILT FROM
🌍WHY IT MATTERS
Every result I present opens with two groups and what makes them alike. If I cannot name the comparison, I do not have a finding yet, only a number.
☑️NEXT STEPS
- Open the next readout with the two groups and what makes them alike
- Put a comparison row beside the weighted score in the scorecard prototype
- Redo my own confounded comparison the way Hana suggested
An 1855 study in which the comparison of water companies matters more than the famous map, and a modern account of framing the question before doing the arithmetic.
Compared with what is the third question in 💎 A chart is an argument and can be cross-examined. The most basic comparison of all is with the ordinary rate: 💎 A number without its base rate is a story.
A single figure leaves the counterfactual to the listener's imagination. Setting two otherwise similar groups side by side supplies it, removes most rival explanations in advance, and lets the difference itself make the case.
People are moved by one story far more than by two columns. Snow's comparison did not persuade the authorities of his day; a single case often does what a table cannot.
A fact I keep about how a diagram of two causes of death changed an army's hospitals, and my reading of why the scorecard review went the way it did.