What is statistics — description, inference, and the gap people fall into

The first job is summarising numbers you actually have. The average is the one everyone knows:

In plain words: add everything up and divide by how many there are. Alongside it sits the standard deviation — a single number that says how spread out your values are. Small spread: everything close to the average. Big spread: results all over the place.

Descriptions can still mislead — one average can hide two very different groups mashed together — but they can't be wrong about the future, because they say nothing about it.

The second job makes a jump: from the people you surveyed to the people you didn't, from last month to next month. Every jump like that carries uncertainty, and statistics puts a number on it:

This is the "standard error", and in plain words it says: your estimate gets more trustworthy as you collect more data — but slowly. Because of that square root, to be twice as sure you need four times as much data. Managers forget this constantly.

The classic mistake is treating a description as if it were a conclusion. "Complaints fell 12% this month" quietly becomes "our fix worked". But every process bounces around on its own, with no cause at all. Until you know how big the normal bounce is, you can't tell a real improvement from a lucky month.

That's the honest heart of statistics: it isn't a machine for proving things. It's a discipline for stopping yourself from seeing meaning in randomness — because randomness imitates meaning very, very well.