Estimators and Estimates

The one idea: A point estimate — one sample mean, one sample proportion — is your single best guess at a population parameter you can't observe directly. It's built from a formula (an estimator) that's provably unbiased and efficient, which earns it your trust as a guess. It never earns it as an answer, because a single number carries no information about how far off it might be.

⚠️ The trap: "Our sample says average deal size is $51,814, so that's what deal size is." That feels right and it's wrong. A point estimate is a single draw from a distribution of possible estimates — the same study run again would land on a different number, and the point estimate alone gives you no way to tell whether this draw was close to the truth or way off.

Every dashboard number that starts with "our average is..." or "conversion rate is..." is a point estimate. It's the most common statistical object in business, and it's also the most commonly overtrusted one — treated as the answer instead of an answer, with an unstated amount of static riding along underneath it. This lesson is about where point estimates come from, why some formulas for computing them are better than others, and exactly what gets lost the moment you report a single number instead of a range.

What a point estimator actually is

Start with vocabulary, because business conversations blur two different things. The estimator is the formula — a fixed recipe applied to any sample: "add up the values and divide by how many there are." The estimate is the specific number that formula…