Sample size, practically.
THE QUESTION
How many records before a conclusion is safe?
THE HONEST ANSWER
More than most small businesses have for most questions.
WHAT SMALL NUMBERS PRODUCE
Patterns that are coincidence.
Ten records will show a pattern. It means nothing.
WHAT TO DO ABOUT IT
Treat small-sample findings as prompts to investigate, not conclusions.
WHAT MAKES A FINDING MORE TRUSTWORTHY
More records A larger difference The pattern persisting across periods
THAT LAST ONE
The most practical test available to a small business.
A pattern repeating across several months is real.
WHAT TO AVOID
Deciding from a single period Comparing segments with very few records Reporting percentages of tiny numbers
THE PERCENTAGE POINT
Two out of three is sixty-seven percent and means nothing.
Give the underlying counts.
WHAT TO DO WHEN DATA IS THIN
Say so, and decide on judgement.
That is legitimate. False confidence is not.
