The summary.
SMALL NUMBERS SHOW PATTERNS THAT ARE COINCIDENCE
Ten records will always show something. Treat small-sample findings as prompts, not conclusions.
The practical test available to a small business: does the pattern persist across several periods?
MOST MOVEMENT IS VARIATION, NOT CHANGE
Establish the typical range for each figure from twelve or more periods.
Then investigate what falls outside it, or several consecutive periods moving the same way.
Do not demand explanations for normal fluctuation.
ALWAYS GIVE THE UNDERLYING NUMBERS WITH A PERCENTAGE
Two out of three is sixty-seven percent and means nothing.
Percentage change and percentage points are different. So are markup and margin.
AVERAGES DESCRIBE NOBODY WHEN THE SPREAD IS WIDE
Report the median alongside, and check whether a few records dominate the total.
They usually do — a small proportion of customers, orders or products accounts for most of it.
INVESTIGATE OUTLIERS, NEVER REMOVE THEM SILENTLY
They are errors, exceptions, or the most important records you have.
THE QUESTION THAT PREVENTS BIAS
Who is missing from this data?
Analysing only customers who stayed, and concluding customers are satisfied, is the commonest business version.
