The summary.
COMPARE LIKE WITH LIKE
Against the same period last year, not just last month. Seasonality makes month-to-month comparison misleading.
Watch for months of different lengths and part periods compared against whole ones.
AVERAGES HIDE THE SPREAD
An average order value tells you nothing about whether orders cluster near it or a few large ones pull it up.
Look at the median and the range at least once, and check whether a few records dominate the total.
SEGMENT EVERYTHING
Flat revenue overall is usually one segment growing and another falling — two different businesses inside one total.
Split your main figure by product, customer type and source.
TWO POINTS ARE NOT A TREND
Plot twelve periods. A chart reveals in seconds what a table conceals.
Establish what normal variation looks like, so you know what is genuinely unusual.
CORRELATION IS NOT CAUSE
Both may follow from a third thing, which is extremely common and constantly overlooked.
Ask what else changed at the same time.
WHEN TESTING A CHANGE
Record the baseline first, change one thing, and measure long enough to exceed normal variation.
WHEN INVESTIGATING A FALL
Segment it first, then rule out a data problem, before looking for business causes.
