Reading a sales trend without fooling yourself

The human brain is a pattern-finding machine with no off switch, and a sales chart is where it does its worst work. Three good weeks become "growth". Three bad ones become "decline". Decisions get made. And most of the time, the pattern was never there — just noise wearing a trend's clothing.

The three comparisons that keep you honest

Compare with the same period last year, not last month. Retail and hospitality breathe seasonally. August down on July is usually just August; August down on last August is information.

Ask "compared with what?" before reacting. A 15% dip means nothing until you know your normal variation. If weekly sales routinely swing 20% around their average, a 15% dip is a Tuesday, not a trend. Eyeball your last six months of weekly totals: the spread you see is your noise floor, and anything inside it deserves no decision at all.

Separate volume from price. Revenue can hold while units collapse (you raised prices; fewer, bigger sales) or collapse while units hold (discounting is eating you — see The 20% sale that costs you 35%). A revenue line alone hides both stories.

When a wiggle becomes a signal

The honest tests: has it persisted for longer than your normal noise runs? Does it survive the year-on-year comparison? Does it show up in more than one place — sales *and* basket size, sales *and* returns? A signal that fires on all three deserves a decision. A signal that fires on one deserves a note in your diary and another look next week.

This discipline — modest thresholds, corroboration before conclusions — is exactly how we built trend detection into Kauzio: it will not shout about a dip until the dip has proven itself against your own history, because an alert that cries wolf teaches you to ignore alerts, and that costs more than the dip.

The goal of trend analysis is not to react faster. It is to react *only when reacting is right* — and to sleep through the noise.