Spotting Patterns, Trends, Anomalies & Signals

Data is just numbers and facts sitting there. The skill that turns it into opportunity is seeing what the numbers are doing: where they're heading, what repeats, and what looks weird. This chapter teaches you to read patterns the way a doctor reads a chart — calmly, asking "what changed and why?" — and to tell a real signal from random noise. Master this and you'll spot openings other people walk right past.

The four things patterns can be

When you look at any set of numbers over time — sales, website visits, complaints, your own gym reps — almost everything you notice falls into one of four shapes. Learn these four and you have a checklist for your eyes.

Trend

A sustained direction over time — generally up, down, or flat — even if it wobbles day to day. The key word is sustained: one good day is not a trend.

Cycle / seasonality

A pattern that repeats on a schedule. Ice cream sells more every summer. A café is busy every Monday morning. The shape comes back around.

Anomaly / outlier

A point that breaks the pattern — far higher, lower, or just out of place. The "that's weird…" moment. This is where discovery hides.

Correlation

Two things that move together. When one goes up, the other tends to go up (or down). It hints at a connection — but it is not proof one causes the other.

TREND        CYCLE         ANOMALY
   /\___/        /\  /\  /\        ___    *
  /             /  \/  \/  \      /   \  (one
 /                                       point
                                         off)

Signal vs noise: the most important distinction

Numbers jiggle naturally. Sales are never exactly the same two days running. That random jiggle is noise — meaningless wiggle. A signal is a real change underneath the noise. The big beginner trap is treating noise as if it means something: "Sales dropped 3% yesterday — panic!" when 3% is just normal wobble.

A simple home test: is this change bigger than the usual ups and downs? If your daily sales normally swing between 80 and 120, then a day at 110 is noise — it's inside the normal range. A day at 220 is a signal — it's way outside. You don't need fancy maths; you need to know your normal range first.

"What changed, and why?" — the question that finds gold

Most useful discoveries start with a change. So the single most powerful habit is, whenever a number moves, ask two linked questions: what exactly changed? and then why? Don't accept the first answer. Keep asking "why?" until you hit something you could actually act on (this is the "5 Whys" idea, popularised at Toyota).

1. Spot the move. "Returns doubled this month."
2. Narrow it. Which product? Which region? Which customers? New or repeat? Big moves usually hide in one segment, not everywhere.
3. Ask why, repeatedly. "Why? — Mostly one size. Why? — A new supplier. Why? — They run small."
4. Land on an action. "Fix the size chart / switch supplier."

Compare segments and find the fastest mover

A segment is just a slice of the whole: by product, place, age, new-vs-returning, day-of-week. A single big number ("revenue is up 5%") hides the real story. Break it into slices and the story jumps out: maybe one slice is up 40% and another is collapsing, averaging to a sleepy 5%.

Then hunt for the fastest mover — the slice growing or shrinking quickest in percentage terms, not just in raw size. The thing growing 3% a week may be small today but is your future. The thing shrinking fastest is your warning. Averages hide; slices reveal.

The danger: seeing patterns that aren't there

Your brain is a pattern-making machine — so good that it sees patterns in pure randomness. Psychologists call this apophenia (finding meaning in unrelated things) — the same wiring that makes us see faces in clouds. In data this is deadly: you'll "discover" that sales rise when you wear your lucky shirt. With enough numbers, some will line up by chance alone.

Two guardrails, straight from the bias chapter:

• Correlation is not causation. Two things moving together may both be caused by a hidden third thing (ice cream sales and drownings both rise in summer — heat causes both; ice cream doesn't drown anyone).
• Could chance explain this? If you tested 20 ideas, one looking "significant" is expected by luck. A real signal usually repeats and has a plausible reason behind it.

Why experts pattern-match faster: schemas

Back in Chapter 2 we met the schema — a mental template built from experience. Experts spot patterns quickly not because they're smarter in the moment, but because they've stored hundreds of past patterns. A veteran shop owner glances at a sales dip and instantly thinks "looks like the post-holiday slump" because they've seen that exact shape before. The pattern matches a stored schema. You build these the slow way: by paying attention, every time, to what a change turned out to mean — so next time, recognition is instant.

Your single best habit: the "That's Interesting" log

The hardest part of pattern-spotting isn't analysis — it's noticing the moment something feels off, before you talk yourself out of it. Capture it. Keep a running log (notebook, notes app, one spreadsheet tab) and write down anything that makes you go "huh, that's interesting" or "that's weird".

Each entry, one line: what I noticed → why it's surprising → a possible why → what I could check. Most entries die. But every few weeks one connects to another, and that collision is where opportunities and ideas are born. (Science writer Steven Johnson found big ideas usually start exactly this way — as a vague hunch logged and left to mature, not a lightning bolt.) This log also directly fixes a problem you've named: not generating ideas from what you read — because now you're forced to react, not just absorb.

Practice

1. Name the shape. Take any number you track (steps, spending, messages sent). Look at the last two weeks. Label what you see: trend, cycle, anomaly, or just noise. Write one sentence defending your label.
2. Slice it. Take one "flat" or "fine" number from your life or work and break it into 3 segments. Find the fastest-growing and fastest-shrinking slice. What story was the average hiding?
3. Chase one anomaly. Pick the weirdest data point you can find this week and run "what changed → why → why → why" four times until you reach something you could act on.
4. Apophenia check. Take a "pattern" you currently believe. Ask: does it repeat, and is there a real reason? If either answer is no, downgrade it to "unproven".