Correlation is one of the most useful ideas in analytics — and one of the most misused. This page shows what a correlation coefficient actually measures, works through a real example, and stress-tests the classic trap: correlation without causation.
The correlation coefficient, r, is a single number between -1 and +1 that summarizes how tightly two variables move together in a straight-line pattern.
r = +1 → perfect positive · r = 0 → no linear relationship · r = -1 → perfect negative
Positive means both variables rise together (ice cream sales and temperature). Negative means one rises as the other falls (tire tread depth and miles driven). Near-zero means knowing one tells you nothing about the other.
correlation ≠ proof that A causes B
Two variables can correlate because A causes B, B causes A, both are caused by a third factor, or it's pure coincidence. Correlation only tells you they move together — never why.
Ten students tracked their study hours the week before an exam. Switch between this real pattern and two other common shapes to build intuition for what different r values actually look like as a point cloud.
Students who studied more tended to score higher — a strong, but not perfect, positive relationship (r ≈ 0.87). A couple of students beat the trend either direction, which is normal.
Each point is one student: x = hours studied, y = exam score. As points trend from bottom-left to top-right, that's a positive correlation.
A couple of students studied a lot but scored only average, and one studied little but scored well — real relationships almost never fit a perfectly straight line.
Does studying cause the higher score, or do more disciplined students both study more and score higher for unrelated reasons? Correlation alone can't answer that.
A genuinely causes B. More hours of sunlight really does cause faster plant growth.
The relationship you're hoping forIce cream sales correlate with drowning incidents — both are actually driven by hot weather, not each other.
The most common trap in real dataStruggling companies often hire more consultants — it can look like "consultants cause struggle" when it's the reverse.
Easy to get backwardsWith enough variables, some will correlate by chance alone — famously, US spending on science correlates with suicide rates. Purely coincidental.
Why large-scale "data mining" needs care