Correlation Analysis Template for PMs

Pearson measures the linear relationship between two variables. It assumes both variables are normally distributed and the relationship is roughly a straight line. Spearman measures the monotonic relationship (does one variable consistently increase as the other increases?), making no assumptions about distribution or linearity. Use Pearson for continuous metrics with roughly normal distributions. Use Spearman for ordinal data, skewed distributions, or when you suspect a non-linear but consistently directional relationship.

As a rule of thumb, you need at least 30 data points per metric pair for a meaningful correlation, and 100+ for high confidence. For user-level analysis, 30 users is often too few because individual behavior is noisy. Aim for 200+ users minimum. For weekly cohort-level analysis, you need 12+ weeks of data. Fewer data points increase the risk of spurious correlations that disappear with more data.

No. A p-value of 0.08 means there is an 8% chance the observed correlation is due to random chance. The standard threshold is p < 0.05 (5%). With p = 0.08, you should collect more data to see if the correlation strengthens and becomes statistically significant. Do not make product changes based on correlations that do not pass significance testing.

Yes. Google Sheets has `=CORREL(range1, range2)` for Pearson correlation. For Spearman, rank both columns first using `=RANK()`, then apply `=CORREL()` to the ranks. Excel offers the same functions. For larger datasets (10K+ rows) or multiple analyses, SQL or Python (pandas `.corr()`) is more practical. The analysis method does not change; only the tool does.

Correlation coefficients are scale-independent. A metric measured in milliseconds and a metric measured as a percentage will produce a valid Pearson r value. You do not need to normalize or standardize your data before calculating correlations. However, for scatter plot visualization, you may want to standardize (z-score) both metrics so the plot is easier to read.