Zero-Upload CSV/JSON Pivot Table, Correlation Matrix & Outlier Explorer (2026)

Pearson correlation (r) measures strict linear relationships and is highly sensitive to extreme outliers. Spearman rank correlation (rho) converts values to ordinal ranks before computing correlation, making it ideal for monotonic non-linear relationships (such as exponential growth) and datasets with heavy skew or outliers.

The Z-Score method (`z = (x - mean) / stdDev`) assumes the data follows a normal bell curve, but extreme outliers inflate both the mean and standard deviation, potentially masking anomalies. Tukey's Interquartile Range method (`IQR = Q3 - Q1`) relies on medians/quartiles, flagging values below `Q1 - 1.5*IQR` or above `Q3 + 1.5*IQR` without being distorted by skewed tails.

In skewed metrics like API latency, cloud spend, or customer deal size, a single massive outlier pulls the arithmetic Mean upward. Reporting the Median (50th percentile) alongside P95 (95th percentile) accurately reflects both the typical user experience and tail behavior.

When two independent predictor variables exhibit a pairwise Pearson correlation exceeding `|r| > 0.80` to `0.85` (corresponding to a Variance Inflation Factor VIF > 5), linear and logistic regression coefficient estimates become unstable and hard to interpret.

No. CSV parsing, pivot aggregation, matrix math, and quartile sorting run 100% inside your browser's local JavaScript memory.