See every relationship in one correlation matrix.

Build and interpret a correlation matrix

This free correlation matrix maker summarizes the relationships among several numeric variables in one square heatmap. Each cell reports a coefficient from −1 to +1, while color makes strong positive, negative, and near-zero relationships easy to scan.

Import numeric columns

Upload CSV, TSV, TXT, JSON, XLSX, or XLS data, or paste a table. Select at least two numeric columns, preview the observations, and exclude rows you do not want to analyze. All processing stays in your browser.

Choose Pearson or Spearman deliberately

Pearson correlation measures linear association and is sensitive to influential observations. Spearman correlation first ranks the values, making it useful for monotonic relationships, ordinal measurements, and data where spacing is not reliably linear.

Decide how missing values are handled

Complete rows requires every selected value. Listwise deletion uses one common set of complete observations for the entire matrix. Pairwise deletion uses all complete observations available for each pair, so sample sizes may differ between cells.

Read color and coefficients together

The diagonal equals one because every variable is perfectly correlated with itself. Off-diagonal cells are mirrored across that diagonal. A coefficient near +1 indicates a strong positive relationship, near −1 a strong negative relationship, and near zero little linear or monotonic association.

Correlation is not causation

A strong coefficient does not prove that one variable causes another. Hidden variables, time trends, repeated measurements, nonlinear patterns, and outliers can all create misleading correlations. Inspect the underlying data and use subject-matter knowledge before drawing conclusions.

Export a readable matrix

Adjust colors, decimal precision, labels, matrix shape, and cell borders before downloading PNG or SVG. For many variables, use a larger square export so names and coefficients remain legible.

Practical example: screen relationships in a survey

A survey dataset might contain satisfaction, response time, purchase frequency, and support contacts. The matrix provides a compact first look at which pairs move together and which are nearly unrelated. It is a screening view, not a final model: each interesting cell should lead back to the raw observations, question wording, sample design, and a plot of the pair.

Coefficient size needs context

A value that is meaningful in one field may be weak in another. Sample size, measurement reliability, restricted ranges, repeated observations, and outliers all affect a coefficient. Pairwise missing-value handling can also give different cells different sample counts, so record the chosen policy and review counts before comparing colors.

Follow the matrix with a scatter plot

Two variables can have a coefficient near zero while following a curved relationship, and one unusual point can create a strong coefficient. Open a relevant pair in the scatter plot maker to inspect shape, clusters, gaps, and outliers. Use domain knowledge and an appropriate statistical analysis before making causal or high-stakes claims.