Assign subjects to groups, balance by covariates, and generate Latin squares. Fully auditable and reproducible — data never leaves your browser.
Paste subject IDs (one per line) or enter a count. Add a weight column (tab-separated) to balance groups by body weight.
Try it out
Load example study randomizer data to see the full workflow
Group Labels
Enter condition names (one per line) or a number of conditions. Balanced squares control for carry-over effects in within-subjects designs.
When to use
Do not use for
If you have 3 treatment groups and use block size 4, one group will always be over-represented within each block. Use block sizes that are exact multiples of group count (e.g., 6 or 9 for 3 groups) to maintain balance within every block.
When balancing on body weight, the algorithm minimizes group mean differences for that variable. Adding multiple covariates dilutes the balancing power for each one. For most preclinical studies, body weight alone is sufficient. If you need multi-covariate balancing, consider dedicated software like R’s blockrand.
A standard Latin square controls for first-order position effects, and a balanced (Williams) Latin square additionally controls for first-order carry-over. Neither handles residual effects that persist beyond one period. If your treatment has a long washout, extend the inter-period interval rather than relying on the square design alone.
The audit trail is only useful if reviewers can reproduce the randomization. Record the seed in your protocol, lab notebook, and IACUC application. The SHA-256 hash lets anyone verify that the output was not modified after generation.
Seedable mulberry32 PRNG for reproducibility. Latin square generation via Williams’ method for balanced sequences that control first-order carry-over effects. Covariate balancing by greedy minimization of group mean differences across 1,000 Monte Carlo iterations. SHA-256 audit hash computed over the full randomization output for tamper detection.
Last validated 2026-04-03. Calculations are designed for planning and documentation support; verify procurement decisions against manufacturer specifications or institutional SOPs.
ConductScience Study Randomizer (v1.0). ConductScience, Inc. 2026. Available at: https://conductscience.com/tools/study-randomizer
Randomization is the gold standard for eliminating selection bias in experimental research. Without it, confounding variables — known and unknown — can systematically differ between groups, making it impossible to attribute outcomes to the treatment.
In animal studies, randomization is particularly important because researchers often select animals from the same breeding colony. Subtle differences in cage position, handling order, or weight can bias results if not properly controlled.
The Group Randomizer uses a Fisher-Yates shuffle powered by a seedable PRNG (mulberry32). For simple randomization, subjects are shuffled once and dealt round-robin into groups.
For covariate-balanced randomization, the tool performs 1,000 Monte Carlo iterations — each time shuffling the subjects, dealing them into groups, and computing the maximum pairwise mean difference in body weight. The allocation with the smallest difference wins.
The Latin Square Generator creates either standard or balanced (Bradley, 1958) Latin squares. Standard squares ensure each condition appears once per row and column. Balanced squares additionally control for first-order carry-over effects — critical for crossover and within-subjects designs.
