Randomization Basics

Learn Randomization Basics for free with explanations, exercises, and a quick test (for Data Analyst).

Published: December 20, 2025 | Updated: December 20, 2025

Who this is for

Next steps

Standardize your hash + seed method across teams
Add SRM checks to your experiment template
Practice with the exercises below, then take the quick test

Ready to check yourself?

Take the quick test below. It’s available to everyone; only logged-in users will have progress saved.

Practice Exercises

2 exercises to complete

Instructions

You have a table/list of 10,000 eligible users with user_id. Create a stable 50/50 assignment to variants A and B using a fixed seed. Then run an SRM check.

Generate a bucket with a hash of user_id and a fixed seed (e.g., 202501). Map 0–49 to A, 50–99 to B.
Count users in A and B.
Run a chi-square goodness-of-fit test against the expected 50/50 split.
Document the seed and assignment logic you used.

Tips (Spreadsheet)

Use a hash-like function or RAND() with a seeded surrogate if available; otherwise, simulate by hashing text via a custom formula if your tool supports it. Ensure the random value is derived from user_id so it is stable, not from volatile RAND() without a seed.

Tips (SQL sketch)

SELECT CASE WHEN MOD(ABS(HASH(CONCAT(user_id, ':', 202501))), 100) < 50 
            THEN 'A' ELSE 'B' END AS variant,
       COUNT(*) AS n
FROM users
GROUP BY 1;

Expected Output

Variant counts close to 5,000 each (± a small random fluctuation) and a chi-square p-value > 0.05 (no SRM flagged).