A/B testing and CRO, in plain English
Short, answer-first guides to running split tests that hold up: how many visitors a test needs, how long to run it, bandits vs classic A/B tests, significance, what to test first, and how AI split testing works. Written by the team that builds Luria.
What is AI split testing?
AI split testing uses software to write page versions, test them against your current page and keep the winners. How it works and how to judge it.
How many visitors does an A/B test need?
A/B test sample size table: visitors per version to detect a 10% to 50% lift at 1% to 10% conversion rates (95% confidence, 80% power), and how long.
How to A/B test a Shopify store
How to A/B test a Shopify store in 2026: what to test first, the app or Shopify Rollouts, tracking real orders, how long to run it and how to read the result.
How long should you run an A/B test?
How long to run an A/B test: at least one full week and until each version has enough visitors for your conversion rate. Worked examples and when to stop early.
Multi-armed bandit vs A/B testing
Multi-armed bandit vs A/B testing explained: how each splits traffic, what each costs you in lost sales during the test, and when to use which.
What is a holdout control group?
A holdout control is a share of visitors who keep seeing your original page, so a lift is proven against the same traffic at the same time.
Statistical significance in A/B testing, in plain English
What 95% significance means in an A/B test, why it is not a 95% chance your version is better, the Bayesian view, and mistakes that fake winners.
What to A/B test first on an online store
The first A/B tests worth running on an online store, ranked by how much they move orders: headline, offer, product page, social proof, layout.
Shopify Rollouts vs A/B testing apps
Shopify Rollouts adds native theme and checkout experiments. What it tests, what it leaves out, and when a testing app is still worth it.
VWO and AB Tasty are combining: what it means for you
VWO and AB Tasty agreed to combine in January 2026, backed by Everstone Capital. What is known, what customers should watch, and alternatives.
Why every ad should have its own landing page
Message match explained: a landing page that continues the exact ad a visitor clicked beats one page for every ad. How to build and test one.
How to run website A/B tests from Claude, ChatGPT or Cursor
Connect Luria's MCP server to Claude, ChatGPT, Codex or Cursor and let the agent add your site, install the tag, start split tests and report back.
What is conversion rate optimization (CRO)?
Conversion rate optimization explained: getting more visitors to buy, sign up or book by finding what stops them and testing fixes against your page.