# A/B testing with low traffic: how to get answers you can trust > A/B testing with low traffic works if you test bigger changes on busier pages and goals. The sample-size maths, what to avoid, and when to skip the test. Source: https://cuttl.io/blog/ab-testing-low-traffic/ · Last updated: 2026-10-30 A/B testing with low traffic works if you test bigger changes on busier pages and goals. The sample-size maths, what to avoid, and when to skip the test. ## Can you A/B test with low traffic? Yes, but only changes big enough to show up in the traffic you have. With a 5% conversion rate, catching a 10% relative lift takes about 31,200 visitors per version; a 30% lift takes about 3,800. On a small site, the change you test decides whether a test can finish at all. A useful way to judge low traffic is by time. A test is too big for your site when it would need more than six to eight weeks to reach its sample size, because over longer runs, seasons, campaigns and other site changes start to blur the result. ## How much traffic an A/B test needs The visitors a test needs depend on three things: the conversion rate of the goal, the smallest lift you want to catch, and how sure you want to be. The table uses the standard calculation for a two-version test at 95% significance and 80% power. - 5% conversion rate, 10% lift (5% to 5.5%): about 31,200 visitors per version. - 5% conversion rate, 20% lift (5% to 6%): about 8,200 per version. - 5% conversion rate, 30% lift (5% to 6.5%): about 3,800 per version. - 20% conversion rate, 20% lift (20% to 24%): about 1,700 per version. - 2% conversion rate, 20% lift (2% to 2.4%): about 21,100 per version. Two levers do most of the work. Halving the lift you want to catch roughly quadruples the visitors needed, and a goal that converts more often needs far fewer. The guide to minimum detectable effect explains how to pick the lift, and how long to run an A/B test turns visitors into weeks. ## Five ways to test with less traffic Each of these makes a test finish sooner without bending the statistics. - Test bolder changes. A new offer, a shorter form or a rewritten headline can move conversions by 20% or more. A button colour rarely does, and a lift that small needs traffic most small sites never get. - Test on your busiest pages. The home page, a top landing page or the cart see more visitors than a single product page, so the same test finishes weeks earlier. - Measure a goal closer to the change. Clicks on the main button or starting the form happen several times more often than a purchase, so they reach a result sooner. Keep the purchase as a second metric to check the change doesn’t hurt it. - Keep to two versions. Every extra version splits the same traffic further. Save multivariate tests for pages with plenty of visitors. - Run each test on everyone who reaches the page. Narrow targeting, such as one country or new visitors only, cuts the traffic a test can use. Change Impact in Web Experiments. See how Cuttl handles A/B testing with low traffic. ## What not to do when traffic is low The common shortcuts make a small test faster by making its answer less reliable. - Stopping when the result looks good. Checking every day and stopping at the first significant result raises the chance of a false winner well above 5%. Decide the length before you start, or use a method built for looking often, such as sequential testing. - Lowering the bar to 80%. It shortens the test, but one in five tests where nothing changed can then name a winner. Treat any result at that level as a lead to confirm with a follow-up test or a before and after check. - Running a test for months. After eight weeks or so, cookies are cleared, campaigns change and the visitors in each version stop being comparable. - Reading a test as a loss because it is inconclusive. On a small site, an inconclusive result usually means the change was too small to see with that traffic, and it may still have had an effect. ## When to skip the A/B test Skip the A/B test when the planner says it would take longer than about two months, or when the change is a clear fix rather than a choice between two ideas. There are better tools for both. - For clear fixes, such as a broken button, a confusing error message or a field nobody can fill in, make the change and measure it with a before and after analysis. - For finding what to change, use heatmaps and session recordings. They work from the first day with any traffic, because they show what individual visitors did. - For questions about why, a one-question on-page survey or five short user interviews can answer what an underpowered test can’t. A before and after check is less precise than an A/B test, because other things change over time too. Comparing the changed pages with the rest of the site over the same weeks removes much of that noise. ## A/B testing with low traffic in Cuttl Cuttl’s Web Experiments are built to tell a small site early whether a test can finish. On every test: - The pre-test planner estimates the weeks a test needs from your page’s real traffic and conversion rate. When it would run past the longest acceptable run (6 weeks by default, at most 8 with Bayesian statistics), it says so before launch. - From that warning, “Log as change” skips the test: you ship the change and Change Impact compares the changed pages with the rest of the site before and after. - Change Impact also picks up site updates on its own, with no release process, and gives a verdict of improved, worse, inconclusive or too early. - With the default Bayesian statistics, a test says “Too early to tell” until it has run 7 days and has 1,500 unique visitors, with at least 100 in each version, so a small test can’t be called on a handful of visitors. - The primary metric can be a click on an element or a form submit, which happen more often than a purchase. Heatmaps, session recordings and issue detection in Journey Insights work with any amount of traffic, so a small site can find what to fix first and then pick the right way to check it. ## Common questions ### How much traffic do you need for an A/B test? It depends on your conversion rate and the lift you want to catch. At a 5% conversion rate, a 20% lift needs about 8,200 visitors per version, and a 10% lift about 31,200. ### Can you A/B test with 1,000 visitors a month? Only for very large changes on a goal that converts often. For most changes at that level, a before and after check, heatmaps and recordings give more reliable answers. ### Is it OK to use 80% significance for low-traffic tests? It shortens tests, but up to one in five tests where nothing changed can then show a winner, so confirm any result before you act on it. ### What is the alternative to A/B testing for small websites? A before and after analysis that compares the changed pages with the rest of the site, plus heatmaps, recordings and short surveys to decide what to change. ### What is an MCP server? How AI assistants connect to your tools ### MCP analytics: ask Claude or ChatGPT about your website visitors ### How to measure a change without an A/B test