> ## Content Index > Fetch the complete content index at: https://conversion.club/llms.txt > Use this file to discover other available public pages before exploring further. # How Do I Optimize My Site Without Enough Traffic to Test? - URL: https://conversion.club/playbook/how-do-i-optimize-my-site-without-enough-traffic-to-test/ - Published: 2026-10-01T20:20:12.000Z - Updated: 2026-10-19T20:26:04.000Z - Description: Replace statistical proof with a body of converging evidence. When five different sources point to the same problem, you can fix it without a test. But look at the sessions that buy too, or your evidence proves nothing. - Author: Jeremie Werner - Tags: Improve conversion, #playbook, CRO, ux, A/B Testing Replace statistical proof with [converging evidence](https://conversion.club/playbook/how-many-sessions-do-you-need-to-run-an-a-b-test/). When five different sources point to the same problem, you can fix it without a test. --- ## The benchmark Rank your evidence sources from fastest to slowest: requests received by customer support, [return reasons](https://conversion.club/fashion-e-commerce-return-rate-how-to-cut-it-without-killing-your-sales/), searches typed into your internal search engine, session recordings, a question asked after purchase, and five people observed buying on their phones. The recording tool is almost always installed, and almost never watched. A founder told us this during a diagnostic: > "I've never managed to make use of what I saw on [Clarity](https://conversion.club/how-to-optimize-your-fashion-product-page-a-full-cro-diagnosis-with-clarity/)." It took us a year to turn it into a routine: 45 minutes a week, five product pages chosen from those that convert below average or whose returns are rising, six sessions per page, three on desktop and three on mobile, with the sound off. --- ## The rule The less traffic you have, the more every change must answer an observed problem, and the bolder it must be. A small tweak on small traffic can't be measured. A blocker spotted by three independent sources is more likely to be real than a 3% gain declared significant by a tool you check every morning. --- ## The steps 1. Customer support. List the ten most frequently asked questions: each one points to information missing from the site. 2. Returns. Read the reasons for the twenty most-returned products. 3. Sessions. Watch twenty cart abandonments, then twenty purchases. If a behavior shows up in 40% of abandonments, check that it doesn't also show up in 40% of purchases. If it does, it explains nothing. 4. The five people. At a ready-to-wear brand, a test had won a 23% increase in add-to-carts, not orders yet, among new visitors. Before generalizing, five customers observed during a shopping session revealed a barrier the numbers had missed, summed up by the manager: "What they're missing is knowing how it looks on someone 5'3", how it looks in a size 12." 5. Before and after. Change one thing only, date it, compare four full weeks before and four after, with no campaign in between, and keep a control metric that shouldn't move. 6. The [logbook](https://conversion.club/playbook/how-do-i-keep-track-of-what-moved-my-numbers/). Record every change. A sentence to say to your leadership, in your own words: > "We won't test. We'll fix what customer support, returns, and twenty sessions point to together, and we'll compare four weeks to four weeks." --- ## The trap The most common trap is believing twenty sessions are a sample. They're twenty anecdotes: they prove nothing, they make a hypothesis credible, and that's already a lot when you can't test. Two other traps take one line each: copying the "best practices" of a brand that has neither your customers nor your prices, and comparing the week of the change to a promotion week to conclude what you hoped. #### Sources - [Why You Only Need to Test with 5 Users (NN/g)](https://www.nngroup.com/articles/why-you-only-need-to-test-with-5-users/?ref=conversion.club)