Conversion rate optimization: get a strategy, not just a button color

Before you touch an A/B test or redesign a page, you need a strategy behind it. That's the whole difference between conversion rate optimization that compounds and "we tried a green button once and nothing happened." CRO isn't a bag of tricks — it's a repeatable loop. Here's the loop we run, updated for how the tools actually work in 2026.

1. Analyze — find where you're leaking

Don't jump straight to ideas. First find out where the funnel actually leaks and for whom. In GA4, use Explorations and funnels to see which pages and which traffic sources convert poorly, where people drop, and how new vs returning and mobile vs desktop visitors behave differently. Then add the qualitative layer the numbers can't give you: heatmaps and session recordings (Microsoft Clarity is genuinely good and free; Hotjar is the paid standard) to watch where people hesitate, rage-click, and bail. Segment everything — a flat sitewide conversion rate hides the page that's quietly bleeding money.

2. Research — understand WHY they don't convert

Analytics tells you what and where; it never tells you why. For that you need real human input: user testing, on-site surveys, five-second tests, and mining your reviews and support tickets for the language and objections of real buyers. Do a heuristic pass yourself, too — browse your own funnel with fresh eyes for the obvious friction (slow pages, confusing forms, missing trust signals, the classic UX gaffs). One warning: research competitors for ideas, but don't blindly copy them. Their "winning" page may be losing — you have no idea what their data says.

3. Prioritize — you can't test everything

You'll generate more ideas than you have traffic to test. Rank them with a simple framework (PIE or ICE — potential, importance/impact, ease/confidence) and start where the math is friendly: high-traffic, high-value pages where a lift actually moves the business. Testing a page with 200 visits a month is a great way to never reach significance.

4. Hypothesize — not "let's try stuff"

Every test should be a real hypothesis grounded in steps 1–2: "Because users tell us they don't trust the checkout (research), adding security badges and clearer return policy (change) will increase completed purchases (metric)." A hypothesis you can be wrong about is what makes a test worth running. "Let's try a different color" is not that.

5. Test — properly

Heads up if you're returning to CRO after a few years: Google Optimize was shut down in 2023, so there's no free first-party Google testing tool anymore. The current options are dedicated platforms — VWO, Optimizely, AB Tasty and the like. Whatever you use, the discipline matters more than the tool: size the test before you launch, run full business cycles, and respect statistical significance — don't peek and call a winner the moment it crosses 95%. Our free A/B significance calculator will tell you whether a result is real and how much traffic you need.

6. Review and iterate — CRO is a program, not a project

When a test ends, ask whether the result is both statistically significant and a big enough lift to matter (the two are not the same). Roll out winners, document everything — what won, what flopped, what you learned — and feed it into the next round. A loser still teaches you something about your customers. The teams that win at CRO aren't the ones with the cleverest single test; they're the ones who never stop running the loop.

The honest version

Most "CRO" disappoints because people skip straight to step 5, test something trivial on a low-traffic page, peek until it looks like a win, and ship noise. Do the analysis, do the research, prioritize ruthlessly, and test with discipline, and the conversion gains compound for years. If you'd rather a senior team run that loop on your funnel, that's exactly what our CRO & landing pages work is.