Most people think CRO means A/B testing the button colour. It does not. The button-colour test is about the lowest-leverage thing a CRO program can run, and teams that start there tend to get stuck making tiny bets that never add up to anything.
Real CRO is about finding where you actually lose people and fixing that, in the right order. Here is how to do it.
What is a good conversion rate in 2026?
The number to anchor on is your vertical median. The cross-industry average hides too much variation to be useful.
| Vertical | 25th percentile | Median | 75th percentile |
|---|---|---|---|
| Apparel & Fashion | 1.6% | 2.4% | 3.6% |
| Beauty & Cosmetics | 1.9% | 2.8% | 4.1% |
| Consumer Electronics | 1.0% | 1.7% | 2.6% |
| Food & Beverage | 2.2% | 3.3% | 4.7% |
| Health & Supplements | 1.8% | 2.7% | 4.0% |
| Home & Furniture | 0.8% | 1.4% | 2.2% |
| Jewelry & Luxury | 0.6% | 1.0% | 1.7% |
| SaaS B2B | 2.4% | 4.0% | 6.5% |
| SaaS B2C | 3.0% | 4.8% | 7.4% |
| Lead Gen / Service | 4.0% | 6.5% | 10.0% |
The full breakdown by vertical and traffic source is in the Conversion Rate Benchmarker. Pick your vertical and main traffic source and read the cell that matches.
How to read your result: below the 25th percentile, you have structural problems and A/B testing is premature. Between the 25th and 50th, behaviour-driven help is your highest-leverage lever. Above the 75th, shift attention to order value and retention rather than conversion rate.
Where do conversions actually leak?
Most CRO writing treats the funnel as a single number. It is at least four, each with its own leak.
| Stage | What it measures | 2026 median | Most common cause of leak |
|---|---|---|---|
| Landing → engaged | Visitor scrolled, clicked, or stayed > 30s | 64% | Slow page, weak headline, irrelevant traffic |
| Engaged → cart-add (or signup-start) | Visitor reached the conversion-intent surface | 8.5% | PDP / signup form friction, missing trust signals |
| Cart → checkout | Visitor started checkout | 50% | Surprise costs, missing wallet pay, account-creation requirement |
| Checkout → completion | Visitor finished the order | 70% | Form length, payment friction, technical errors |
A site running 64% × 8.5% × 50% × 70% lands at about 1.9% blended. To reach 3.0%, you need several stages improving together: a few points on each sub-funnel compound into a 50%+ blended lift. So the first question for any CRO program is simple — which sub-funnel is worst relative to your vertical median? Fix that one first.
Which CRO levers actually move the needle?
Each lever below stands on its own; the usual playbook combines three to five. They are sorted by how much they move conversion, with rough effect sizes and how long they take to ship.
Tier 1 — the big movers
Behaviour-driven in-session help. The onsite conversion agent category. Average lift: 8-15% relative on a 30-day holdout. Time to ship: under an hour.
Cut shipping-cost surprises. Show shipping in the cart, not at checkout. Time to ship: 1-2 days.
Add wallet pay above the fold on mobile. Apple Pay, Google Pay, Shop Pay. Time to ship: 1-3 days.
Tier 2 — steady, worthwhile gains
Behaviour-driven product recommendations. Average lift: 2-4% relative. Time to ship: 1-2 weeks.
Visible trust signals near checkout. Reviews, security badges, return and shipping policy. Time to ship: 1 day.
Faster product pages, below 2.0s LCP. Average lift: 2-5% relative on mobile. Time to ship: 2-4 weeks.
Comparison content for high-AOV products. Side-by-side tables for products over $100. Average lift: 3-6% relative on the comparison-shopper subset. Time to ship: about a day per product.
Tier 3 — diminishing returns
A/B test CTA microcopy. Average lift: 0-1% relative. Useful only for mature, top-quartile sites.
A/B test button colours. Average lift: 0-0.5% relative. The lowest-leverage change in the playbook.
A/B test hero image variants. Average lift: 0-2% relative. Rarely worth the testing cost on sub-100k-session sites.
The rule of thumb: spend about 80% of your CRO time on Tier 1, 20% on Tier 2, and none on Tier 3 unless you are a top-quartile store that has run out of bigger moves.
How do you run a 90-day CRO program?
Six phases over 90 days.
Days 1-7 — diagnose. Pull your GA4 funnel data, add Microsoft Clarity for session replays, and find the worst-leaking sub-funnel relative to your vertical median. Output: a one-page diagnosis with three observations and three hypotheses.
Days 8-21 — ship the in-session layer. Install Yokaify or build the equivalent (see the onsite conversion agent guide). Let it gather 14 days of data.
Days 22-35 — ship the highest-leverage Tier 1 fix. Usually shipping-cost transparency, wallet pay on mobile, or a shorter form. One change, measured against a holdout.
Days 36-50 — ship the second Tier 1 fix. With the first lever now sitting on two weeks of data, layer the next change on a fresh holdout so you can isolate its effect.
Days 51-70 — add the Tier 2 layer. Product recommendations, trust signals near checkout, page-speed work. These are not one-day ships; budget two to three weeks to build and monitor.
Days 71-90 — read the lift. Pull the holdout data across every change, compute the blended lift, and write down what worked, what did not, and what surprised you. Then set the next 90 days.
By day 90, a typical store moves from below-median conversion to median or better.
How has AI search changed who lands on your site?
A 2026 wrinkle most CRO playbooks skip: the visitors arriving on your site are different now. AI Overviews and chat assistants answer a lot of pre-purchase questions before anyone clicks through, so the people who do arrive are a more self-selected, higher-intent group than they were in 2022.
Two consequences follow.
Top-of-funnel content has been compressed. Visitors landing on your product page have often already read an AI summary of it. Repeating what they just read costs you attention; moving straight into "here is what matters for your situation" earns the click.
And the traffic is fewer-but-better. Calibrate your benchmarks against 2026 data, not 2022 data — old benchmarks will flatter you, telling you that you are outperforming when you are actually below the new median.
When is CRO the wrong investment?
CRO is the highest-margin work most teams can do — but not always, and not first. There are cases where the money should go elsewhere.
Further reading
- ToolConversion rate benchmarkerWhere your CVR sits relative to the 2026 vertical × source percentiles.
- GuideThe Onsite Conversion Agent: A 2026 Field GuideThe category that produces the tier-1 in-session lift.
- GuideThe cart-recovery field referenceThe deep dive on the highest-leakage stage of the funnel.
- GlossaryAverage Order Value (definition)AOV decomposed.
Frequently asked questions
CRO is the discipline of increasing the percentage of visitors who complete a desired action without increasing traffic. The work is a mix of qualitative research, quantitative analysis, and structured experiments against a holdout.
Last updated May 31, 2026.