Last month, CRO consultant Ana Kostic shared a case study that ricocheted around marketing Twitter: her client lost 40% of revenue after implementing a widely-cited 'best practice' pattern. It's the reminder every strategist needs heading into Q4 — best practice is a hypothesis, not a decision. The only thing that tells you what works on your site is a test.
Add to that the seismic changes happening around search — Google testing forced sign-in walls for extended results, Microsoft Clarity now surfacing branded and non-branded AI queries, and AI visibility becoming a real signal in PPC performance — and the pool of visitors landing on your site in August 2026 is materially different from a year ago. They arrive later in the funnel, with more context, and less patience for friction.
Below are 15 conversion rate optimisation tests we're actively running for clients across the US, UK, and EU right now. Each has moved the needle on at least one account in the last 12 months. Not all will win for you — that's the point.
Why CRO Looks Different in 2026
Three shifts matter before you queue up a single experiment:
- Traffic is more qualified but scarcer. AI answer engines pre-filter tyre-kickers. The people who click through have usually seen a summary of your offer already.
- Intent signals are richer. With tools like Clarity segmenting AI-sourced sessions, you can now A/B test against traffic that arrived via ChatGPT versus organic search versus paid.
- First-party data is the new competitive edge. Google's Data Manager API expansion and the death of third-party cookies mean your on-site conversion isn't just a sale — it's a data capture event that feeds every downstream channel.
If your CRO programme still optimises for a generic 'add to cart' rate, you're leaving strategy on the table. Pair this piece with our broader conversion rate optimisation guide for the underlying framework.
The 15 Tests Worth Running
1. Replace hero copy with the exact query your AI traffic is asking
Pull the top branded and non-branded AI queries from Clarity. If people are arriving because ChatGPT told them you 'offer flat-rate pricing for SaaS onboarding,' put that phrase in your H1. Message match is worth more than clever copy.
2. Test a single primary CTA vs. two CTAs of equal weight
The 'best practice' says one CTA. The reality is nuanced — for high-consideration B2B, offering a 'Talk to sales' and 'See pricing' side by side often outperforms funnelling everyone into a demo request.
3. Move social proof above the fold — but be specific
Generic '5,000 happy customers' badges are wallpaper. Test named logos with a one-line outcome ('Cut CAC 34% in 90 days'). Our social proof marketing guide details which formats hold up.
4. Sticky mobile CTA vs. inline CTAs only
Mobile bounce spikes at scroll depth 60–70% on most sites we audit. A sticky bottom-bar CTA recovers 8–15% of would-be exits. Test it against traffic segments — it often loses on desktop.
5. Long-form product pages vs. accordion-collapsed pages
Same content, different scroll behaviour. Accordions win for returning visitors; long-form wins for first-touch. Segment your test by traffic source.
6. Test form length reduction one field at a time
Not 'six fields to three' — one field per test. You need to know which field is the killer. Phone number is usually it.
7. Price transparency vs. 'Request a quote'
The uncomfortable truth: showing starting prices often lifts qualified leads by 20–30%, even for premium services. It filters out mismatches before your sales team wastes time.
8. Add a comparison table for AI-summarisable content
AI crawlers love structured comparisons. A well-built table lifts both on-page conversion and the odds your page gets cited in AI answers. Combine with proper schema markup.
9. Personalise the hero by referral source
Traffic from LinkedIn Ads sees a B2B-framed hero; traffic from Google Shopping sees a product-framed hero. Dynamic content is now trivial to implement with modern CMS platforms.
10. Test 'no sign-in required' friction points
With Google testing forced sign-in for extended search results, users are already fatigued by gates. Sites that let users get value before capture (interactive calculators, instant quotes) are outperforming gated equivalents.
11. Trust bar directly beneath the CTA
Money-back guarantees, security badges, and 'no credit card required' sit better under the button than above it. Test placement, not existence.
12. Video hero vs. static image with faster load
Video hero pages often lose on mobile due to Core Web Vitals penalties compounding with lower engagement on autoplay. Static + fast beats video + slow. See our Core Web Vitals guide.
13. Exit-intent offer differentiation
Test the offer, not the popup mechanic. A 10% discount, a free consultation, and a downloadable checklist all convert wildly differently by industry.
14. Reduce navigation options on landing pages
PPC landing pages with full site nav leak 15–25% of clicks to non-converting pages. Test a stripped nav or footer-only navigation. Detailed patterns in our PPC landing page guide.
15. Test the thank-you page as a conversion asset
Most sites treat the confirmation page as an afterthought. Test upsells, referral prompts, calendar bookings, or content offers. It's the highest-intent moment on the entire site.
Client Case Study: A B2B SaaS Pricing Page Rebuild
A mid-market HR software client came to us in Q1 2026 with a 1.8% pricing-page-to-demo conversion rate. The 'best practice' recommendation from their previous agency had been to hide pricing entirely behind a 'Contact sales' CTA — the pattern Ana Kostic warned about.
We ran a three-arm test over six weeks:
- Arm A: Existing gated pricing page (control)
- Arm B: Starting price displayed, single 'Book demo' CTA
- Arm C: Full three-tier pricing table, two CTAs ('Start free trial' and 'Book demo'), FAQ accordion below
Arm C won decisively. Demo requests rose 61%, and — critically — sales-qualified lead rate on those demos rose from 34% to 52%. Fewer wasted sales conversations, better close rates. Revenue attributed to the pricing page climbed 138% within 90 days. The 'best practice' of hiding pricing had been costing them roughly £40K/month in pipeline.
What Most Businesses Get Wrong
After running hundreds of tests, the same five mistakes come up:
- Testing without a hypothesis. 'Let's try a green button' isn't a test — it's a coin flip. Every experiment should state: what you believe, why, and what result would disprove it.
- Calling tests too early. Statistical significance at 95% confidence with 200 conversions is not the same as a durable win. Watch for two full business cycles.
- Optimising the wrong metric. Lifting click-through on a CTA while tanking downstream conversion is a loss dressed as a win. Track to revenue, not to micro-conversions.
- Copying competitors. Their audience, offer, and traffic mix are not yours. Their winning pattern may be your losing one.
- Ignoring qualitative data. Session recordings and user interviews generate better hypotheses than any analytics dashboard. Clarity is free — use it.
Building a Testing Cadence That Compounds
One test in isolation is a data point. A programme is what moves revenue. We recommend clients aim for:
- 2–4 concurrent tests on non-overlapping page templates
- Two-week minimum test duration, four weeks preferred
- Documented learnings library — even losing tests are worth their weight, provided you record what didn't work and why
- Quarterly review of test velocity, win rate, and revenue impact
The clients who compound CRO gains aren't running exotic tests — they're running boring ones, consistently, with discipline. Pair this with a healthy first-party data strategy and every test also strengthens your audience data for paid channels.
Ready to Prioritise Your Testing Roadmap?
If you're sitting on traffic that isn't converting, or you've plateaued at a conversion rate that hasn't moved in six months, we can help. Our team runs CRO programmes for brands across the US, UK, and EU — and we start every engagement with a diagnostic, not a template.
Build a proposal with our team and we'll map the three highest-impact tests we'd run on your site in the first 60 days.