valentinyeo.com: CRO results
2026-10-15 · 10 results · about 5 min read
Your site loses bookings to missing proof, a hidden price and an unclear call, and nothing measures bookings yet. Five experiments below fix and check that, starting with tracking.
Next: Say yes to E1 (booking tracking). The other four start once it runs.
All results
| # | Skill run | Score | Result |
|---|---|---|---|
| 1 | Site audit | 6.1/10 | 61/100: weak copy, trust and measurement; speed is fine. |
| 2 | Page speed | 10/10 | Fast: every page shows its main content in under 1 second on a throttled phone, so speed is not what costs bookings. |
| 3 | Tracking check | 1/10 | Nothing counts visits or bookings; the only script is an A/B test tool that cannot finish a test at your volume. |
| 4 | Test readiness | no score | Not testable yet: about 3 bookings a month is far too few for A/B tests, so we fix, research and measure before and after. |
| 5 | Persona walk-throughs | 5/10 | One buyer found the booking page after 15 steps; the other nearly left and never saw a price. |
| 6 | Skeptic review | no score | 14 of 15 findings held up; 11 needed rewording and 1 was dropped. |
| 7 | Synthetic visitor swarm | 0.8/10 | 2 of 25 AI buyers reached your booking page; 4 problems were confirmed on screenshots. |
| 8 | Fix wireframe | no score | Four changes put the free call, the price and the inne result on the first phone screen. |
| 9 | Client deck | no score | 14 slides for you as the client: score, readiness, tracking, 5 findings, before and after, 90-day plan. |
| 10 | Message test (MiroFish) | no score | Did not finish: MiroFish hung in round 1, and it simulated your client companies instead of buyers. |
Result 1 of 10 · done
Site audit
How well do the key pages convert, page by page, across UX, copy, trust, visual and tracking?
61/100: weak copy, trust and measurement; speed is fine.
- C-1: No analytics or booking tracking
- H-1: Contact, case study and blog pages have no Book a Call
- H-2: The case studies are hidden
- H-3: Book a Call says nothing about the call
- H-4: Price is about 17 phone screens down
- H-5: Home says no retainers, service offers one
- Critical: no analytics, so no booking can be traced to a page or LinkedIn post
- High: the case studies (inne +14.6%) are commented out of the home page
- High: the price is about 17 phone screens down; the service page shows none
- High: Book a Call never says free, 30 minutes or no obligation
- High: home says no retainers, the service page offers a monthly retainer
Result 2 of 10 · done
Page speed
How fast do the key pages load on a throttled phone, against Google's Core Web Vitals limits?
Fast: every page shows its main content in under 1 second on a throttled phone, so speed is not what costs bookings.
Nothing left to fix on this test.
- blog: LCP 0.9 s (FAST), CLS 0.000 (FAST), TBT 115 ms (FAST), 5 requests
- casestudy: LCP 0.9 s (FAST), CLS 0.000 (FAST), TBT 153 ms (FAST), 4 requests
- contact: LCP 0.8 s (FAST), CLS 0.000 (FAST), TBT 34 ms (FAST), 7 requests
- home: LCP 0.8 s (FAST), CLS 0.000 (FAST), TBT 19 ms (FAST), 7 requests
- service: LCP 0.9 s (FAST), CLS 0.000 (FAST), TBT 123 ms (FAST), 2 requests
Result 3 of 10 · done
Tracking check
What measures visits and conversions today, is anything doubled or broken, and what access do we need?
Nothing counts visits or bookings; the only script is an A/B test tool that cannot finish a test at your volume.
- C-1: No analytics or booking tracking
- M-1: Headline test cannot finish
- No analytics on any page (Google Analytics, Hotjar and PostHog are listed under Tools I Use but not installed)
- Endless Testing runs a headline test on home, contact and blog
- Book a Call opens Google Calendar, which never learns which page sent the visitor
- Cloudflare counts about 68,000 page views in 30 days, mostly bots: an edge count, not people
- Still needed: Search Console access and your booked calls per month
Result 4 of 10 · done
Test readiness
Are there enough conversions to run A/B tests, or is research the right method?
Not testable yet: about 3 bookings a month is far too few for A/B tests, so we fix, research and measure before and after.
- Inputs: 3 conversions/month, CVR 0.50%, ~600 visitors/month, traffic share 1
Result 5 of 10 · done
Persona walk-throughs
Can a realistic buyer reach the main conversion in a real browser, and where do they struggle?
One buyer found the booking page after 15 steps; the other nearly left and never saw a price.
- Booking link on the contact page (the head of product lost 4 steps there)
- Price on the service page and the first phone screen (the store founder never found one)
- Head of product (desktop), judging whether you can run parallel tests: reached the booking page in 15 steps, after a detour to a contact page with no booking link
- Online store founder (phone), looking for a price and proof: reached the contact page in 7 steps, nearly left, never saw a price
Result 6 of 10 · done
Skeptic review
Does every finding survive an adversarial re-check against screenshots and the live site?
14 of 15 findings held up; 11 needed rewording and 1 was dropped.
- Corrected: the price is 17 phone screens down, not 12.5
- Corrected: all three case studies are hidden, not only inne
- Dropped: the Tools I Use list, which no visitor would notice
Result 7 of 10 · done
Synthetic visitor swarm
When many AI buyers browse the real site, how far do they get and what stops them?
2 of 25 AI buyers reached your booking page; 4 problems were confirmed on screenshots.
- 22 visitors: 'Increase testing speed by 10x' / '10x more tests' claim has no baseline, method or example
- 14 visitors: Testimonials give praise but no metrics, baselines, timeframe or test details
- 15 visitors: Hero headline and subtext are buzzword-heavy ('AI-driven', 'full potential') with no concrete deliverable
- 10 visitors: No price or price range shown for any engagement or the Book a Call path
- 22 of 25 asked what the 10x claim is compared to
- 14 asked for numbers behind the testimonials
- 10 could not find a price
- 6 could not tell if the logos are clients
- Dropped: a blank booking page (our screenshot fired too early; fixed)
Result 8 of 10 · done
Fix wireframe
What would the fixed page look like, change by change, tied to the findings?
Four changes put the free call, the price and the inne result on the first phone screen.
Result 9 of 10 · done
Client deck
Can the owner understand the results and decide in one sitting?
14 slides for you as the client: score, readiness, tracking, 5 findings, before and after, 90-day plan.
- Slide types: appendix, before-after, bottom-line, finding, next-steps, roadmap, score, strengths, test-readiness, tracking
Result 10 of 10 · failed
Message test (MiroFish)
Which offer, price or headline does a simulated crowd of buyers prefer, and why?
Did not finish: MiroFish hung in round 1, and it simulated your client companies instead of buyers.
- Built its knowledge graph in 3 minutes; agents were posting after 4
- Then no activity for 50 minutes, so the run was stopped
- It mostly repeats what the visitor swarm does; parked until you decide
Experiments: what we test next (5)
Every result above feeds these. Each one says what we believe, which results show the problem, how to run it at this site's traffic, and when to call it.
E1 · Critical · Setup, then baseline
Count every Book a Call click and booked call
Because nothing measures visits or bookings today, we believe adding click and booking tracking will give a baseline that every later change can be judged against.
3. Tracking check, 4. Test readiness, 1. Site audit
- Turn on Cloudflare Web Analytics for the Pages project (cookieless, no banner needed)
- Fire an event on every Book a Call click with the page name, and add ?src=<page> to the calendar link
- Each week, write down booked calls from Google Calendar
- Pause the Endless Testing headline test so all visitors see one version
E2 · High · Ship + before/after, checked by synthetic visitors
First screen: free 30-min call, price and the inne result
Because buyers could not find a price, doubted the claims and did not know what the call is, we believe showing the free 30-minute call, the day rate and the +14.6% inne result on the first screen will raise Book a Call clicks for product and growth leads.
1. Site audit, 7. Synthetic visitor swarm, 5. Persona walk-throughs, 8. Fix wireframe
- Build the wireframe as a real page on a Cloudflare preview link
- Send 25 synthetic visitors to today's page and 25 to the preview; compare how many reach booking and what stops them
- Show the preview to 5 product or growth leads on a call and note where they hesitate
- Ship it and log the date
E3 · High · Synthetic A/B on two previews, then interviews
Day rate only vs a monthly option
Because home says no retainers while the service page offers one, we believe one clear price format will remove a reason to hesitate; we do not know yet which format buyers prefer.
1. Site audit, 6. Skeptic review, 7. Synthetic visitor swarm
- You choose the monthly price and terms
- Publish two previews: A with day rate only, B with a monthly option
- Send 25 synthetic buyers to each and compare objections about price
- Ask the 5 interviewees from E2 which they would rather approve
E4 · Medium · Ship + before/after
Proof with numbers instead of claims
Because 22 of 25 visitors doubted the 10x claim and 14 asked for numbers behind testimonials, we believe replacing the claim with a measured example and adding one number to each testimonial will make buyers trust the offer enough to book.
7. Synthetic visitor swarm, 1. Site audit
- Ask each testimonial author for one number (tests a month, lift, time saved)
- Replace '10x' with a measured example from a named client, or remove it
- Caption the logo wall and link the logos that have a case study
- Rerun 25 synthetic visitors and count trust objections
E5 · Medium · Ship + before/after
Book a Call on every page
Because the contact, case study and blog pages have no booking link, we believe adding one at the top of contact and the end of each case study and post will turn more readers into calls.
1. Site audit, 5. Persona walk-throughs, 6. Skeptic review
- Add the same Book a Call button to contact, each case study and each blog post
- Add a sticky phone button on long pages (the style already exists in styles.css)
- Tag each with its page name (from E1)
How we verify results at this traffic
Too few conversions for an A/B test means no single number proves a change. We stack weaker checks that point the same way:
- Rescore before and after: rerun the same tests (audit, 25 AI visitors, the two buyer walk-throughs) on the new page. The scores on this page must go up. Free or cheap, done in a day, but it shows the page got clearer, not that it sells more.
- Five real buyers on a call: show the new page to 5 product or growth leads and note where they hesitate. Five people find most usability problems.
- Count clicks, not only bookings: Book a Call clicks happen far more often than bookings, so a change shows up in clicks within weeks (needs E1 tracking first).
- Before and after on bookings: 4 weeks before vs 4 weeks after, with LinkedIn posting kept steady. At about 3 bookings a month only a large change is visible, so we treat it as a sanity check, not proof.
- Ask every new lead: one question on the booking form, "What made you book?". It tells us which change people noticed.
How these results were made
Each section is one skill run on the same site, shown the same way. Numbers come straight from the run files; the one-line results are the reviewer's reading of them. Simulated visitors and simulated opinion are early signals, not measured conversion results.
Made with the free claude-cro pack.
Want an expert to run these experiments for you? Book a free 20-minute teardown.