Work / Website conversion

Builder Funnel · 2026

Making the agency's own website book more calls.

Builder Funnel's site was turning 2.41% of sessions into new contacts, and the goal was 3%. Before touching copy, I wanted to know what was slowing people down and where the buttons actually were on a phone.

CROSite speedHubSpot CMSAnalytics
1.0s → 0.2sserver response on every page after the caching fix
3.36%sessions to new contacts in the four weeks after, vs a 2.41% baseline
1.4 → 2.9homepage discovery calls booked per week, first four weeks

TL;DR

  • A theme setting was keeping HubSpot from caching pages, so every page took 0.7 to 1.9 seconds just to start loading. One fix took that to about 0.2.
  • On a phone, the homepage's first call button sat about 960 pixels down. I added a three-step booking form and a phone-only call button near the top.
  • In the first four weeks, site conversion came in at 3.36% against a 2.41% baseline. It's still running, so these are early reads.

The problem

I measured the baseline first: 2.41% of sessions became new contacts from January through early October, counting every online source and leaving out offline and giveaway pages.

Then I audited 19 key pages. Phones were downloading hidden background videos. Mobile load time was over 4 seconds on 8 of them. Most had no way to convert above the fold on a phone, and the homepage form didn't give sales much to work with.

What I did

  1. Fixed caching. Theme modules were checking every request for the page editor, which blocked HubSpot's edge cache. After the fix, 32 of 32 scanned pages were cached.
  2. Stopped wasting phone data. Preloaded the hero images and stripped the background videos on mobile.
  3. Rebuilt the booking path. A three-step form with the calendar opening right in the card, first on the homepage, then on Contact and Results.
  4. Moved the button up. A phone-only "Book a Free Strategy Call" button, about 450 pixels down instead of 960.
  5. Fixed what was broken. Lead magnet checklists had been returning 404s since launch, and 21 redirects pointed nowhere.

Server response (time to first byte, 75th percentile)

Sep 11 - Oct 8
1,172 ms
Oct 12 - Nov 8
447 ms

Largest content loaded (75th percentile)

Sep 11 - Oct 8
2,145 ms
Oct 12 - Nov 8
888 ms
From Google's Chrome UX Report, which measures real visitors to the site. The Pricing page's lab mobile score also went from 61 to 77.

Early results

Site conversion was 3.36% for October 9 to November 9, and 3.56% through November 11. Homepage discovery call bookings went from 1.4 a week to 4.7 in the first two weeks, then settled at 2.9 a week over four.

Read the clean signal. Part of the early jump came from fewer junk paid-social sessions after an ad placement change, which shrinks the denominator. Non-paid traffic is the cleanest read on the site itself, and it went from 1.52% in September to 2.73%.

More reads are scheduled through early December for the phone button, the blog calls to action and a layout-shift fix.

What I'd do differently

I'd exclude my own traffic before measuring anything. My automated audits were showing up as direct sessions until I filtered my IP out.

I'd also run audits one or two at a time. A big parallel crawl got the site's firewall serving bot challenges, which made the next round of numbers useless.

Built withHubSpot CMSLighthouse + Chrome UX ReportPlaywrightPythonClaude Code