A med spa owner in Ridgewood called us last March with a complaint that sounded backwards. Her PageSpeed score had jumped from 71 to 98. Her booking form submissions had dropped 41% over the same two weeks. She wanted to know if we had broken something.
We had not broken anything. The previous agency had chased a 100/100 score by removing every element that scored as “render-blocking” — including her booking widget, her reviews carousel, and her before-and-after slider. The page was lightning fast and converted nothing.
The PSI score is a means, not the end
PageSpeed Insights measures how fast a page loads and how stable it feels. Those are real things that matter. But the score itself is a proxy for user experience, not a guarantee of one. A 100/100 page that nobody books from is a worse business asset than a 78/100 page that converts 4% of visitors.
Google does not pay your rent. Bookings do. Score-chasing without revenue context is one of the most expensive trap doors in technical SEO right now, and it’s getting more common because the score is easy to measure and conversion is not.
What “load-blocking” actually means
When PSI flags an element as render-blocking, it means the browser has to fetch and parse that resource before it can paint the page. Booking widgets, live chat, scheduling embeds, and review-pull scripts often fall into this category because they call out to third-party servers.
The fix is not always to remove them. Sometimes it is to defer them, lazy-load them, or replace the third-party version with a self-hosted alternative. The med spa’s previous agency took the shortcut — they deleted the elements entirely and high-fived over the 98 score.
The four elements that score badly but convert well
These are the usual suspects we see stripped out in score-chasing rebuilds. Each one is fixable without deleting it.
- Booking and scheduling widgets. Calendly, Acuity, Square Appointments, GoHighLevel embeds — these add 200ms to 800ms of load time but convert 6 to 12% of qualified traffic. Deferring them until after the fold or replacing the embed with a button-triggered modal keeps the conversion and recovers the score.
- Live chat widgets. Intercom, Drift, and Tidio scripts are render-blocking by default. The fix is loading them on first user interaction (scroll, click, mousemove) instead of on page load. Same conversion, no score penalty.
- Review carousels. Google review pulls, Trustpilot embeds, and Birdeye widgets fetch external data. They build trust at the bottom of the page where they belong. Lazy-load them and the score recovers.
- Before-and-after sliders. Heavy JavaScript libraries for image comparison. Convert beautifully for med spas, contractors, and dentists. The fix is using a CSS-only slider or loading the JS only when the image enters the viewport.
The trade-off framework we use
Before we touch any conversion element, we ask three questions. First, is this element above or below the fold? Above-fold elements deserve more optimization budget because they hit every visitor. Second, what is the conversion contribution? If the booking widget drives $48,000 in monthly revenue, we are not deleting it to gain four PSI points. Third, is there a deferred or lazy-loaded version that preserves function?
The med spa’s booking widget was the biggest revenue driver on the page. Killing it for a score gain made no sense. We re-added it with a defer attribute and a viewport-trigger, and the score landed at 91 instead of 98. Her bookings recovered within ten days.
What 91 versus 98 actually costs you
Almost nothing. Google’s ranking algorithm uses Core Web Vitals as a tiebreaker, not a primary factor. Once you are in the “good” range — LCP under 2.5s, INP under 200ms, CLS under 0.1 — additional gains do not move rankings meaningfully. The difference between 91 and 98 is invisible to Google in most competitive niches. The difference between 91 and a working booking flow is $14,000 a month.
How AJD handles this
We optimize for the “good” range, not for the 100. Our internal rule is hit 90+ on mobile and 95+ on desktop while preserving every revenue-touching element on the page. If a fix would require removing a conversion driver, we route around it with deferred loading, lazy-loading, or a self-hosted swap.
Before any speed work, we map the conversion contribution of every above-fold element. If a previous agency stripped your booking widget or reviews to chase the score, that is the first thing we put back — with a smarter loading pattern so the score holds up too.
The PSI score is a diagnostic tool, not a scoreboard. Whether you work with us or not, before you delete anything for a speed gain, look at what the element converts. Speed without bookings is a beautifully framed loss.
Book Free Discovery Call →
