One tenth of a second. That is the speed improvement Google and Deloitte isolated in their 2020 study “Milliseconds Make Millions”, and it moved real money across the 37 retail, travel, luxury and lead generation brands they measured.
A 0.1 second improvement in mobile site speed lifted retail conversions by 8.4% and average order values by 9.2%. Travel conversions rose 10.1% (Deloitte and Google, 2020).
Not a redesign. Not a replatform. A tenth of a second, worth more than most ad copy tests I have ever run.
I have managed over $8M in ad spend across 22 years, and page speed is still the audit item most likely to be skipped by the people paying for the clicks. Advertisers will rebuild their campaign structure three times a year and never once open a performance waterfall for the page the ads land on.
A slow page taxes every click you buy
The arithmetic is unforgiving. You pay for the click at the auction, but you collect after the page loads. Everything that happens between those two moments is leakage, and you fund that leakage at full CPC.
A landing page that suppresses conversion rate does not show up as a problem anywhere in your ad account. The auction does not care. Your cost per acquisition simply runs higher than it should, on every campaign, every day, while the reporting blames your bids, your audiences or your creative.
The auction does eventually notice, though, in the worst way. Landing page experience is a documented component of how Google grades ad quality, so a page that frustrates visitors can quietly raise the price you pay per click on top of suppressing what happens after it. You get charged twice for the same slow page.
This is why I audit landing pages before scaling spend, never after. Scaling multiplies whatever efficiency you already have. If the page leaks, you are paying to scale the leak.
The numbers are old. The behavior is not.
The most quoted statistic in this space comes from Google’s mobile speed research: 53% of mobile site visits are likely to be abandoned if a page takes longer than three seconds to load (Google, 2016).
Be honest about that number. It was published in September 2016 and turns ten years old this autumn, measured on 2016 phones over 2016 networks. Anyone presenting it as fresh research is not reading their own sources.
I still use it, for a reason. Devices and networks got faster while pages got heavier at nearly the same rate, and user patience has moved in only one direction. Google’s follow-up research with SOASTA in 2017 found that as mobile load time stretches from one second to seven, the probability of a visitor bouncing increases 113% (Google and SOASTA, 2017).
The direction and rough magnitude of these findings have been replicated enough times, across enough verticals, that the burden of proof now sits with anyone claiming their slow page is fine.
INP is the speed number worth watching now
Load time is only half the experience. The other half is what happens when a visitor actually taps something, and that is what Interaction to Next Paint (INP) measures.
INP became an official Core Web Vital on March 12, 2024, replacing First Input Delay (web.dev, 2024). Google’s documented threshold for good responsiveness is an INP of 200 milliseconds or less (web.dev).
The difference matters for buyers. First Input Delay graded only the first interaction on a page; INP grades all of them. On an e-commerce template that means the variant selector, the quantity stepper, the add-to-cart button and the coupon field. A page can load fast and still feel broken at exactly the moment someone tries to give you money.
In audits, the usual INP culprits are the same ones behind slow loads: too much JavaScript fighting for the main thread, most of it belonging to tags and widgets nobody remembers installing. Which is convenient, because it means one cleanup usually improves both numbers at once.
My take: treat Core Web Vitals as a conversion diagnostic first and a ranking factor a distant second. The SEO industry argues endlessly about ranking weight, and I mostly do not care. The same 200 milliseconds that keeps a URL green in Search Console keeps a buyer from rage-tapping your checkout button.
What I check before I scale a budget
A recent example, anonymized. I was asked to scale an e-commerce account that had flatlined. Before touching a single bid I ran the landing pages through PageSpeed Insights, and the field data told the story: mobile product pages failed every Core Web Vitals threshold, dragged down by oversized images and a wall of third-party scripts. The pages got fixed before the budget moved, and only then did the account respond to extra spend the way the projections said it should.
The pre-scaling checklist is short:
- Field data over lab data. The Chrome UX Report numbers inside PageSpeed Insights reflect real visitors on real devices, not your office connection.
- Test the actual landing pages on mobile, including the tracking parameters your ads append. Redirect chains and parameter-triggered scripts add weight that a bare-URL test never sees.
- Check INP on the templates where money changes hands: product page, cart, checkout, lead form.
- Cut or defer third-party scripts. Tag bloat is the most common speed problem I find in audits, and the marketing team usually owns it.
- Re-test after every promo season. Regressions arrive with every new pixel, popup and testing snippet.
Speed work is not glamorous, and nobody screenshots it for a case study. But agencies will happily charge five figures to test button colors on a product page that takes four seconds to become usable. Fix the four seconds first. Halfway through 2026, it is still the cheapest conversion lift you can buy.