Website performance benchmarks for heavy-asset B2B: why 90ms load time still loses deals
·6 min read
A 90ms page load sounds fast. In industrial B2B sectors like crane hire, maritime and energy, it is often still slow enough to kill an inquiry. This post breaks down real Core Web Vitals medians across nine verticals and shows exactly where performance gaps are costing firms work.
The number everyone quotes is wrong.
Firms in construction, energy and maritime often point to a Google PageSpeed score of 80-plus and call the job done. That score is not a load time. It is a weighted index of several signals, and a site can post a decent score while still delivering a 3.8-second Largest Contentful Paint to a buyer on a mobile connection in Houston or Singapore.
The number that matters is what your actual visitors experience. In heavy-asset B2B, those visitors are often a project director on site, a procurement lead traveling between offices, or a port operations manager checking suppliers from a vessel terminal. Their connection is not a London fibre line.
Median Core Web Vitals across nine verticals.
We audited the top 10 ranking sites for each of our nine target verticals using field data from Chrome User Experience Report (CrUX) and PageSpeed Insights. All figures are 75th-percentile field data, which is what Google uses for ranking signals.
Largest Contentful Paint (LCP) medians by sector
Sector
Median LCP (desktop)
Median LCP (mobile)
Construction
2.9s
5.1s
Working on a project?
If this resonated, tell us what you're building.
We'll read your brief and reply within one business day. No mailing list. No follow-ups unless you ask for them.
Most projects start with a short conversation. We listen to the brief, ask the right questions and tell you straight what we can handle and how. No retainer required to start.
Google's "good" threshold is 2.5 seconds. The majority of these sites fail on desktop and almost all fail on mobile.
Shipping and maritime and crane and heavy lift are the worst performers. Both sectors rely heavily on image-rich portfolios of equipment and project work. Those images are typically uncompressed, unformatted and loaded without lazy-loading or modern formats like WebP or AVIF.
What delay actually does to inquiry rates.
Google and Deloitte published research showing a 0.1-second improvement in mobile load time increases conversion rates by 8 percent for B2B sites. That sounds abstract until you attach a number to it.
A mid-size crane hire firm generating 40 qualified web inquiries per month, at an average project value of $180,000, is leaving roughly $576,000 in annual pipeline on the table for every full second of unnecessary load time. The math is not precise. The direction is.
The more specific effect is abandonment at the inquiry form. In industrial sectors, buyers who reach a contact or quote form and then leave are not coming back. They found a competitor. The form load time, not the homepage load time, is the conversion-critical moment. Most firms have never measured it separately.
Why commodity CDNs fail for global maritime and energy buyers.
A Content Delivery Network caches static assets on edge servers around the world. Most managed hosting platforms include a CDN in their base tier. The problem is cache hit rate.
For a shipping and maritime website or an energy and petrochemical website, buyers are often located in the Middle East, Southeast Asia, West Africa or the Gulf Coast. Commodity CDNs from shared hosting providers have thin edge coverage in those regions. A buyer in Abu Dhabi or Lagos may be pulling assets from a London or Frankfurt origin, adding 200-400ms of latency before a single byte of your page renders.
Enterprise CDNs (Cloudflare Enterprise, Fastly, AWS CloudFront with proper configuration) have genuine global coverage. The difference in time-to-first-byte between a commodity CDN and a properly configured enterprise network is often 300ms or more in these regions. That is enough to push an already marginal LCP score from borderline-pass to clear fail.
Next.js versus Webflow at scale.
This comparison comes up constantly among firms considering a rebuild. The honest answer is that both can deliver fast sites and both can deliver slow ones. The outcome depends on implementation, not the platform name.
That said, there are structural differences worth knowing.
Webflow renders pages server-side and serves them through Cloudflare's CDN. For a standard brochure site with limited dynamic content, a well-built Webflow site will hit sub-2.5s LCP consistently. The ceiling is lower when you add complex animations, large CMS collections or third-party embeds. Webflow does not give you server-side control, so optimizations like streaming HTML, edge caching logic or custom image pipelines are off the table.
Next.js gives you full control. A Next.js site built correctly, deployed on Vercel or a comparable edge-native platform, can achieve sub-1.5s LCP with server components, image optimization built in and route-level code splitting. The risk is that "built correctly" requires developer discipline. A poorly structured Next.js app with unoptimized images and client-side data fetching will perform worse than a default Webflow build.
For the crane and heavy lift and transport and logistics sectors, where sites carry large equipment catalogs and project galleries, Next.js with a proper image pipeline is the more defensible choice. For firms that need marketing teams to update content without developer involvement, Webflow with a strict build process is often the better trade.
The Interaction to Next Paint problem nobody talks about.
LCP gets the attention. INP (Interaction to Next Paint) is the metric that is quietly failing on most industrial sites.
INP measures how quickly a page responds after a user clicks or taps. Google moved INP into Core Web Vitals in March 2024, replacing First Input Delay. The "good" threshold is 200ms. The median for sites in our audit was 340ms on mobile.
For B2B sites, INP matters most on pages with quote calculators, filter-based project portfolios and multi-step contact forms. These are exactly the interactive elements that engineering, property development and logistics firms invest in. If the UI stutters after every click, buyers notice. They do not diagnose it as INP failure. They just form an impression that the firm is not on top of things.
What to do about it.
Three actions make the biggest measurable difference, in order of effort versus return.
First: fix your images. Convert all site images to WebP or AVIF. Set explicit width and height on every img element. Lazy-load everything below the fold. This alone typically improves LCP by 0.8-1.4 seconds on image-heavy industrial sites.
Second: audit your third-party scripts. Chat widgets, analytics tags, CRM trackers and video embeds each add blocking load time. Run a WebPageTest waterfall and identify every third-party call. Defer or remove anything that does not directly contribute to inquiry conversion.
Third: measure at the form, not the homepage. Your inquiry form page may be loading assets from a different CMS template, a different widget stack or a different CDN path than your homepage. Test it separately. Most firms discover their form page is 40-60 percent slower than their homepage and have never noticed.
For a fuller picture of what separates the top performers in each sector, the nine sector landing pages benchmarks post covers design and content patterns alongside performance signals. And if you want to understand what a properly instrumented launch looks like before you touch performance optimization, The Conversion-Instrumented Launch sets out the measurement framework.
Slow pages do not announce themselves. They just quietly hand inquiries to whoever loads faster.