Mobile conversion rate drops at specific page speed thresholds, not continuously as speed worsens. LCP (Largest Contentful Paint) above 2.5 seconds produces measurable conversion loss, above 4 seconds the loss becomes severe, and above 6 seconds conversion collapses to roughly half of the fast-page baseline. INP (Interaction to Next Paint) above 200ms adds additional loss on top because the page feels unresponsive even after it loaded. The thresholds are not Google's Core Web Vitals targets — Google's targets are set for crawlability, not conversion. The conversion thresholds are tighter and worth measuring separately because that is where revenue actually lives.
Mobile conversion drops at LCP > 2.5s, drops sharply > 4s, collapses > 6s. INP > 200ms adds additional loss. Conversion thresholds are tighter than Google's Core Web Vitals targets. Measure field data, not lab data, and benchmark against the conversion thresholds.
What is the actual LCP threshold where mobile conversion starts dropping?
Mobile conversion starts dropping measurably at LCP above 2.5 seconds. Below 2.5s, speed has minimal independent effect on conversion — the page is fast enough that the user does not perceive wait, and conversion is governed by other factors (content, offer, trust). Between 2.5s and 4s, conversion typically drops 10-20 percent compared to the sub-2.5s baseline. Above 4s, conversion drops 25-40 percent. Above 6s, conversion collapses to roughly half the fast-page baseline — many users bounce before the primary content paints. The thresholds are approximate but consistent across categories because they reflect human attention tolerance for wait rather than any technical property of the page.
Why is INP above 200ms a separate conversion-loss layer?
INP above 200ms produces its own conversion loss layer because it measures how long the page takes to respond to user interaction, independent of how long the initial load took. A page that loads in 2 seconds (good LCP) but takes 500ms to respond to a button tap (bad INP) feels broken to the user even though visually it looked fast. The user taps, sees no response, taps again, the second tap registers as a new interaction, the page misbehaves, the user bounces. INP loss is multiplicative with LCP loss — a fast-loading page with slow interaction loses as much conversion as a slow-loading page with fast interaction, and a page bad on both can lose 40-60 percent of conversion compared to a page good on both. The two metrics are independent and both need to be measured and optimised.
Why are Google's Core Web Vitals targets looser than the conversion thresholds?
Google's Core Web Vitals targets (LCP under 2.5s is 'good,' 2.5-4s is 'needs improvement,' above 4s is 'poor') are set for ranking and crawlability, not conversion. Google wants pages that are usable enough to deserve ranking position; the targets separate acceptable from unacceptable. Conversion is a different question: within the 'good' Google band, pages at 1.5s LCP convert noticeably better than pages at 2.3s LCP, even though both pass Google's threshold. The conversion curve is continuous within the Google 'good' band and steps down sharply above it. Teams that treat Google's thresholds as the goal often leave 10-15 percent of conversion on the table because they stopped optimising at the ranking threshold rather than the conversion threshold. Mobile drives over 80 percent of landing page traffic, which makes the gap between the two thresholds materially consequential.
What are the five most common causes of LCP above the conversion threshold?
1) Unoptimised hero images — PNG or uncompressed JPEG at desktop resolution served to mobile devices; typically adds 1-2s to LCP. 2) Blocking third-party scripts in the head — analytics, tag managers, consent platforms loaded synchronously before the page can render; adds 500ms-2s. 3) Server response time above 500ms — slow TTFB from an under-provisioned backend or a non-CDN origin; adds 300ms-1s. 4) Font loading without font-display: swap — the browser waits for custom fonts before painting text; adds 200-800ms. 5) Full-width hero video autoplaying from an uncompressed source — adds 1-3s to LCP depending on video size. The five causes together typically account for 80 percent of LCP problems in Advize-audited DTC landing pages.
How do you measure LCP and INP on actual live traffic?
1) Install the web-vitals library or use the Chrome User Experience Report (CrUX) for real-user monitoring. 2) Report LCP and INP at the 75th percentile — the slowest 25 percent of users is what determines conversion loss, not the median. 3) Segment by mobile vs desktop (the targets and thresholds are different). 4) Segment by device tier (low-end Android devices perform worse than iPhones at the same page). 5) Benchmark against the conversion thresholds: p75 mobile LCP under 2.5s and INP under 200ms is the floor. 6) Lab tests (Lighthouse, PageSpeed Insights) are useful for debugging but do not represent what real users experience — always benchmark against field data. 7) Rerun monthly as code and third-party dependencies evolve.
What is the most impactful first fix if page speed is the primary conversion bottleneck?
The highest-leverage first fix for mobile LCP is hero image optimisation. Most DTC and B2B landing pages serve a hero image that is 2-5x larger than it needs to be for mobile — a 2000px wide image on a 400px wide viewport, uncompressed. Replacing the hero image with a mobile-optimised version (responsive srcset, WebP format, compressed to around 100-200KB) typically shaves 800ms-1.5s off LCP and produces immediate measurable conversion lift. The fix takes an hour to implement and does not require backend changes. In Advize's experience, hero image optimisation produces more conversion lift per hour of engineering time than any other page speed intervention and should be the first lever before anything more complex is attempted.
Conclusion
Page speed is a conversion lever at specific thresholds, not a continuous optimisation. LCP under 2.5s and INP under 200ms is where mobile conversion operates at full strength. Above those thresholds, measurable conversion loss accumulates and crosses structural failure zones at 4s and 6s. The practical next step for any team with mobile conversion below 2.5 percent is to measure LCP and INP on live traffic (not lab tests) and benchmark against the thresholds. In Advize's experience, roughly a third of underperforming mobile pages are losing conversion to page speed thresholds the team has not measured directly.