Deloitte: 0.1s faster mobile speed lifted retail conversion 8.4%

Deloitte and Google-commissioned research firm Fifty-Five spent four weeks in late 2019 measuring what a fraction of a second of mobile page speed is worth to 37 brand sites. Their headline number, an 8.4% lift in retail conversion rate for a 0.1 second improvement, still gets quoted constantly. Shopify has since published its own, more recent read on the same relationship. Here is what each study actually measured, and where the two claims are Deloitte’s and Shopify’s findings, not ours.

What Deloitte actually measured#

Thirty-seven brands, one four-week window. Fifty-Five recruited over 70 retail, travel, luxury and lead-generation brands across Europe and the US, and 37 passed its analytics, technical and qualitative audits. From 28 October to 24 November 2019, it collected mobile site speed data through Google Lighthouse and combined it with each brand’s own web analytics: page views, bounce rate, funnel progression, conversion rate and average order value.

A regression model, not a controlled experiment. Deloitte and Fifty-Five built a logarithmic regression model to correlate four key mobile speed metrics with each site’s KPIs, requiring at least 95% statistical significance before a result was reported. Speed fluctuations were natural, arising from each site’s own performance work and traffic conditions, never artificially induced. The published 0.1 second figure is the combined, cumulative effect of all four speed metrics moving together, not one isolated change.

Different verticals, different KPIs. Retail and travel counted a completed transaction as conversion. Luxury counted an add-to-basket or “contact us” click. Lead generation counted a completed form. Retail conversion rose 8.4% and average order value rose 9.2%; travel conversion rose 10.1% with a smaller 1.9% average order value gain; luxury page views per session rose 8%; lead-generation informational pages saw an 8.3% bounce rate improvement.

Why a correlation study is not a universal speed rule#

This measures sites that were already live and already changing. Deloitte’s own limitations section is explicit: the 37 sites were “aggregated by vertical,” which “may not fully reflect an individual site’s product offering, design, or seasonality.” The study found a correlation between naturally occurring speed changes and business metrics on sites that were already fast enough to be running production ecommerce, not a controlled test of speed at every point on the scale.

It does not extend to desktop, or to every speed range. Deloitte says it looked for the same correlation on desktop and “found a lot of contradicting parameters,” attributing that to desktop users being less sensitive to load time than mobile users. Nothing in the study implies a 0.1 second gain is worth the same 8.4% at every starting speed; a site already loading in under a second and one loading in eight seconds are not shown to behave the same way for the same improvement.

Shopify’s newer figure is a similar shape of claim, not a replication. Shopify’s enterprise blog reports that across actively selling Shopify stores, every 100 milliseconds of slower load correlates with roughly 3.5% lower conversion, based on 28 days of Largest Contentful Paint data from around the turn of 2026. It also found stores with a 2.5 second LCP converting roughly 30% worse than stores at 1.5 seconds. Different platform, different metric, same broad conclusion: mobile speed and conversion move together, and the relationship looks steep at the low-millisecond end most publishers actually operate in.

Why we care#

The direction is well established; the exact multiplier is not portable. Two independent studies, six years apart and built on different site populations, both find meaningful conversion sensitivity to small mobile speed changes. That is a solid reason to treat Core Web Vitals work as revenue work, not just a ranking checkbox.

Quote the vertical, not just the headline number. “0.1 second equals 8.4% more conversions” is true only for Deloitte’s retail sample in 2019. Travel, luxury and lead-generation sites in the same study moved by different amounts on different metrics, and neither study claims the effect is linear at every speed level. Cite the study, the vertical and the metric together, or don’t cite the number at all.