The Evidence That
Speed Costs You Money

Portent analyzed conversion data across B2B and ecommerce sites, finding a consistent 4–7% relative decline in conversion rate for each additional second of page load time. The effect is steeper at lower load times (1–3 seconds) and flattens somewhat for very slow sites where only high-intent users remain.

This is the primary source for our 4% conversion penalty rate. We deliberately use the bottom of the 4–7% range to produce conservative estimates.

Deloitte conducted this study across 37 retail and travel sites for Google, measuring the impact of site speed improvements on mobile conversion. The finding that a 0.1-second improvement produces an 8% conversion lift is one of the most cited figures in the performance-to-revenue literature.

Critically, the study controlled for other variables — the speed improvement was the isolated change. This strengthens the causal claim more than correlational studies. It also found that speed improvements affected average order value, not just conversion rate — suggesting speed builds trust and reduces purchase anxiety.

Analysed data from 900 million user sessions on retail sites. Top-quartile sites loaded in approximately 2 seconds; bottom quartile averaged 5 seconds. The conversion gap between these groups was substantial and consistent across verticals.

While this study is from 2017, subsequent research has consistently confirmed its directional findings. User expectations for speed have increased since then, meaning the penalty for slow sites has likely grown, not shrunk.

Google analysed millions of mobile landing page sessions and found a steep, consistent relationship between load time and bounce rate. The relationship is non-linear — the first additional second of latency causes proportionally more damage than later seconds.

This study is the basis for our bounce rate model, where we estimate the fraction of users who leave before any content loads. For ad-supported businesses, this bounce directly translates to wasted spend.

Google's official documentation on TTFB as a performance metric establishes the thresholds used in Chrome's User Experience Report and PageSpeed Insights. Sites in the "Poor" TTFB range cannot achieve a "Good" LCP score, which feeds into the Page Experience ranking signal.

Searchmetrics analysed ranking data before and after Google's Core Web Vitals ranking update. Sites in the top quartile of CWV scores held significantly better rankings than equivalent-content sites in the bottom quartile. The LCP correlation was strongest — and LCP is the CWV metric most directly influenced by TTFB.

The SEO impact of slow hosting is a compounding factor our revenue calculator does not include. Sites penalised in search rankings receive less organic traffic, reducing the total revenue base — meaning our estimates undercount the full cost of slow hosting.