Site speed and sales: what the numbers say

Tell someone their site is slow and you usually hear "it loads in three seconds, that is fine". The trouble is that three seconds sounds harmless until you convert it into money. Let us do that.

What the research shows

• Amazon measured that 100 ms of extra latency cost roughly 1% in sales. At their scale that was over a billion dollars a year.
• Walmart recorded a roughly 2% conversion increase for every second of load time removed.
• Google''s analysis of mobile pages found bounce probability rises 32% when load time goes from 1 to 3 seconds, and 90% when it reaches 5 seconds.
• A Deloitte retail study found that a mere 0.1 second improvement lifted conversions by around 8% on average.

The figures vary by industry, but the direction is always the same and surprisingly consistent.

Calculate your own loss

Take a conservative estimate: 7% conversion drop per 100 ms. A small shop example:

• Monthly traffic: 8,000
• Conversion: 1.5% → 120 orders
• Average basket: €38 → €4,560 per month

Cut 500 ms off load time and, conservatively, you gain around 15-20% more orders — roughly €680 to €910 a month. Hosting that makes this possible costs €16. The ratio is not remotely close.

Where the seconds actually go

On a typical WordPress site the breakdown looks like this:

• Server response (TTFB): 300-1200 ms. The largest single slice and the least visible. Driven by hosting, PHP version and caching.
• Images: 1-4 seconds. Uncompressed photos and no modern formats.
• Scripts and plugins: 0.5-2 seconds. Sliders, chat widgets, three separate analytics tools.
• Fonts: 100-400 ms. Five weights of a family where two would do.

Order of work by return on effort

Hosting and caching first — the fastest, largest gain. Then images (WebP, correct dimensions, lazy loading). Then plugin cleanup. Only then fine-grained code optimisation, which usually buys you tenths.

How to measure honestly

PageSpeed Insights gives both lab and field data — trust the field data (CrUX), because it reflects what real users experience. Test the mobile version, since that is where most traffic and the worst conditions are. And measure at least three times, at different hours.

When hosting is the bottleneck

If TTFB is above 600 ms, no front-end optimisation will fix it — the server simply spends too long thinking before sending the first byte. NVMe storage, a modern PHP version and a CDN bring that under 200 ms. See our plans or the details on SEO hosting — with a 30-day money-back guarantee, testing costs you only time.