The claim that speed affects revenue is everywhere, and the direction is right: a page that renders late is abandoned before it is read, and an unread offer does not sell. But the widely quoted “one second costs X percent” figures came from particular sites under particular conditions. They are not your numbers.
So this part does not repeat anyone else’s statistics. It covers how to measure where your own visitors drop off.
The path from speed to revenue
Break the path apart first. Speed does not touch revenue directly; it travels through several stages, and each stage is a place you can measure.
Slow pages cost you in two places. First, immediately after the click: a long blank screen produces a back button, and that exit often does not register as a visit at all, which makes it hard to notice. Second, on forms and checkout, where someone who has already decided to buy is left waiting. That is the most expensive loss on the site.
How to see where people leave
Three tools are enough, and all three are free.
The key discipline is separating lab scores from field data. A score from developer tooling is measured under controlled conditions; field data comes from real devices on real connections. When the two disagree, trust the field.
Comparison matters too. A single absolute number tells you very little, so compare pages within your own site. If the home page is quick and only the pricing page is slow, the cause is on that page — usually a handful of images or one particular script.
Measure and fix the conversion path first
Trying to make an entire site fast at once leaves you with no obvious starting point. The order is settled: start with the screens the money passes through.
Then measure before and after under identical conditions: same URL, same device profile, same time of day, several runs, and take the median. A single measurement may just be the state of the network that afternoon, which makes it useless as evidence.
Do not borrow someone else’s number as your evidence
There is a strong temptation to tell customers that improving speed lifts revenue by some percentage. That sentence is a promise you never measured, and when it fails to land it takes the credibility of everything else with it.
Say this instead: “this screen responded like this before, and like this afterwards.” Showing before and after, measured the same way — that is genuinely all you can prove, and it is usually enough.
Measurement technique and the improvements themselves live in the Performance archive; the full improvement job including before-and-after measurement is handled by our optimization program, and the exact order we measure and fix in is written out on the process page.
Next part
One part left. Payment feels like the end of the transaction, but for the customer it is the beginning. The final part covers what happens after the money arrives — confirmation, handover and closing, automated.