What It Actually Means
The distinction explains a common frustration: a fast connection that still feels slow. Upgrading bandwidth does nothing for latency, which is why a site can be sluggish on excellent broadband.
For a page load, latency is usually the larger part of the experience, because a page is not one request. It is dozens, each carrying its own delay.
Why It Compounds
This is the part worth understanding, because it explains where page time actually goes.
A page requests the HTML, which references stylesheets, scripts, fonts and images, some of which reference more. Where those requests happen in sequence rather than in parallel, the delays add up. Thirty requests at 100 milliseconds each is three seconds, and none of it is your server being slow.
The same applies behind the scenes. A page that makes six database queries in sequence, each taking 20 milliseconds, spends over a tenth of a second waiting. Add a call to a third-party service and it can be considerably worse, because you are now waiting on somebody else’s system.
Where It Comes From
Distance. A round trip to another continent costs well over 100 milliseconds and nothing removes it except moving closer, which is what a CDN does.
Number of requests. Each one carries a delay, which is why combining files and removing unnecessary third-party scripts helps disproportionately.
Third-party services. Analytics, chat widgets, advertising, payment providers. Each adds a dependency on somebody else’s response time, and when they are slow, you are slow.
Sequential work. Anything done one after another where it could be done at once, or done at request time where it could have been done in advance.
Why It Matters Commercially
Latency is felt rather than measured by visitors, and it affects behaviour before anyone consciously notices it. Delay in a checkout, a search or a form submission reads as the system being unreliable rather than slow, which is a worse impression to leave.
It also affects internal systems in a way that is easy to underestimate. A screen your team uses forty times a day, taking two seconds longer than it needs to, is a real cost that never appears anywhere.
What To Ask
- Where is the time actually going? Browser developer tools show this directly and it is nearly always surprising. Frequently one third-party script.
- How many third-party scripts are we loading? Most sites carry several added for campaigns that ended.
- Are we waiting on any external service during a page load? If yes, what happens when it is slow.
- Is anything being done at request time that could be done in advance? Caching is usually the answer.
The Practical Point
Most latency problems on ordinary business sites are not solved by faster hardware. They are solved by doing less: fewer requests, fewer third parties, less work at request time, and less waiting on other people’s systems.
That is unglamorous and it is where the improvement actually is. Buying a larger server to fix a page waiting on a slow analytics script is a common and entirely ineffective response.
More terms are in the glossary.
