How Speedrunners Teach Site Owners to Read a Page Waterfall
A client of ours, a tailoring shop with a booking page, told us their site "felt slow" and asked whether they should switch hosts. Their PageSpeed score was 41 on mobile.

A client of ours, a tailoring shop with a booking page, told us their site "felt slow" and asked whether they should switch hosts. Their PageSpeed score was 41 on mobile. Nobody had opened the waterfall chart, the one view that shows every file the browser fetched, in what order, and how long each one made the visitor wait. Ten minutes with that chart found the problem: a 2.4 MB slider image and a chat widget that pulled in eleven scripts before the page could show its first line of text.
I learned to read charts like that long before I touched a business website. In my twenties I ran a small Discord group of speedrunners for an old platformer, and the habit of that community was to go through a run frame by frame, looking for the exact moment time was lost. A page waterfall rewards the same patience. You are not looking at the total. You are hunting for the few frames, or the few requests, where the time actually goes.
Why speedrunners never trust the final time
A speedrunner who finishes a level in 1:42 instead of 1:38 does not conclude "I was slow." They load the recording, step through it at 60 frames per second, and find that they jumped four frames late at one ledge and took a wide line around one enemy. Four seconds of loss usually lives in two or three places, not spread evenly across the run.
Page load time behaves the same way. A page that takes 6.8 seconds to become usable on a mid-range phone is rarely slow everywhere. In the sites we audit, roughly two thirds of the delay usually sits in three or fewer requests: one oversized image, one render-blocking script, or one slow server response at the very top. A score out of 100 hides this. The waterfall shows it.
So the first speedrunning lesson is simple. Stop reacting to the summary number. Open the timeline and find where the frames were lost.
How to open and read a waterfall in about ten minutes
You do not need paid tools. Chrome DevTools has a Network tab that draws one for any page, and WebPageTest draws a cleaner version from a real device in a real location. For a small business site serving customers in Bangladesh, I test from a Mumbai or Singapore location on a throttled 4G profile, because that is closer to what customers experience than a fibre connection in Frankfurt.
Here is the reading order I teach shop owners, the same way a runner reviews a recording from start to finish:
- Look at the very first bar. That is the HTML document. If the green or teal "waiting" section of that bar is longer than about 600 milliseconds, the server or the hosting plan is slow to answer, and no amount of image compression will fix it.
- Find the vertical line for first paint. Everything to the left of that line is something the visitor waited for while staring at a white screen. Those are the requests that matter most.
- Sort the left side by size. Any single image above 300 KB before first paint is a suspect. Any font file above 100 KB is a suspect.
- Count the domains. Each new domain (a chat tool, an analytics script, a font host) adds a connection setup. Five or six extra domains before first paint can cost a full second on mobile data.
- Look for staircases. When bars start one after another instead of side by side, a file is waiting for a previous file to finish. That chain is usually a script loading another script.
- Write down the three worst offenders and stop. Fix those, test again, and only then look for more.
Step six is the one owners skip. They find twenty small problems and try to fix all of them in one evening, often breaking the site along the way.
The advice I disagree with: chase a perfect score
A lot of performance guides treat 100 out of 100 as the goal. Speedrunners have a phrase for that kind of thinking: optimising the wrong split. A runner who spends a week saving two frames in a level that already looks clean, while ignoring a twelve-second mistake in a later level, will never beat the record.
Websites work the same. I have seen owners spend a weekend removing a 9 KB icon font to push their score from 88 to 91, while their hosting plan took 1.9 seconds to send the first byte of every page. The score moved. Customers noticed nothing. In our experience a small business site gets most of the value by reaching "fast enough" on the slowest pages people actually visit, usually the home page, the contact page and one or two service pages. Past that point, effort is better spent on content or on keeping plugins updated.
I will admit a limit to this. If you run a shop with thousands of product pages and paid advertising, small gains do add up across traffic, and the extra work can pay. For a clinic, a tailor or a local consultant, I have not seen it pay.
What a typical before and after looks like
For the tailoring shop, the three problems from the waterfall were:
- A homepage slider with four photos, each around 600 KB, all loaded before first paint even though only the first one was visible.
- A chat widget loading in the page head, which pulled eleven further scripts from three domains.
- A Google Fonts request for six weights of one typeface, of which the theme used two.
We replaced the slider with one compressed WebP image of 94 KB, moved the chat widget so it loads only after the visitor scrolls or waits five seconds, and cut the fonts to two weights served from the site itself. Mobile load time dropped from 6.8 seconds to 2.9 seconds on the same test profile. The hosting plan did not change. Their PageSpeed score went to 78, which is not perfect, and it does not need to be. Bookings from mobile went up in the following month, though I would not claim the speed was the only reason.
If your waterfall shows the first bar itself is slow, that is a different story, and our hosting and domains section covers how to tell whether the plan or the setup is at fault. The way game server rentals map onto web hosting tiers is also covered in our piece on renting a Minecraft server.
Practise on someone else's site first
Runners learn routes by watching other runners. You can learn waterfalls the same way, with zero risk. Pick three local business sites you visit, maybe a pharmacy, a restaurant and a competitor, and run each one through WebPageTest on a mobile profile. Spend five minutes on each chart and name the worst request. After three sites you will start to recognise the usual suspects on sight: the giant hero photo, the stacked tracking scripts, the slow first byte.
Then do your own homepage. Write the three worst offenders on paper with their sizes and timings. That paper list is your route for the week. Fix one item per day, test after each fix, and keep the old file so you can roll back if a change breaks a layout.
This week, if you only do one thing, open the waterfall for your slowest page and find the single largest file loaded before first paint. In most of the small business sites we see, that one file is a photo, and compressing or resizing it is a fifteen-minute job.
FAQ
What is a page waterfall chart?
It is a timeline of every file a browser downloads to show a page, drawn as horizontal bars in the order they were requested. Longer bars took longer, and bars that start late were waiting on something earlier.
Is PageSpeed Insights enough without a waterfall?
It is a good starting point, and it lists suggestions, but it does not show the order and timing of requests as clearly. The waterfall tells you which suggestion actually matters on your page.
How fast should a small business website load on mobile?
Aim for the main content to appear within about 2.5 seconds on a 4G test profile. Under that, most visitors stop noticing the wait.
Do I need a developer to read a waterfall?
No. Reading it takes practice, not code. Fixing some causes, such as a script loaded by the theme, may need a developer or a support plan.
More in Games
Games
What Online Game Health Bars Teach About Uptime Monitoring
In Valheim, when your character gets hungry, cold or wet, small icons appear in the corner of the screen long before your...
Games
What Budget Phone Gamers Teach Shops About Lightweight Sites
Walk past any tea stall in Dhaka in the evening and you will see someone playing Free Fire or PUBG Mobile on a phone that cost...
Games
How Lost Online Gaming Usernames Explain Domain Expiry Risks
My nephew lost his Xbox gamertag last year. He had used the same short name since he was eleven, stopped playing for a couple...