It's the same trick that Apple and Dell use. The left side of the graph is not 0% but out of habit we think it is.
This is a learned behavior and it can be fixed. I'm mostly broken of it; I assume it's not until I see that it is. Only mostly, not claiming to be a bias-free superhuman here, but as a for instance I had to check the comments to see what was offending people about this chart because I didn't assume it was anchored at zero in the first place. I recommend making the effort if you can; most charts out of school aren't anchored at zero, many wouldn't even be useful if they were.
What jumped out at me actually was the lack of units. At least it links to the original report, and I think this is the Lighthouse Performance Score, which... is itself in units I don't understand? It's some sort of percentage, not speed measure, I think... looks like percentage conformance to an SEO/performance test suite? I'm not sure this is really something that can be bar graphed in the first place sensibly. Percentage conformance to test suites is at least a marginal case at best for a bar graph... how "numeric" such a quantity is is very debatable. e.g., a 40% compliant product may in fact be "less" compliant than a 30% compliant product if the former is compliant on a whole bunch of incidental check points but misses the most important ones, where the latter is compliant on the important ones and misses the incidentals, so such "percentages" can break simple numeric properties like comparability or the transitivity of such comparisons and really aren't numbers. Compliance to such a test suite will also be highly dependent on how you use the system; odds are its competitors can be tuned to go higher without much work and that very easy-to-make mistakes will send a Astro-based site plummeting because of what a user might naturally do with it, through no fault of the framework.
What would make it useless to start a graph at zero?
Whether that is "useless" depends on your point. If you want to make a point that they are not large percentage fluctuations you may prefer to anchor it at zero. But if you want to talk about the fluctuations themselves, you want the chart to be full of those fluctuations, not anchored at zero and all the fluctuations just little wiggles relegated to the top.
So it does depend on what you want to do, but there are use cases where anchoring at zero is either useless, or even itself at times deceptive, especially in cases where "zero" is not a particularly distinguished value on its own. (e.g., voltage. zero voltage is itself a relative measure to some standard of voltage, so a voltage chart being anchored to zero may not mean anything if zero voltage in some context isn't special but is itself somewhat arbitrary.)
The problem that people have with not starting at zero is how arbitrary it is. You can tell a completely different story depending on where you choose the starting point. So there should be some kind of justification for the chosen Y range.
Wouldn't you then plot the _fluctuations_ anchored at 0?
Same with voltage, using some 'absolute' voltage value doesn't make sense. If you put ground and zero at 100V and the positive node to 105V, your 5V device will work just fine.
Always good reading, find at a library near you!