Yes, and the honest range is bigger than most people expect. The BBC lost about 10% of users for every extra second a page took to load. Google's own research shows a half second delay on mobile measurably drops conversion too. That doesn't mean shaving 200 milliseconds off your homepage will move revenue this quarter, though. Most of what clients call a "speed problem" is something else wearing a stopwatch.
There's a real mechanism behind the numbers, not a marketing line. Nielsen Norman Group's research on load time found that a delay of around one second is enough to interrupt a person's conscious train of thought. Call it a small system check rather than impatience: something feels broken, and the visitor starts looking for a way out before they've consciously decided to leave. That's the moment you're actually losing, not the page weight itself.
The data trail behind these claims is older and more specific than the plugin vendors let on. Google and Bing found half second delays cutting conversions back in 2009. Mozilla cut load time by two seconds in 2010 and saw a 15% lift. By 2016, more than half of mobile sessions were abandoning a page once it passed three seconds. Real, dated, named studies, not a universal law you can apply to any site at any traffic level. Treat a stat with no source attached the way you'd treat a client testimonial with no name on it.
Here's where the plugin fix runs out of road. Speed gets decided at build time: the framework, the image pipeline, what has to load before the page is even usable. A caching plugin bolted onto an existing site buys you a little, sure, but it can't undo a decision made months earlier when someone chose the stack or shipped every image at full resolution. That's a build decision, made once, by whoever built the thing. Fixing it after launch with a settings panel is treating a structural problem like a stain.
The studies also don't cleanly separate speed from everything else that's broken. Baymard's checkout research finds that "the site had errors or crashed" accounts for about as many abandoned carts as delivery being too slow, and a complicated checkout flow sits right beside both. The average US checkout still asks for more than twenty fields when twelve or so would do. A technically fast page that throws an error, or interrogates the visitor for their life story, loses the sale for reasons no speed score will ever flag.
So before you spend a budget chasing a page speed score, watch where people actually drop off. Run a real test, sure, but look at the step in your funnel where visitors leave too, and ask honestly whether it's load time or friction doing the damage. If it turns out to be the build itself, that's a conversation worth having with whoever built it, not another plugin to install.