A client told me his site "loads instantly" and he wasn't lying — on his office fiber connection, on a two-year-old MacBook, with nothing else open, it snapped up in well under a second. His mobile conversion rate was a third of desktop. He had no idea why, because the one test he kept running told him everything was fine.
It wasn't the product. It wasn't the price. It was eleven seconds of load time on a mid-range Android over LTE in a parking lot, which is where a lot of his actual customers were standing when they tried to buy something.
The test you're running isn't the test that matters
Every owner checks their own site the same way: good wifi, a decent laptop, first thing in the morning before the coffee's cold. That's the fastest your site will ever load, for anyone, ever. It tells you nothing about the customer on a three-bar connection scrolling through your checkout page on a phone that's two generations old. Those two experiences aren't close. They're not even the same product, from a loading-time perspective.
Nobody emails you to complain about slow
That's the part that keeps this invisible. If your checkout throws an error, you hear about it — angry email, one-star review, a call. If your site just takes six seconds to show a add-to-cart button, nobody writes in. They don't experience it as a bug. They experience it as mild annoyance, and mild annoyance on the internet resolves itself by closing the tab and buying from whoever loaded first. There's no complaint, no ticket, no signal in your inbox. Just a slightly lower number in an analytics dashboard that's easy to shrug off as "soft month."
What was actually on that page
When we pulled the client's site apart, the HTML and product photos were a reasonable 400KB. The page was shipping 4.6MB. A chat widget. Two different tracking pixels installed eighteen months apart by two different agencies, neither aware the other existed. An autoplaying hero video nobody had re-encoded since it was shot. A review carousel plugin that loaded its own copy of jQuery, on top of the jQuery already loaded for something else. None of these were added maliciously or even carelessly — each one got bolted on by someone solving a real problem at the time, and nobody was ever assigned the job of adding up the total weight.
The fix, and it's not a redesign
You don't need a rebuild. You need someone to open the site on a throttled connection — Chrome's dev tools will fake "slow 4G" for free — and cut whatever doesn't survive that test. Most of what gets removed in that exercise isn't core to the business; it's a tracking script nobody's checked the dashboard for in a year, or a carousel that was added because a template came with one. Do this once a quarter, the same way you'd review a subscription bill, because scripts accumulate exactly like subscriptions do: one at a time, each individually justified, collectively enormous.
The takeaway
Your site being fast for you tells you nothing. Test it the way your actual customer experiences it — on a bad phone, on a bad connection, on their third tab — and you'll usually find the thing that's quietly taxing every visitor isn't your product. It's three plugins nobody's looked at since the person who installed them left the company.