If you search for what a slow website costs, you will find a great many confident numbers. A one second delay reduces conversions by some precise percentage. Slow sites lose a specific share of visitors. Multiply by your revenue, and there is your loss, in bold, with a currency symbol.
I want to explain why I do not use those numbers, and then show you a more honest way to get at the same question.
Where those statistics come from
The widely quoted page speed conversion figures almost all trace back to studies run by very large companies on very large sites. Retailers with millions of monthly visitors. Travel booking platforms. Media companies. These are real studies and the researchers were not doing anything dishonest.
The problem is what happens to the findings afterward.
A number derived from a retailer processing an enormous volume of transactions gets lifted out of context and applied to a six person landscaping company. The traffic is different by orders of magnitude. The customer’s decision process is different. The purchase is different. The competitive set is different. On a huge site, a fractional percentage change in conversion is a meaningful figure because it is calculated across a vast number of sessions. On a site receiving forty visitors a week, that same percentage is a rounding error applied to a rounding error.
So when someone tells a small business owner that their slow site is costing them a specific amount per month, that figure was produced by taking research about a company nothing like theirs and performing arithmetic on it. It looks like measurement. It is closer to astrology with a spreadsheet.
I am not being pedantic for its own sake. I care about this because business owners make real decisions based on these numbers, and when the number turns out to be fiction, the reasonable response is to distrust the whole subject, including the parts that are true.
What is actually established
Strip out the invented precision and a solid core remains.
Google publishes thresholds for what constitutes a fast page, based on research into human attention and on data about what real sites achieve. Their standard is that the main content of a page should render within 2.5 seconds, measured at the 75th percentile of loads. That is a real, checkable, publicly documented standard, and you can measure your own site against it for free in about a minute.
Google also uses page experience as a factor in ranking. Not the dominant factor, and not one that outweighs actually being relevant to the search, but a real one.
And directionally, the finding that people abandon slow pages is consistent across essentially every study that has looked, at every scale. The magnitude varies enormously by context. The direction does not.
That is what is known. It is less dramatic than the usual pitch and it is enough.
The e-commerce case, where numbers do exist
If you sell online, there is one genuinely well documented figure worth knowing, and it is worth knowing precisely because of how it is constructed.
The Baymard Institute maintains a running meta analysis of shopping cart abandonment research. Rather than running one study and publishing one number, they collect every credible study they can find and average them. As of their September 2025 update, that average across 50 documented studies sits at 70.22 percent.
The number itself is interesting. What makes it more useful is that Baymard also publishes the spread. The individual studies range from roughly 55 percent to 84 percent. That is an enormous range, and it is the honest part. It tells you that cart abandonment varies hugely depending on what is being sold, to whom, and how, and that any single figure presented without that range is hiding something.
So if you run a store, roughly seven in ten people who put something in a cart do not complete the purchase, and that has been broadly stable for years despite everything the industry has thrown at it. Speed is one of several reasons. It is not the only one, and Baymard’s own research on checkout usability points to a range of causes, many of them about unexpected costs and forced account creation rather than performance.
How to get a number that is actually yours
Here is the approach I would use if I wanted to know what speed is costing a specific business, without importing anyone else’s research.
Establish where you stand. Run your key pages through Google PageSpeed Insights. Record the mobile scores and the LCP figures. This takes minutes and costs nothing. Now you have a baseline that is about your site rather than an average.
Find out how many people are affected. If you have any analytics, look at mobile sessions specifically. If you do not have analytics, this is the moment to add something simple. Even a rough count of monthly mobile visitors changes this from a hypothetical into arithmetic.
Know what a customer is worth to you. Not revenue, and not per transaction if you have repeat business. What is the value of one new customer over the time they typically stay with you. You are the only person who can supply this and it is the number that makes everything else meaningful.
Fix the speed. The common fixes are unglamorous and fast: resize oversized images, defer scripts that are blocking rendering, remove widgets nobody uses. For most small business sites this is days of work, not months.
Measure again, after enough time has passed. This is the step people skip and it is the only one that produces a real answer. Compare the same period, ideally against the same season the year before, because a landscaping company comparing March to June has learned nothing about page speed.
At the end of that you have a number that belongs to your business. It might be large. It might be small. Either answer is worth having, and either is worth more than a borrowed percentage.
Why to fix it even if the number turns out small
Suppose you go through the exercise and the measurable revenue difference is modest. Is the work still worth doing?
Usually yes, for reasons that do not appear in a conversion calculation.
The speed fixes are cheap and permanent. Resizing images is a one time job that keeps paying. Unlike advertising, you do not have to keep buying it.
Everyone sees it. Speed affects every visitor, including the ones who arrived through a referral, the ones checking you out after meeting you, and the ones who already decided to buy and just want to find your phone number. It is infrastructure, not a campaign.
The people you lose are your best prospects. Every visitor to your site got there through interest. They are further along than a stranger. Losing them to a loading screen is losing the most qualified people you have.
And the fix rarely stands alone. In practice, whoever is compressing your images also notices that your phone number is not tappable, and that a widget from a cancelled service is still loading, and that the hero photo is four years old. The speed audit is a general inspection that happens to have a number attached.
The honest summary
A slow site costs you something. I do not know what, and neither does anybody quoting a percentage at you.
What I can tell you without hedging is that the standard is public, at 2.5 seconds, that you can check yourself against it for free right now, and that the fixes are almost always cheaper and faster than the people selling them imply.
Start with the measurement. If you come out under the threshold, congratulations, genuinely, go and spend your attention on something else. If you come out well over it, you have found a real problem with a known fix and no mystery attached.
That is a better position than most businesses are in, which is not knowing either way.
The cost that is easiest to overlook
There is one cost of a slow site that does not appear in any conversion calculation, and for small businesses it may be the largest.
Slow sites do not get maintained.
When loading your own site is a mildly unpleasant experience, you stop visiting it. When you stop visiting it, you stop noticing that the phone number is old, the top service is no longer offered, the photos are from four years ago and there is a typo in the heading.
I have seen this repeatedly. The speed problem and the neglect problem arrive together, and the neglect problem usually does more damage. A fast site tends to stay current because its owner can bear to look at it.
So even setting aside every visitor you lose to loading time, fixing the speed changes your own relationship with the thing, and that has a compounding effect nobody measures.
What to ask whoever built it
If somebody else maintains your site, three questions produce useful answers and do not require you to understand the replies in technical detail.
“What is our mobile PageSpeed score and our LCP?” A competent answer includes actual figures. An answer explaining that scores do not matter is a warning sign, because whether or not the score is a perfect proxy, the underlying measurements are real and they are what your customers experience.
“Are our images sized for how they display?” This is the most common cause and the easiest to confirm. If the answer is vague, it is worth checking yourself in the PageSpeed report, which names the specific files.
“What is loading that we do not use?” Old tracking scripts, widgets from cancelled services, plugins nobody remembers installing. This audit is quick and frequently removes a surprising amount.
None of these are gotcha questions. Somebody doing a good job will have ready answers and will probably be pleased you asked.
Core Web Vitals thresholds are published by Google at web.dev. Cart abandonment figures are from the Baymard Institute’s meta analysis of 50 studies, updated September 2025, average 70.22 percent with a documented range of roughly 55 to 84 percent. This article deliberately does not estimate what a slow site costs any specific business.