A Bad PageSpeed Insights Score Is Not Evidence of an Emergency

An email lands in your inbox. Sometimes it's a stranger, sometimes a "marketing audit" you never asked for. Inside is a screenshot: a number in a red or orange circle, maybe a 40, maybe a 54, and a sentence designed to make your stomach drop. Your website is slow. It's costing you patients. Let's talk.

Before you forward that email to your web person in a panic, or worse, hire the person who sent it, give us four minutes. Almost everyone who gets that email is reading the wrong number, and the people sending it are counting on exactly that.

We settled this the only honest way, by pointing the same test at the companies with the most to lose from a slow website. Here is what Google's own tool said about them.

Why You Got That Email

Start with Amazon, because nobody can argue with Amazon. You used it this week. It employs some of the best performance engineers working, and it has published its own research on what slowness costs it. If this number measured commercial success, Amazon would score 100 or somebody would be out of a job. It scores 59.

Then Apple, a company that agonizes over the curve of a corner, selling the most successful consumer product in history. It scores 57. Then Google, which wrote the test, scoring 67 on its own front door.

Now the tell. Anyone on earth can run that test on your website. It needs no access to your analytics, your phone records, or your booking system, and no knowledge of your market, your patients, or your competitors. It is the only number about your business a total stranger can produce without knowing a single thing about your business. That's not a diagnosis. That's an opening line.

What You're Actually Looking At

This is the whole secret, and once you see it you can't un-see it. A Google PageSpeed report has two halves, and they measure completely different things. The scary circle everyone screenshots is the bottom half. The half that actually matters is at the top, and the email never mentions it.

Look at what the two halves are saying in that report. Real visitors loaded the page in 1.4 seconds and passed every check. The robot handed back a 54. Same website, same moment. The gap between those two numbers is the entire business model of the email in your inbox: the circle makes a scary screenshot, and almost nobody knows to scroll up to the part that describes actual human beings.

How far apart can the halves get? When we ran Apple's own homepage through this test, the simulation claimed eight full seconds before anything appeared. Apple's real visitors got it in 1.2. One report, one moment, a sixfold disagreement, and it is the gap hiding behind every forwarded screenshot.

The simulation was wrong about Apple by a factor of six. Anyone who sends you the left panel and never mentions the right one is not running an audit.

So it helps to be precise about what that circle is, and what it isn't.

If one example should end the argument for a dental practice, it's this one.

The Part Nobody Explains

Here is what nobody attacking your score wants to say out loud. The features that make a website produce patients are the same features that drag the number down. Every one of them costs you points, and every one of them earns its place.

A brochure site with three paragraphs, one stock photo, no tracking and no booking tool will score in the nineties. It will also be unable to tell you where a single patient came from. That is the trade the score is quietly asking you to make, and no competent marketer would take it.

Which is exactly why Amazon, Apple, Google and Zocdoc score the way they do. Those sites are packed with tracking, personalization, media and interactivity, because that is what makes them work. They optimized for the customer. The score is what happened afterward, and all four decided they could live with it.

One honest note, because we have a stake in this: Einstein builds and maintains websites for medical, dental, and legal practices, so we are usually the ones a client forwards that email to, and we are the ones who answer for how a site actually performs. That is exactly why our standard is real visitors, not a lab score. We would rather be judged on whether your site brings in patients than on a number that changes with the weather it was tested in.

When It's Real

None of this means speed never matters. Sometimes a site really is too slow, and no amount of context about robot tests changes that. So here are the three signs of a genuine problem. None of them is the orange circle, and you can check all three yourself.

Sign one: your real visitors are failing

This is the green section from the graphic above. Open the "Core Web Vitals" report in Google Search Console, or just read the top panel of any PageSpeed run, and look at whether real visitors are marked passed or poor. If a group of your pages comes back poor, that's a real problem worth fixing, and it has nothing to do with what the simulation claimed. If they pass, you're done: your site is fast for the people who actually use it.

One catch for smaller practices: this real-visitor data needs traffic to exist. A quieter site may not have enough visitors for Google to report on, so the panel falls back to your whole domain or simply shows nothing. An empty panel isn't a passing grade or a failure. It just means Google can't judge yet.

Sign two: every single page is slow to even start

Here's a simple tell. If one page is slow, something on that page is heavy, usually a giant image or a bloated widget. But if every page hangs for a beat before anything shows up at all, the delay is happening before your website even starts loading. That points at the server, the machine that hands your website out to visitors, and it's the one problem you can't fix by shrinking a photo.

Why does one slow machine wreck the whole page? Because that wait happens first, before the browser can begin drawing anything. It pushes everything else back:

The developer name for that first dark bar is Time to First Byte, and Google's rule of thumb is plain: under 0.8 seconds is good, over 1.8 is poor. You won't shrink an image out of that bar, and no plugin will paper over it. The usual culprits are cheap shared hosting, a server sitting a continent away from your patients, or caching that isn't doing its job. This is the one speed problem that's more ours to solve than yours, which is why Einstein sites run on real cloud infrastructure instead of the bargain hosting that causes it. It's the one part of this nobody can design around, so we'd rather control it than explain it.

Sign three: it got slow, and you can name the week

Websites do get heavier slowly over the years. But the sharp drops, the ones people actually notice, almost always trace back to the week something specific got added. When we find a real slowdown, it's usually one of these:

• ›A new chat widget in the corner
• ›A review carousel pulling stars from another site
• ›A booking script from a vendor with no reason to care about your load time
• ›Forty gallery photos uploaded straight off the camera at six megabytes each
• ›Three tracking tags doing the job of one
• ›A script still firing for a vendor you stopped paying two years ago

And there's a human version of this test that beats any tool: pull up your own site on your phone, off the office wifi, out in the parking lot on whatever signal you really get. If your front desk says the site feels slow, believe them, and go look at what changed.

Run that check and you already know more about your website than whoever sent you the screenshot. You'll also bump into the edge of what ten minutes can reach.

The Honest Part

What Ten Minutes Can and Can't Tell You

The check above tells you whether something is wrong. It can't tell you what it's costing you, or what's safe to remove. That takes things a stranger with a screenshot has never seen: your analytics, your call records, your booking data.

That's the honest division of labor. The ten minutes tells you whether something is wrong. Answering what it's worth, and what to pull without breaking something that was quietly working, takes the numbers only you have. That's the work we do, and it's why a real review looks nothing like an eleven-second test.

Send This Back

If a competitor emails your practice a screenshot of an orange circle, you don't need a technical argument. You need four questions. Reply with these:

Nobody selling a page speed audit has a good answer to the fourth question. That's the tell, and it's worth listening for.

The Bigger Picture

Speed Is a Gate, Not the Engine

Even when speed is perfect, it doesn't win you patients on its own. A slow site can hold you back, but a fast one doesn't push you ahead. It's a bar to clear, not a scoreboard. Four things actually decide whether your website brings in patients:

Can people find you? Local search, your Google Business Profile, your reviews, and content that answers what someone actually typed.

Does the page convince them once they arrive? Real photos, real proof, and your voice instead of a template's.

Is it easy to act? Booking that works on a phone, a short form, and someone picking up when it rings.

Can you tell what worked? Call tracking and attribution tied back to what earned the call.

Speed matters most when it gets in the way of one of those four. A fast page nobody can find is still a page nobody sees. So if one of the three signs above describes your site, it's worth fixing, and we'll tell you so plainly. If none of them does, and someone is still emailing you a screenshot of an orange circle, you're being sold a problem, not shown one.