Your Star Rating Predicts Nothing About Your Website. We Measured 156 Local Businesses to Check.
Everyone treats the Google star rating as the scoreboard. We tested it against real mobile load times across 156 businesses in Bristol County and nearby Rhode Island, then ran the twenty restaurants Google shows for New Bedford. The correlation is 0.11. Here is what to measure instead.
New Bedford restaurants are good at the hard part.
We pulled the twenty restaurants Google actually shows a customer searching "restaurant New Bedford" — not the twenty we liked, not the twenty with the worst websites, the twenty in Google's own order. Between them they hold 15,786 Google reviews at a weighted average of 4.47 stars, and not one of them sits below 4.2.
You cannot buy that. You cannot automate it. That is a decade of getting the fish right and remembering somebody's name.
So here is the uncomfortable question this article exists to answer: does any of it tell you whether the rest of your online presence is working?
The myth#
It gets said in every chamber meeting and every Facebook group on the SouthCoast, usually as reassurance:
"We're at 4.6 with over a thousand reviews. Our online stuff is fine."
The logic feels airtight. Reviews come from customers. Customers are the whole point. A high rating means happy customers, and happy customers mean the machine is working.
It is the one number almost every owner can quote from memory, and it is the number almost every owner uses to decide whether anything needs fixing. That makes it worth testing properly rather than arguing about.
How we tested it#
Nothing clever here, and you could repeat it yourself.
Avalon has been running the same measurement pass on local businesses since July across 20 categories and towns in Bristol County and nearby Rhode Island — hair salons, HVAC, dentists, florists, print shops, restaurants. That corpus now holds 216 businesses. For each one we record the Google rating and review count from the Places API, then measure the site the way a customer on a phone experiences it: Largest Contentful Paint on mobile, via Google's PageSpeed Insights, using real-visitor field data where Google has enough of it and a lab simulation where it does not.
Of the 216, 156 have both a star rating and a mobile LCP we could actually measure. Those 156 are the test set. The rest either had no website, blocked our checker, or had a site Google could not profile — and they are excluded rather than counted against anyone.
Then we did the simplest thing possible: check whether the rating moves with the load time.
The result#
A correlation of 0.11 is, in plain English, nothing. A correlation of 0.03 is less than nothing. If you handed me a business's star rating and asked me to guess whether its website loads in two seconds or twelve, the rating would not help me.
Broken into bands, it gets worse for the myth:
| Google rating | Businesses | Median mobile LCP | Over 2.5s | Poor (over 4.0s) |
|---|---|---|---|---|
| 4.8 – 5.0 | 83 | 9.4s | 90% | 83% |
| 4.5 – 4.7 | 37 | 3.2s | 57% | 43% |
| 4.0 – 4.4 | 27 | 4.5s | 78% | 59% |
| Under 4.0 | 9 | 7.5s | 89% | 78% |
Read the top row again. The best-rated businesses in the entire corpus have the slowest websites of any band — a median of 9.4 seconds to show a phone anything, with 83% sitting in Google's "poor" bracket.
What New Bedford's restaurants actually look like underneath#
Back to the twenty.
Their ratings are excellent and remarkably tight — 4.2 to 5.0, no weak links. Now the parts a rating cannot see:
- 1 of 20 has no website at all — and it holds the highest rating in the entire set, a perfect 5.0.
- 8 of 16 readable sites have no analytics installed of any kind. No Google Analytics, no tag, nothing. They have no mechanism for finding out that any of this is happening to them, which is usually why it still is.
- 5 of 16 have a phone number you cannot tap on a phone — on restaurant sites, where tapping to call is the conversion.
- 9 of 16 have no contact form, so there is no way to reach them outside opening hours.
- 2 of 19 failed our security-certificate check, which puts a "Not secure" marker in the browser's address bar in front of the menu.
- Among the 9 restaurants whose mobile LCP we could measure, the median was 2.5 seconds — right on Google's threshold, and genuinely decent. But 4 of those 9 failed it and 4 landed in "poor." The slowest took 15.8 seconds.
That first bullet deserves its own sentence. Somewhere in New Bedford is a restaurant that has never had a single dissatisfied reviewer, ranks well ahead of most of its competitors in Google's own results, and has no website for a customer to land on. Tap the listing at 6pm on a Friday and you hit a dead end.
That business does not have a website problem it can see. Its dashboard says five stars.
The number that is actually load-bearing#
Core Web Vitals are measured by the real users of your website, and different users will have different results. They are not "what Google thinks" nor "what googlebot thinks", but what your website's actual users have experienced.
That is the whole distinction in one sentence, from the person at Google whose job is this.
Your star rating is the opinion of the people who made it all the way through — who found you, got the page to load, tapped through to the menu, drove over, ate, and cared enough afterwards to come back and type. It is a survey of your survivors. It is a fantastic thing to have and it says something true about your food and your staff.
It is silent on everyone who bounced. The person whose phone sat on a white screen for nine seconds never left a one-star review. They just went to the place next door, and they left no trace in the only number you were watching.
Only field measurement can accurately capture the complete picture.
The fix you can do today, for free, in four minutes#
Do not hire anyone for this. Do it now, on the phone in your hand.
- Open pagespeed.web.dev and paste in your own website address. Choose Mobile.
- Scroll to the panel headed "Discover what your real users are experiencing." That is the field data — actual New Bedford customers on actual phones, not a simulation.
- Read one number: Largest Contentful Paint. Green means under 2.5s. Orange is 2.5–4.0s. Red is over 4.0s.
- If that panel says there is not enough data to show, that is information too: almost nobody is loading your site on mobile.
Then, on your own phone, open your Google listing exactly like a customer would, tap the website link, and count out loud. Everything you learn in those four minutes, you now have and did not have this morning.
If you are one of the eight with no analytics, that is the second free job. Google Analytics 4 is free, and on Squarespace, Wix, WordPress or Shopify it is a paste-one-code-into-one-settings-box task. Fifteen minutes. You cannot fix what you have never once looked at.
What a New Bedford restaurant should actually do Monday#
Cheapest first. Real time estimates.
Free, 4 minutes. Run the PageSpeed check above and write the LCP number down. This is the metric to watch instead of the star rating.
Free, 15 minutes. Install analytics if you have none. Half the restaurants on this list are flying blind.
Free, 20 minutes. Make your phone number tappable. It is one change: the number on your site has to be a tel: link, not plain text. Whoever built your site can do it in one email, or you can do it yourself in the page editor of any modern site builder. Five of sixteen restaurants here are losing calls to this.
Free, 30 minutes. If you have no website, claim and fill your Google Business Profile properly — hours, real photos, menu link, ordering link. It is not a website, but it stops the dead end. If that is you, do this before anything else on this page.
Free, one afternoon. Fix a failing security certificate. Every reputable host issues one at no charge through Let's Encrypt, and most will switch it on for you if you ask. A "Not secure" marker sitting in front of your menu costs you more traffic than any marketing spend will win back.
A weekend, or a professional. If your LCP is over 4 seconds, it is almost always uncompressed hero photography and a stack of plugins. Compressing the images alone usually moves it more than anything else you will try.
The one-line version#
Keep earning the stars. They are the hardest thing on this list and you have already done it.
Just stop reading them as an all-clear. The rating measures the customers who arrived. The load time measures the ones who didn't. Only one of those two numbers is telling you about the money you are currently leaving on the table, and it is not the one on the wall.
Avalon Partner builds websites, local SEO and business systems for owners across Fall River, New Bedford and the SouthCoast. If you want the four-minute check run properly on your own site, with the specific fixes ranked by what they are costing you, that conversation is free.
774.559.8992 · Joshua.Amado@AvalonPartner.com
Every business measured for this article is anonymised. They did not ask to be graded and it would be unfair to name them. The measurements are real, the methodology is above, and you are welcome to repeat it.
Filed under
Want help putting this into practice?
Avalon Partner helps Fall River and South Coast businesses fix the gaps that cost them leads. Call 774.559.8992 or email Joshua.Amado@AvalonPartner.com.
