We Measured 182 Local Business Websites. A Five-Star Rating Predicted Nothing.

Every print shop owner in New Bedford I have spoken to has said some version of the same sentence: "Our reviews are great, so the website must be doing its job."

It is a reasonable thing to believe. It is also the single most testable claim in local marketing, and nobody had tested it. So we did.

First, the respect, because it is earned#

The 14 print shops Google returns for New Bedford hold 532 Google reviews between them at a weighted average of 4.77 stars. Not one of them sits below 3.1. That is a decade of getting the wedding invitations right, the banner spelled correctly, the funeral programs done overnight because somebody called at 4pm. You cannot buy that. You cannot fake it at that volume.

Nothing in this article is a criticism of the work those shops do. It is a criticism of an assumption about what that work proves.

The myth, stated plainly#

"If our reviews are good, our website is fine."

The logic seems sound. Happy customers, good business, good business, good site. If something were badly broken online, surely it would show up in the reviews.

We can check that. And it is wrong.

The method, so you can repeat it#

Between 13 July and 8 August 2026 we ran the same audit across 16 business categories in 9 towns and cities, seven of them SouthCoast Massachusetts and two just over the Rhode Island line. For every business Google Places returned, we pulled the star rating and review count from the listing, and the mobile Largest Contentful Paint from Google PageSpeed Insights.

LCP is the moment the biggest thing on the screen finally appears. It is the closest single number to "how long the customer stared at a blank phone." Google calls under 2.5 seconds good and over 4.0 seconds poor. Those are Google's thresholds, not ours.

182 businesses survived into the final sample. Then we simply asked whether one number predicts the other.

Since field data represents what real users are experiencing, it's the most accurate way to really understand what your users are struggling with and what needs to be improved.

The result#

A correlation of 1.0 would mean the rating perfectly predicts the speed. Zero would mean it tells you nothing whatsoever. We ran it twice, on two independent kinds of measurement, and got 0.06 and 0.07.

And what little signal there is points the wrong way from the myth: both correlations are positive against seconds, meaning higher-rated businesses were, if anything, slightly slower.

Here is the like-for-like lab comparison, where every site faced the identical simulated phone on the identical simulated network:

Read the first and third rows again. The near-perfect group was the slowest group in the study. The businesses rated 4.0 to 4.49 loaded a median of 4.9 seconds faster than the businesses rated 4.8 to 5.0.

Now the real-user data, which is the more trustworthy of the two and also the smaller. Sample sizes are small enough that you should not lean hard on any single row:

Different tool, different absolute numbers, same answer: the top-rated group is again the slowest group, and the rating does not track the speed anywhere in the table.

Why the myth survives, and it is not stupidity#

Reviews measure the job you did for the customer who already found you. Site speed determines how many people become that customer in the first place. They are measurements of two different populations, and only one of those populations ever writes to you.

The person who tapped your listing on a phone in the CVS parking lot, waited eleven seconds, gave up and called the next shop on the list did not leave you a one-star review. They left no trace of any kind. Your rating cannot fall because of them, because they never met you.

This is why the myth is self-sealing. The feedback mechanism people trust is structurally incapable of reporting the failure.

Now the part that actually stings#

Of the 228 live, reachable business websites where we could determine it, 90 had no analytics installed at all. Just under two in five. No Google Analytics, no Plausible, no Fathom, nothing.

Among New Bedford's print shops specifically: 4 of the 13 live sites have no analytics of any kind. Those four owners are not choosing to ignore their numbers. They do not have numbers. The star rating is not their preferred evidence, it is their only evidence. Of course they lean on it.

And the shops that do have analytics installed are barely better off, because a standard GA4 install does not put load time in front of you. It shows you sessions. Sessions look fine, because sessions only count the people who waited.

The New Bedford print shops, specifically#

Nine of the 14 got a usable speed measurement. The pattern in the region held inside the category:

• Four of the shops are rated a perfect 5.0. The most-reviewed of them, with 197 reviews behind that 5.0, recorded a 17.4 second simulated mobile LCP. Another 5.0 shop came in slower still at 18.4 seconds.
• The slowest site in the category, at 24.0 seconds, belongs to a shop rated 4.9.
• The fastest lab-measured site belongs to a shop rated 4.7, several tenths below the leaders, at 4.7 seconds. It is also one of the four with no analytics installed, so nobody there knows it is winning.
• Only one shop had enough real traffic for Google to report field data. That one came in at 1.2 seconds, and it is genuinely quick. It is rated 4.8, not 5.0.

Four of thirteen also list a phone number on the website that does not match their Google listing. Six of thirteen have a phone number you cannot tap on a phone.

I want to be precise about what those 17-second figures are and are not. They are simulations on a throttled mid-range Android, and a real customer on good home wifi with a warm cache will see better. They are not a claim that every visitor waits 17 seconds. They are a claim that on the device and connection Google chose as its yardstick, those sites sit far outside the range Google considers acceptable, and that flawless review scores gave their owners no hint of it.

The fix you can do yourself today, free, in about two minutes#

Go to pagespeed.web.dev, paste your homepage URL, and choose Mobile.

Then read it in this order:

1. Look at the very top of the report first. If you see a section headed "Discover what your real users are experiencing", that is field data, the real thing. That number is your answer. Stop there.
2. If that section is missing entirely, Google does not have enough traffic to your site to report on it. That is itself a finding worth knowing.
3. Only then scroll to the lab score. Treat it as a smoke alarm, not a stopwatch. A terrible lab number does not tell you exactly what customers feel, but it reliably tells you something on the page is heavy.

That is the whole exercise. It costs nothing, needs no login, and takes longer to read this paragraph than to run.

The second free fix, five minutes: open your website on your own phone and try to tap your phone number. If nothing happens, it is plain text, not a link. Your web person needs one line changed: the number wrapped in <a href="tel:5085551234">. Six of thirteen New Bedford print shops fail this, and it is the single cheapest lost-call repair in local business.

What a New Bedford print shop should actually do Monday morning#

1. Run the PageSpeed check above. Two minutes. Write the number down; you now have one more piece of evidence than your rating.
2. Install analytics if you have none. Google Analytics is free. Plausible and Fathom are roughly $9 to $14 a month and simpler to read. Any of them beats the current situation, which is guessing.
3. Check the tap-to-call. Five minutes on your own phone.
4. Make your Google listing phone number and your website phone number identical. Four of thirteen shops in your category currently disagree with themselves.
5. Stop treating the star rating as a systems check. It is a wonderful thing. It is evidence that you do the work well. It is not, and was never, a report on whether people can reach you.

---

If you run a shop in New Bedford, Fall River or anywhere on the SouthCoast and you want the actual numbers for your own site rather than a sales pitch, I will run the same audit and send you the results. No charge and no obligation, whichever way the numbers come out.

Joshua Amado, Avalon Partner · 774.559.8992 · Joshua.Amado@AvalonPartner.com

Method note: 182 businesses across 16 categories in 9 towns, measured 13 July to 8 August 2026 via Google Places and Google PageSpeed Insights. Correlations are Pearson, computed separately for lab and field data and never on a pooled set. Businesses that blocked our checker or could not be reached are excluded from the sample entirely, never counted as failing. All businesses anonymised.