What Core Web Vitals Actually Cost a Providence Restaurant in 2026

Somebody ran a free speed test on your restaurant's website, screenshotted a red number, and emailed it to you. Now they want money to make it green.

Here is what that work actually costs in 2026, line by line, with a link next to every price. And here is the measured reason most Providence restaurants should not buy it.

First, the number that scared you is probably the wrong number#

When you paste your URL into pagespeed.web.dev you get two different things on one screen. The big coloured dial at the top is a lab score — a simulation, run once, on a throttled virtual phone in a data centre. Below it, if Google has enough real traffic to your site, is field data: the trailing 28 days of what actual people on actual phones actually sat through.

The lab score is the one that gets screenshotted and emailed to restaurant owners. The field data is the one that matters.

Here is what happened when we measured all eight in July.

Every one of those LCP figures is field data — real Providence diners, real phones, trailing 28 days, not a simulation. Seven of the eight are comfortably inside Google's "good" band. The eighth, at 2.6 seconds, misses it by a tenth of a second and is nowhere near the 4.0s "poor" line.

The businesses are anonymised as Site A through Site H, and I have deliberately left out their star ratings and review counts, because a rating paired with an exact review count is a fingerprint that would identify them in about thirty seconds. They did not ask to be publicly graded, and grading them publicly is not the point.

The line items, with real prices#

Everything below was checked in August 2026 against the vendor's own published pricing page, or — where the pricing page hides its numbers behind JavaScript — against the vendor's own published announcement or knowledge-base article. Where a vendor refuses to publish a price at all, I have said so rather than guess.

Tier 0 — the free tier, which is genuinely most of the win#

Worth knowing: the Chrome team's own documentation states the CrUX API is "offered without charge" and that "it is not possible to pay for an increased quota." Nobody can sell you privileged access to your own field data.

Tier 1 — paid tooling, per year#

Tier 2 — hosting, where the real money quietly sits#

WP Engine's own footnote says renewal pricing is "subject to change" and does not state the year-two rate anywhere on the page. Budget for a rise you cannot currently size.

Tier 3 — monitoring, which you almost certainly do not need#

A forty-seat restaurant on Atwells Avenue does not need $900/yr of continuous synthetic monitoring. It needs somebody to open PageSpeed Insights once a quarter. I say that as someone who could bill you for the former.

Tier 4 — labour#

This is the line item everyone fudges, so here is a defensible floor and ceiling.

The Bureau of Labor Statistics puts the mean wage for web developers in the Providence-Warwick RI-MA metro at $49.22/hr — median $47.18, 90th percentile $63.69 — from the May 2025 reference period, published May 2026 (BLS OEWS, area 0039300). That is what the work costs as labour.

What it costs as a service is higher, because agencies carry overhead. Codeable publishes $80–$120/hr plus a 17.5% platform fee — call it $94–$141/hr effective — and a fixed-price "Speed and Security Optimization from $420" (codeable.io/pricing). Clutch's 2026 guide puts the US average at $100–$149/hr, though it publishes no sample size, so treat that as directional (clutch.co).

Notably, I could not find a single US agency publishing a hard-numbered Core Web Vitals package price on a public page. The "$225 to $1,000" range you will see repeated on marketing blogs traces back to nothing citable. Not measured is the honest answer, and it should make you ask harder questions when a quote arrives.

Nor could I price GoDaddy. It is one of the most common places a small restaurant's site actually lives, and in August 2026 its US plan pricing is not published as readable text anywhere I could find — the builder has been rebranded around a credits model behind a "start for free" button. A mainstream small-business platform with no machine-readable price is its own kind of answer.

Three real bundles, totalled#

The free Saturday — $0. Cloudflare Free, LiteSpeed Cache, run your menu photos through Imagify's free tier, measure with PageSpeed Insights. Realistically three hours of your own time. For most of the eight sites we measured, this is the entire job.

The sensible year — $250.88. Cloudways Micro ($132/yr) + WP Rocket ($59/yr) + Imagify Growth ($59.88/yr) + Cloudflare Free ($0). Under twenty-one dollars a month, all in, and it is genuinely fast.

The full build — roughly $977/yr plus one-time labour. Kinsta ($350/yr) + WP Rocket ($59) + ShortPixel Unlimited ($99.90) + Cloudflare Pro ($240) + SpeedVitals Starter ($228) = $976.90. Then labour on top. Six hours is my own working estimate for a typical restaurant site, not a measured figure — adjust it to your situation — but at six hours that is $283 at the BLS Providence median, or $564–$846 through Codeable.

The line item nobody puts on the invoice#

Here is the part that reframes the whole exercise. The single heaviest thing on a typical restaurant homepage is not your logo or your hero shot. It is the reservation or ordering widget — a third-party script, loaded from someone else's server, that blocks your page while it wakes up.

Unfortunately, this puts a lot of your ability to be fast in someone else's hands, and the vast majority of performance slowdowns I encounter in my work are, indeed, caused by external resources that we have little-to-no control over.

What those widgets cost, from the vendors' own pages:

Read the OpenTable Basic row again. On that tier you pay $0.25 per cover for a reservation made through the widget on your own website — unless you pay $49/mo to flat-rate it. You are being metered on the same script that is slowing your page down. On Core and Pro that particular fee is included in the subscription instead.

OpenTable Core at $299/mo is $3,588/yr. That is fourteen times the entire sensible tooling bundle above. The speed problem is real, but the money is not where the speed-test email says it is.

Two more things worth knowing before you sign anything. OpenTable's published plans page states a standard one-year, auto-renewing term. And Square's plans page reserves custom code and embed blocks for Plus and above, so if you are on Square Free and someone tells you to "just drop in the Resy widget," that is a $49/mo upgrade nobody mentioned.

One caveat on this table specifically: OpenTable, Toast and Cloudways all block automated readers, so I checked those pages in a browser rather than a script. If you are about to make a decision on the OpenTable per-cover figure, open the page yourself before you do.

The one free fix, do it this afternoon#

Take the reservation widget off your homepage and replace it with a button.

Not the reservation system — the embedded widget. That iframe or script block sitting on your front page loads OpenTable's or Resy's entire application before a hungry person in the Jewelry District has seen your menu. Replace it with a plain link styled as a button: "Book a table." Send them to /reservations, and put the widget there, on a page they chose to visit.

Your homepage stops paying for a script that most visitors never touch. Your booking flow is unchanged. It takes about twenty minutes in Squarespace, Wix, or WordPress, and it costs nothing.

Then, while you are in there: open your menu photos, run them through Imagify's free tier, and put them back. A 4MB phone photo of your scallops is the other half of the problem, and compressing it is free.

What a Providence restaurant should actually do Monday#

1. Go to pagespeed.web.dev, paste your URL, and scroll past the dial to the field data. If your LCP is under 2.5s, you have no speed emergency. Seven of the eight restaurants we measured are in exactly that position.
2. Pull the reservation widget off the homepage. Twenty minutes, $0.
3. Compress your menu and hero images. Free tier is plenty.
4. Only then, if field data still says you are slow, spend the $250.88.
5. Look hard at the $149–$499/mo you are already paying a reservation platform. That is the real line item.

Four of the eight sites we measured have no contact form at all, five have no LocalBusiness schema so Google is guessing at their hours, and three have no analytics installed at any level — which is usually why nobody has noticed any of this. Those are cheaper problems with bigger payoffs than shaving a tenth of a second off a page that already loads in 1.3.

---

Method. Eight independent restaurants returned by Google Places for "restaurant in Providence, RI," chains and franchises excluded, measured 19 July 2026. Mobile LCP via Google PageSpeed Insights; field data (CrUX, trailing 28 days) was available for 8 of 8, so no lab estimates were substituted. Thresholds are Google's own: under 2.5s good, over 4.0s poor. Medians here are the true medians of the eight published values — the midpoint of the two middle numbers — so you can add up the table and check me. Anyone can re-run the whole thing by pasting a URL into pagespeed.web.dev. Prices verified against vendor pages in August 2026 and are subject to change; anything a vendor would not publish is marked as such rather than estimated.

If you want someone to run these numbers on your own site and tell you honestly whether it is worth fixing, that is a phone call, not an invoice.

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