MargaritasA taqueria moved from a truck into its own restaurant. Its website was hidden from Google and had no menu, so I built the one I'd pitch them.
Measured on the preview. Not launched yet.
TodayProposed
Nobody asked for this one. I noticed the problem as a customer.
We've been going to Margaritas for years, since their truck in Sunnyvale. Then the truck moved closer to Santa Clara, near a hospital. When we looked them up again, they had their own restaurant, which is a huge step for a food truck. But online, the new place was hard to find.
Their website told Google not to list it, it was a template for a different kind of business, and the only menu online was on DoorDash. Some listings still sent people to the old truck spot.
My fiancée looked at their website and said I should help them. So instead of a pitch deck, I built the site, put it on a preview address, and wrote a short proposal around it.
What I found, and what the proposed site does about each one.
A noindex setting, usually left on by accident, tells search engines not to list the site.
FixIndexing is tied to the real domain: it turns on there by itself, with a sitemap, while every preview copy stays hidden.
Pages like "Services" and "Areas Served", and a box for hiring cleaners and plumbers. No food anywhere.
FixFour pages built from scratch around the food and the story: Home, Menu, About and Visit.
Google still listed the old food truck address, and Google and Yelp disagreed on Friday closing time.
FixOne config holds the address, phone and hours, and every page, the footer and the structured data read from it. Fixing the Google listing is part of launch.
To see what they serve, customers had to go through a delivery app, at delivery-app prices.
FixThe full menu on its own page: 34 items in 8 sections, with meat choices, add-ons and labels like Fan fave, Spicy and Veggie.
Phone first, because that's where people decide where to eat.
123456
The decisions that make a small restaurant site cheap to run and hard to break.
Most restaurant sites stack a page builder and a dozen plugins, and each one is something to update and secure.
I wrote a custom WordPress theme that needs no plugins and no page builder. Each page is about 33KB of HTML, with one self-hosted font and no jQuery.
The menu had to go live with real data, then belong to the owners, not to me.
I scraped their DoorDash menu into a starting file that loads once on install. After that, every item is a normal entry in WordPress, with sections, labels and Quick Edit for prices.
A cached page would freeze the badge at whatever time it was saved.
The badge is worked out in the visitor's browser from the saved hours, so it stays correct even when the page itself is cached.
A live preview of their site could end up in Google next to the real one.
Indexing only turns on for their real domain. The preview, staging and local copies all tell search engines to stay away, so moving it over is the only switch.
No calling anyone to change a price. Log in, edit, and it's live when they save.
WordPressA login the owners already understand, with an Editor role so they can't break the theme.
Custom themeNo plugins or page builder; activating it creates the pages and imports the menu.
Structured dataRestaurant, Menu and FAQ markup plus an llms.txt, so search and AI tools read the facts directly.
CloudflareSSL and Cloudflare protection, with caching and daily backups from the host.
GitHub ActionsEvery push is syntax-checked, uploaded and the cache cleared, so changes are checked before they go live.Measured on the preview against their current site. Nothing here has launched yet.
| Tested on a phone | Today | Proposed |
|---|---|---|
| Performance score | 68 | 89 |
| Accessibility score | 85 | 100 |
| Best practices score | 79 | 100 |
| First thing on screen | 2.7s | 1.7s |
| Data downloaded | 1,216 KB | 532 KB |
| Files requested | 58 | 10 |
| Listed in Google | No | Once on their domain |
Google Lighthouse, simulated phone on a slow mobile connection, October 2, 2026, middle of three runs each. The new site's performance score ranged from 77 to 100 between runs.
People searching for the restaurant would find its own site, not just third-party listings.
Customers could see what they serve without opening a delivery app first.
One address and one set of hours on every listing.
Being regulars. We've followed them since the truck days, so this felt like helping friends, not a sales call. Building it before pitching it helped too: instead of mockups and a promise, the proposal could say "it's live, try it on your phone."
I'd talk to the owners sooner. The menu came from DoorDash, so prices, Friday hours and even the spelling of the name still need their sign-off before launch.
If they say yes, I move it to their domain, turn on indexing and point their Google listing at the restaurant. Then a 15-minute walkthrough of editing it from a phone.