Restaurant website mistakes that cost covers

The person on your website is not browsing. They are standing on a sidewalk with two other restaurants open on other tabs, deciding where their group is going in the next four minutes.

Every second of friction is a vote for one of the other tabs. That is the whole frame, and it explains most of what follows.

What does a customer actually want from my website?

Four things, and they want them without scrolling.

Can I eat what you serve. What does it cost. Are you open. How do I get a table or an order in.

That is it. Not your story, not the chef's philosophy, not a slideshow. Those things matter to a customer who already decided, and they belong further down the page.

There is measured support for what people are actually asking. Maple analyzed 1.2 million calls across more than 1,000 US restaurant and local business locations between December 2023 and November 2025 and published the intent breakdown: 22% of restaurant calls are hours, location or directions and 15% are menu questions. Another 12% are takeout or delivery orders. Maple sells restaurant phone software, so treat the taxonomy as vendor research with a disclosed sample.

Better than a third of your phone volume is people asking a website questions. If your site answered them, that call would not exist, and the ones that did come through would be reservations, which Maple puts at 38% of calls and notes that 58% of reservations still originate by phone.

Test it the way the customer experiences it. Open your site on your phone, on cell data, not on the restaurant's wifi, standing outside. Start a timer. If a stranger cannot find your hours, your menu, and a way to book within ten seconds, the site is costing you covers regardless of how it looks.

Which mistakes cost the most covers?

Ranked by how often I see them and how much they cost.

The menu as a PDF. It downloads, it opens in a separate viewer, it is unreadable on a phone, and search engines cannot read the dish names at all. This one item costs both traffic and conversions at the same time. The proper structure is worked through in how to build a menu page Google can read.

Hours that disagree with your Google listing. The customer trusts whichever one is wrong, drives over, and finds a locked door. Then they write about it. Both records need to say the same thing, which is why hours belong on the monthly check of your Google listing rather than in someone's memory.

A homepage that is a full screen video. Slow on cell data, useless on a small screen, and it pushes every fact the customer wants below it.

No prices. A menu without prices makes people assume expensive and leave. It also removes the exact detail a group organizer needs.

A phone number that is not tappable. Written inside an image, hidden in the footer, or replaced entirely by a contact form. Restaurant customers call. Let them.

No address in text. An embedded map is not an address. Write the street, city, and cross street in words, and add a sentence about parking and where the entrance actually is.

A reservation button that breaks on mobile. Test it on your own phone, all the way to a confirmed booking, at least once a season.

Why does the PDF menu keep coming up?

Because it is the one mistake that fails in three directions at once, and because it is so easy to make.

Somebody has the menu file already. Uploading it takes thirty seconds. Updating it takes thirty seconds. It feels efficient, and for the person maintaining the site it genuinely is.

But the file cannot be read by a search engine looking for your dishes, cannot be read comfortably by a person on a phone, and cannot be updated by anyone who does not have the design file. So your most searched content becomes invisible, awkward, and fragile at the same time.

Publish the menu as text on a page. Keep the PDF underneath it for people who want to print it. That is the whole fix and it usually takes an afternoon.

How much does slow actually cost?

More than the design, and the research here is Google's own.

Google and SOASTA trained a model on bounce and conversion data with a stated 90% prediction accuracy. The finding, verbatim: "as page load time goes from one second to 10 seconds, the probability of a mobile site visitor bouncing increases 123%." From the same work: "as the number of elements, text, titles, images, on a page goes from 400 to 6,000, the probability of conversion drops 95%."

Google's related crawl of 11 million mobile pages found 79% were over 1MB, 53% over 2MB and 23% over 4MB, and that 25% of pages could save more than 250KB by compressing images and text alone. A full screen hero video is that entire problem in one element.

Know which version of that statistic you are holding, because there are two and both are Google's. The 2017 edition said one second to seven seconds and 113%. The 2018 update, which is the live page, says one second to ten seconds and 123%. They are one finding, re-run. Anyone presenting them as two independent studies has not opened either.

The second study worth knowing is Google with Deloitte and Fifty-Five, "Milliseconds Make Millions": 37 brand sites, roughly 30.5 million sessions, hourly data over 30 days in October and November 2019. A 0.1 second improvement in mobile load time produced retail conversions up 8.4% and average order value up 9.2%, and travel conversions up 10.1%.

Here is the caveat nobody selling a performance retainer will give you: there is no restaurant in that sample. It was 15 retail brands, 10 luxury, 6 travel and 6 lead generation. The direction is well supported. The specific percentage is not yours.

What is actually actionable is the threshold set. Google's Core Web Vitals define good as Largest Contentful Paint at 2.5 seconds or less, Interaction to Next Paint at 200 milliseconds or less, and Cumulative Layout Shift at 0.1 or less, all measured at the 75th percentile of real visits. Those are pass or fail numbers you can check for free today.

How do I tell whether my site is the problem?

Look at what happens after people arrive, not at how many arrive.

If your listing gets plenty of views and your site gets visits but the phone is quiet and the reservation book is thin, the problem is on the page. If almost nobody is finding you at all, the site is not your first project.

Two habits make this readable. Have analytics installed so you can see how many people land on the menu page and how many reach the booking step. And check whether your inquiry forms are actually being finished, because a form that people start and abandon looks identical to a form nobody found. Form structure decides more of that than the design does, which is the argument in how to structure an inquiry form people actually finish.

If you run more than one room, compare them rather than looking at one in isolation. Two locations with the same site and different results are telling you something about the room or the market rather than the website, and separating those is the point of understanding why two locations perform differently.

What is the abandoned booking actually worth?

Here is the arithmetic. Replace every input with your own.

Say your reservation platform reports 380 booking starts last month and 247 completions. That is 133 people who began and stopped.

Now get two numbers from your POS rather than from the internet. Your average check per person and your average covers per ticket. Every major POS reports both natively. There is no credible published US average check or average party size for independent restaurants, so a modelled figure would be worse than the one sitting in your own system.

Say yours are $38 and 2.4 people:

• 133 abandoned starts, times 2.4 covers, times $38, is about $12,130 in sales that started and stopped.
• Not all of it is recoverable. Some were duplicates, some booked a second restaurant in the same tab, some were never going to come. Assume you recover a quarter: about $3,030 a month.
• At the National Restaurant Association's 2024 full-service food cost median of 32.0% of sales, that is roughly $2,060 of contribution before whatever extra labor those covers need.

That is $24,000 a year of contribution sitting behind a form. Not behind a redesign, not behind a campaign. Behind the number of fields between a person and a table.

Which is exactly where the industry's favorite conversion statistic falls apart.

Does cutting form fields really lift conversions 120%?

There is no study behind that number, and it is worth naming because you will be quoted it.

"Reducing form fields from 11 to 4 increases conversions 120%" circulates on agency sites and landing page tools with no traceable primary source. It is not a Baymard Institute finding, though it is frequently implied to be one.

What Baymard actually publishes is more useful and less dramatic. Its checkout research puts the average flow at 5.1 steps and 11.3 form fields, notes that 17% of users have abandoned a purchase because checkout was too complicated, and reports that step count has barely moved since 2012. Its load-bearing conclusion after more than a decade of testing is that "what really matters to the overall UX of checkout is the number of form fields users must consider."

Two limits on that. Baymard sells UX research and benchmarking, so label it as vendor research. And it is e-commerce checkout work, not restaurant booking work. It supports fewer fields as a direction. It does not support a percentage attached to your reservation form.

So cut the fields, and refuse the number. Every field you delete is a real improvement you can measure against your own baseline, which is the only comparison that means anything at your traffic level.

What to do this week

Open your site on your phone, on cell data, outside the building. Time how long it takes to find hours, menu, and a way to book. Write the number down.

Put your menu on a page as text with prices. Keep the PDF as a secondary download.

Make the phone number tappable and put it at the top of every page.

Write your address, cross street, parking, and where the entrance is, in plain words, on the homepage and the contact page.

Then make your hours identical everywhere: site, Business Profile, reservation platform, delivery marketplaces. Pick one source of truth and copy from it.

Then complete a reservation on your own phone, start to finish, and count the fields. Delete every one you do not use in service.

Questions about your site?

Email me at eric@seod.com with your homepage address. I will open it on my phone on cell data, run the ten second test, and send you the list of what a hungry stranger cannot find, in the order I would fix it. Three or four lines back, usually the same day.

I ran restaurants for sixteen years and I answer these myself. If your site passes, I will tell you to leave it alone and put the budget somewhere it will do more.

Or keep reading more on restaurant marketing.