QR ordering analytics: which numbers actually help an owner

The QR ordering numbers that help an owner are the ones that change a decision. How many people who scan go on to order. When orders arrive. Which dishes sell. What guests asked for that you don’t serve, and which questions went unanswered. How long staff take to confirm orders and answer calls. A scan count on its own tells you very little: it measures curiosity, not service.

The test for every number in this post is the same: name the decision it changes. If you can’t, stop looking at it.

What is QR menu analytics, and what does it usually measure?

Short answer: the reporting your QR tool gives you, which ranges from a scan counter to a picture of how service ran.

A QR menu that only shows a PDF can count scans, and not much else. A QR ordering tool knows more: what was ordered, when, and how long each step took. A chat-based ordering tool knows one more thing that menus don’t, which is what guests asked in their own words. That last category is where the best menu research sits, and most owners never see it.

Before you judge any tool’s reports, keep one fact in mind. Not every guest will order from their phone. In the National Restaurant Association’s 2024 survey, 48% of full-service diners said they would order by QR code, and only 21% of Boomers would (NRA, 2024). Your reports describe the guests who used the tool, not your whole dining room.

Which numbers actually help a restaurant owner?

Short answer: the ones on this list, each tied to a decision. The right-hand column matters as much as the left.

Two numbers are missing on purpose. Total sales belongs to your till, which is the record your accountant uses. Average check is the number vendors like to quote, but it moves with season, price changes and party size, and the best peer-reviewed study found tabletop ordering raised checks by only about 1% while meals got about 10% shorter (Tan and Netessine, reported by Restaurant Technology News, 2019). We explain that study in does tabletop ordering increase check size?.

Is a scan count worth anything on its own?

Short answer: barely. It is the easiest number for any QR tool to show and the hardest to act on.

A scan means a phone opened your page. It doesn’t mean the guest ordered, liked the menu or got an answer. Scans also go up for reasons that have nothing to do with service: a busier night, a bigger party with four phones, a guest who scanned twice because the page was slow to load.

Scans earn their place only as the first step of a funnel. The question is not “how many scanned?” but “of those who scanned, how many ordered, and where did the rest stop?”

How do you read the scanned, chatted, ordered funnel?

Short answer: look for the biggest drop between two steps, then look for the reason in the other reports.

Casa Olivo is a fictional restaurant (an example, not a customer). The figures below are made up to show how to read a funnel; they are not WaiterPilot data.

More people scanned and chatted this Friday, but fewer ordered. The drop between chatting and ordering grew from 10 (80 − 70) to 18 (84 − 66). That points at something in the conversation, so the owner opens two other reports:

• Questions the chat couldn’t answer: six guests asked whether the fish was grilled or fried, and the menu didn’t say.
• Asked for, not on the menu: five guests asked for a non-alcoholic spritz.

Neither number proves why orders fell. They give the owner two cheap things to try: add the cooking method to the fish description, and ask the bar whether a non-alcoholic spritz is easy to add. Next Friday’s funnel shows whether it helped. That is the whole method: a drop, a likely reason, a small change, and the same weekday next week.

What can “asked for but not on the menu” tell you?

Short answer: what your guests want, in their own words, before you spend anything on it.

When guests order by chat, they ask for things: oat milk, a half portion, a dish they had somewhere else, a kids’ option. Each request is a small signal. Over a few weeks, the repeated ones stand out.

Use it with care:

1. Look for repeats, not one-offs. One guest asking for lobster is a wish. Twenty guests asking for a vegan main is a gap.
2. Check whether the dish is really missing or just badly named. “Do you have a burger?” may mean your burger is listed under a name nobody searches for.
3. Price the change before you make it. A request is interest, not a sale.

The questions the chat couldn’t answer work the same way for menu text. WaiterPilot answers from your menu, and when your menu doesn’t say, it tells the guest to ask a staff member. Every unanswered question is a line you can add to a description so the next guest gets an answer. Our guide to preparing your menu for digital ordering covers what to write.

Which service-speed numbers matter?

Short answer: the waits a person controls: confirming first orders, getting food from the kitchen to the table, and answering calls.

Ordering from the phone removes the wait for a server to come and take the order. What’s left are the steps a person still does. WaiterPilot’s owner page reports three of them: time to staff confirm, kitchen time and time to answer waiter calls. On the staff board itself, waiting orders and calls change colour at 1, 5 and 10 minutes, so the team sees delays during service and the owner sees the pattern afterwards.

Read them by hour, not by night. A good average can hide a terrible half hour at 8 pm. If time to confirm climbs every Friday between 7:30 and 8:30, the fix is a named backup confirmer for that hour, not a lecture for the whole team. For the other waits in a meal, our 9-point diagnosis of slow service gives a timing sheet. The rush-hour playbook sets out who does what when every table orders at once.

What does WaiterPilot’s owner page show, and what doesn’t it?

Short answer: the funnel, busy hours, top dishes, guest requests, unanswered questions, a language split and service times. Not payments, not guest profiles, and no published export.

What the owner page shows:

• Today against the same weekday last week.
• Busy hours and top dishes.
• What guests asked for that isn’t on the menu, and questions the chat couldn’t answer.
• The scanned, chatted and ordered funnel.
• A language split.
• Time to staff confirm, kitchen time and time to answer calls.
• A “Tonight at” summary for your restaurant that you can send by text or WhatsApp from your own phone. It opens your messages and you choose who gets it; nothing is sent automatically.

What it doesn’t do, and where to look instead:

For restaurants that serve Arabic-speaking guests, one more owner screen matters: the menu editor shows how much of your menu has Arabic text, and anything missing is shown to Arabic guests in English. If your language split shows many Arabic chats, that coverage panel tells you what to translate first. Multilingual ordering explains how languages work across the chat, buttons and tickets.

How often should you look? A 15-minute weekly routine

Short answer: glance after busy services, and review properly once a week, on the same day.

1. Funnel. Compare this week’s busiest night with the same night last week. Where is the biggest drop?
2. Unanswered questions. Add the top three answers to your menu text.
3. Asked for, not on the menu. Note anything that came up more than a few times. Decide yes, no or test.
4. Top dishes. Check them against prep quantities and anything you ran out of. If a dish keeps selling out, mark it sold out early from your phone rather than disappointing late tables.
5. Service times by hour. Find the worst hour of the week and name one change for it.
6. Language mix. If it moved, check whether your menu text keeps up.
7. Write down one change and look at the same numbers next week.

The routine works for any ordering tool that reports these numbers. If you are in a free trial, the same reports feed the go or no-go decision; how to judge a 30-day ordering trial gives the scorecard.

What to do next

Pick the three numbers from the first table that would change a decision in your restaurant this month, and ignore the rest until those are working. To see where the reports come from, read how WaiterPilot works, or try a sample order to see the guest side. If you are comparing tools, ask each vendor which of the ten numbers above it reports, and add the answers to your buyer’s checklist.