The digital menu
One code on the table.
Eleven things happen.
A menu with photographs of every dish, in the guest's own language, that rearranges itself per person, answers questions in plain words, and builds a guest record from the first scan without a login, an app or an email address. Plus an experimentation engine that lets you test eighteen things on your own tables.
One code can
One code can show every dish with a photograph, rearrange itself for the guest holding it, answer a question about a peanut allergy, translate itself into four languages, take one order from four phones at once, split a bill down to the item, call a waiter without anybody waving, read a dish out loud, fill a quiet Thursday with an event, remember a guest who never logged in, and run an experiment on your own menu.
- No app to download
- Opens in about a second
- Four languages
- Works on any phone
- Guest record from scan one
- 18 built-in experiments
What your guest actually sees
Four numbers this moves in your favour.
These are the industry figures we work from, sourced underneath each one. Your own numbers will differ. The point is which four levers a menu on a phone actually reaches.
1 turn
Faster table turns
Ordering and splitting both stop being queues for a member of staff. Industry turnover benchmarks put 10 to 15 minutes off a dining cycle as worth a whole extra turn during peak service. Source: Eats365, on manual bill splitting and turnover
+35%
A bigger average order
Square measured a 35% rise in sales across businesses in the first 30 days of self-serve ordering by code. Nobody feels rushed by a waiter with a pad, so the second round and the pudding get ordered. Source: Square, QR ordering and ticket size
7%
You finally know who your regulars are
Toast and Resy found around 7% of a restaurant's guests can drive up to half its order volume. Most owners cannot name them. From the first scan onwards, you can. Source: Toast and Resy, Q1 2026 retention data
38%
Decisions made on your own guests
Restaurants that unify guest data and act on it recover an average of 38% of guests who were about to drift away, which Bloom puts at over $53,000 a location a year. Source: Bloom Intelligence, guest lifetime value
What happens between sitting down and paying up.
Each one below has a number under it explaining why we built it, and where that number comes from.
Guest · one
Every dish, one page, big photographs
The whole menu is in front of them at once, each dish with its own photograph, and everything the law wants them to know is attached to the dish rather than printed in six point at the bottom.
- A photograph for every dish. We come and shoot them. It is included in the price.
- Nothing is buried. Categories collapse, so a forty dish menu still fits on a thumb.
- Allergens on the dish itself. The 14 declarable allergens, plus "may contain" traces and frozen tags, all per EU guidance. It lives with the dish, not on a laminated sheet nobody reads.
- Your specials on top. A swipeable row of whatever you want pushed tonight.
- Tap for the full detail. Description, ingredients, allergens, a larger photograph.
- They can save a favourite with the heart, and find it again next visit.
+44%
A dish with a photograph outsells the same dish as a line of text. DoorDash's own merchant data puts individual items with photographs at up to 44% more sales a month. It is the cheapest change you can make to what leaves your kitchen.
Sources: DoorDash data via Snappr · EU Regulation 1169/2011
Photography pending
Guest · two
The menu rearranges itself for whoever is holding it
Two guests at the same table see the same dishes in a different order. A vegan is not shown the ribeye first. Someone who saved a pasta last time gets pasta near the top. Nothing is hidden.
- Their palate moves things up. Dishes like the ones they saved before come first.
- What they cannot eat moves down. Ten things they can flag, from nut allergy to halal.
- The category they order from most is promoted ahead of the ones they never open.
- It carries across your venues if you run more than one.
- A first-time guest sees your order, exactly as you set it. It only adapts once it has been told something.
- Plain rules, not a black box. The panel beside this is the whole of it.
+20%
Personalised digital menus lift upsell conversion by 20 to 30%. The same research puts menu engineering driven by ordering behaviour at 12 to 22% on average check. Guests decide faster when the right things are in front of them, and order more.
Source: RaftLabs, AI menu engineering
Photography pending
Guest · three
They can just ask
A guest types the question the way they would say it to a waiter, and gets a real answer with the actual dishes attached, photograph, price and an add button.
- Real questions work. "Party of four eating family style, one has a peanut allergy" gets a sensible answer.
- It only knows your menu. It searches your own dishes and cannot invent one you do not serve.
- The answer appears as it is written, so nobody watches a spinner.
- It keeps the thread. Follow-ups like "more vegetarian" or "something lighter" are one tap and it remembers what was already said.
- They add straight from the answer without going back to the menu.
- Out loud if they prefer. Any dish can be read to them in a natural voice.
72%
Nearly three quarters of diner complaints are about service, not food. Most of those are somebody waiting: to be told what is in a dish, to be advised, to be noticed. A guest who can ask and be answered in four seconds never joins that number.
Photography pending
Guest · four
Order, tip and split the bill, all in the app
The twenty minutes at the end of a meal where six people do arithmetic and one waiter stands waiting. Guests handle all of it on their own phones, then settle up with your staff.
- They order from the table without waiting to be approached.
- Split by item, not by four. Each person picks what they actually had.
- Shared plates divide properly between however many people had them.
- Nobody pays twice. It tracks what is already covered as the split is worked out.
- Tip prompted at the right moment, straight after a good meal. 15, 18, 22% or their own number, and it goes to the person who served them.
- Then they settle with your waiter, card or cash, on your own terminal. Your takings and your card machine do not change.
+24%
Tips rose 24% in a pilot when the bill moved onto the guest's phone. Around 40% of diners in a group want to split, and every split done at the table costs a server minutes they do not have during service.
Sources: Toast and Tabbed pilot · Blogic Systems, on split checks
Photography pending
Guest · five
A waiter, without the waving
Instead of catching somebody's eye, a guest taps what they need. The right waiter gets it on their phone with the table number and the reason, so they arrive holding the right thing.
- Six one-tap reasons. Ready to order, water, clear the table, the bill, a problem with a dish, or something they type themselves.
- Water arrives specified. Still or sparkling, large or small, already chosen.
- A bad plate reaches you in seconds, while it can still be fixed and comped.
- The reason travels with the request, so nobody walks over to ask what is needed.
- It reaches the waiter's phone even with the app closed.
27%
Over a quarter of diners say long waits alone would stop them coming back, and in the same survey of a thousand people, roughly a third had walked out of a restaurant before they had even ordered. This is the cheapest of all of it to remove.
Photography pending
Guest · six
The whole table orders together
Four people, four phones, one order. Everybody adds what they want from where they are sitting and the whole table watches it appear. Nobody shouts their order across the table at one person.
- The basket updates on every phone in about a second.
- Everyone gets a colour and a name, so it is clear who ordered what.
- Special requests per dish. No onions, extra spicy, sauce on the side.
- Latecomers scan and join the order already in progress.
- One order reaches your floor, not four separate ones.
- A dropped phone reconnects and catches itself up without anybody noticing.
97%
Orders placed by the guest themselves came in at 97% accurate, five points above the same orders taken at a counter. Nobody mishears a dish name, and nobody has to stand at a table for four minutes while six people change their minds.
Photography pending
Guest · seven
Fill a quiet Thursday from the menu itself
Your events sit inside the menu, in front of a guest who is already in the room enjoying themselves. They say yes on the spot, from the table, while they still mean it.
- Put an event in front of everyone dining. Wine tasting, live music, a set menu night.
- They RSVP from the table with the number in their party.
- You see who is coming before you order stock for it.
- Nothing printed, nothing posted. Change the date and every phone has the new one.
- No advertising spend to reach a guest who is already sitting in your restaurant.
60%
Repeat guests drive around 60% of restaurant revenue, across more than 100 million guest records. Yet roughly 85% of first-time guests never return. An invitation handed to them mid-meal is the cheapest second visit you will ever buy.
Sources: Olo, 100m guest records · Bloom Intelligence, first-visit return rates
Photography pending
Now the side your guests never see.
One screen you check with your coffee. You change the menu, fix the photographs, learn who your regulars are, and test what actually sells.
Yours · one
Change anything about the menu, yourself
Every part of it, without ringing us. Add a dish at four in the afternoon and it is live before service. Nothing on the menu needs a support ticket.
- Dishes. Add, edit, reprice, describe, retire. Nothing is locked.
- Categories. Create them, rename them, drag them into the order you want.
- Availability by time and day. The lunch menu appears at lunch and hides itself after. Sunday roast shows on a Sunday.
- Sold out in one tap, back on in one tap, on every phone in the room at once.
- Allergens and dietary tags per dish, so the guest filtering works and the law is met.
- Write once, four languages. Change a price and every language keeps up.
8–15%
Menus actively managed against ordering data improve food cost percentage by 8 to 15%. That only happens if changing the menu is a two minute job. On a printed menu the cost of a change is a reprint, so the change waits, and the margin leaks in the meantime.
Source: RaftLabs, AI menu engineering
Photography pending
Yours · two
Fix the photographs without leaving the app
We shoot your dishes. Then the plating changes, or you want it warmer, or a new dish goes on tonight and you took the photo yourself. AI turns even the worst phone snap into something stunning, in seconds, without a photographer.
- AI enhancement. One tap and it fixes lighting, colour and composition for you. A bad photo becomes a good one without touching a slider.
- Proper manual controls too. Exposure, warmth, contrast, saturation, crop and straighten, for when you want to do it yourself.
- Presets that suit food, so a phone photograph stops looking like a phone photograph.
- It tells you what is wrong with a shot before you publish it. Too dark, too flat, badly framed.
- Raise a flat top-down shot to a proper three-quarter angle, with AI.
- Publish straight to the menu, live on every phone in the room.
+13%
Getting photograph coverage up to half the menu was worth an average 13% sales lift on its own. No advertising, no new channel, just a menu where the dishes can be seen. Which is why we would rather you could fix a bad photograph in a minute.
Photography pending
Yours · three
You know your guests, even the ones who never sign up
This is the part that surprises people. There is no login, no app and no email address in the way, and you still end up with a real record of who eats with you and what they like. Every tap teaches you something about your own customers and menu.
- It starts at the first scan. No sign-up, no form, no friction at the table.
- What they viewed, added, ordered and dropped, dish by dish.
- Which dishes only look good. Lots of opens and no orders is your problem dish, and now you can see it.
- Where you lose people. Scanned, opened a dish, added it, ordered, paid. You see how many drop out at each step.
- What they wanted and you did not have. The questions guests typed, in their own words, most frequent first.
- How long they spent deciding on each dish, which tells you which descriptions are not working.
- Your top guests by spend, ranked, with what each of them actually orders.
- Their allergies and diets carried between visits, so nobody explains it twice.
- Their data stays theirs. They can see it, change it or delete it whenever they like.
$53k
Unified guest data is worth over $53,000 a location a year in recovered revenue, by winning back 38% of the guests who were quietly about to stop coming. You cannot win back a guest you were never able to name.
Photography pending
Yours · four
Test it on your own tables before you commit
Every restaurant argues about whether a price is too high or a dish is in the wrong place. This settles it with your own guests instead of an opinion. Eighteen things you can put to the test.
- Price. The same dish at two prices, and what it does to orders and to margin.
- Position and order. Whether a dish sells because it is good or because it is third.
- The photograph. Two shots of one dish, and which one gets tapped.
- The words. Dish names and descriptions, tested rather than debated.
- Specials, bundles and upsell prompts, each on a split of your real tables.
- Six ways to measure it. Orders, average check, conversion, views and more.
19%
Across 2,288 audited tests, fewer than one in five produced a clear winner. That is the honest case for a testing tool rather than against one. Most hunches are wrong, including ours, and the only cheap way to find the ones that are right is to run a lot of them.
Photography pending
Two places we stop on purpose.
Not everything should be automated. These lines are drawn where they are because a restaurant is people serving people.
Your kitchen
Orders go to your team, not a screen
The order reaches your floor staff, not a printer over the pass. Your waiters carry it in, see the table, catch the thing a screen never would. They keep the part of the job that makes them good at it, and your guest still gets a person who knows what they ordered.
Allergies
A convenience, not a safety system
Anyone with a serious allergy should tell a member of staff, and the app tells them so. We surface the information so your guests can browse with confidence, but we would rather be useful and honest than oversell something this important.
See it on your own menu.
Twenty minutes. We load your real menu so you are looking at your own dishes and your own prices, not a demo restaurant. You will see exactly what your guests would see.
- Phone
- +40 726 025 944
- info@splitbite.ai
- We work in
- Romania, USA, Spain, France and Turkey
- Reply time
- One working day
- The rest of it
- Everything we do →