The thesis
The system of record is a commodity. The guest relationship is not.
Hospitality software has spent twenty years competing to be the till, the ledger or the booking sheet. Meanwhile the most valuable position in the industry, the moment a guest actually interacts with a business, has been left to a printed menu and whoever happens to be standing nearby.
At a glance
- Position
- The interaction layer between a hospitality business and every guest who walks in
- Live today
- A cross-venue guest graph, not a roadmap item
- Categories
- Restaurants and bars, hotels, salons and spas
- Integration
- API, OAuth, or an agent operating the screen where a system offers neither
- Also ours
- The venue's imagery and its campaign artwork, produced in-house in the venue's own style
- Moat
- Density. Each venue that joins sharpens the profiles every other venue draws on
- Based
- Bucharest, deployed across Europe
Deck, unit economics per venue and retention data available on request.
Three things became true at once.
Agents crossed the usefulness line
Software could describe a problem long before it could resolve one. Models that read a messy interface, decide, and act changed what a hospitality platform is allowed to promise, and agent software is now the fastest-growing line in enterprise AI spend.
Margins ran out of slack
Full-service restaurants typically net three to six percent, and accommodation and food services carries the highest quit rate of any tracked US industry. Automation stopped being a preference and became arithmetic.
The guest brought their own screen
Three quarters of venues already put a code on the table. The distribution problem is solved; nobody has done anything intelligent with the surface. That is the arbitrage.
Six steps from a code on a table to a defensible position.
The interaction is the only scarce asset left
Tills, ledgers and booking sheets are well served and heavily commoditised. What no vendor holds is the moment of contact between a business and its guest, and everything valuable downstream, from personalisation to forecasting to margin control, derives from data that only exists if you own that moment.
Roughly four in five on-premise orders are anonymous
Olo's own analysis puts it near 80%: an order keyed in with no guest attached. Every "guest 360" product in this market is therefore built on a fifth of the truth, bought from systems that never met the customer. Owning the interaction means originating that data rather than reconstructing it.
Every incumbent's view of the guest stops at their own front door
Enterprise CDPs unify one brand. Group platforms unify one group. Reservation systems only know guests who booked. The genuinely cross-venue networks in hospitality, such as OpenTable, American Express's Resy and Tock, and Fresha, route bookings rather than taste. Nobody is claiming that a guest's profile at one independent business improves their experience at an unaffiliated one.
Which makes the graph a network effect, not a feature
Ordinary vertical SaaS gets stronger for the vendor with each customer. Here each new venue makes the existing venues better at their job, because it deepens the profiles they all draw on. Value accrues to density, which a competitor cannot ship in a release and cannot buy without buying the network.
Holding cost and guest together unlocks a claim nobody else can make
Guest-facing platforms cannot see margins. Back-office platforms never see a guest. Recommending the dish a guest will love and the one the business earns most on requires both halves in one system, which is also the answer to the operator's real fear about personalisation: that it sells more of the wrong thing.
Then agents turn the data advantage into an operating advantage
Nearly half of restaurant brands are increasing technology spend; fewer than one in ten report meaningful impact from AI. The gap is execution, not insight. Agents that act inside tools an operator already owns, and report the attributed delta, are what convert a data moat into revenue, and into a retention story.
The same layer also creates demand rather than only trimming cost. Because we hold the guest signals, the margins, the venue's photography and its design guide, the platform can invent a campaign, produce the artwork in the venue's own style, launch it and measure it. A marketing tool cannot do that without the operating data, and an operations tool cannot do it without the creative pipeline. Owning both is what makes it a product rather than a feature.
Everyone owns a slice. Nobody owns the whole guest.
Each column below is a well-funded, competently run category. The point is not that they are weak. It is that their strengths are confined by the system of record they are built on.
| Capability | POS platforms | QR ordering | Reservations & CRM | Voice AI | Back office | Splitbite |
|---|---|---|---|---|---|---|
| Identifies the in-venue guest | not available | partial | bookers only | callers only | not available | ● |
| Interface adapts per guest | not available | ranking | not available | not available | not available | ● |
| Live experiments on the guest surface | not available | not available | not available | not available | not available | ● |
| Profile travels across unaffiliated venues | not available | not available | bookings | not available | not available | ● |
| Works across restaurants, hotels and salons | not available | not available | not available | generic | not available | ● |
| Sees true recipe and labour cost | partial | not available | not available | not available | ● | ● |
| Acts inside third-party tools | not available | not available | not available | not available | not available | ● |
| Produces the venue's photography and artwork | not available | not available | not available | not available | not available | ● |
| Invents campaigns from live signals and launches them | not available | not available | scheduled sends | not available | not available | ● |
| Attributes the impact of its own actions | not available | not available | not available | not available | reporting | ● |
| Requires replacing the system of record | yes | no | often | no | often | no |
Assessment based on publicly available product and marketing material as of 2026. Categories are generalised; individual vendors vary.
A large, unglamorous market with an unclaimed centre.
We are not betting on the category growing, although it is. We are betting that the most valuable layer inside it is currently owned by nobody, and that whoever takes it collects from every layer above and below.
Analyst estimates vary widely by definition and method. These are the conservative end of credible published ranges, each linked to source.
Ceiling
$9.1T
Hospitality industry revenue
The pool this software is priced against. Mordor Intelligence
Serviceable
$20B → $76B
AI in hospitality and tourism
2025 to 2030, roughly 30% compound growth. The Business Research Company
Core software
$7.6B → $24.1B
Restaurant management software
2026 to 2033 at about 18% a year, and that is before hotels, salons and spas. Grand View Research
Adjacent
$1.8B → $3.8B
Spa and salon software
A category no restaurant vendor will follow us into, because their system of record does not travel. Grand View Research
The guest relationship is being bought out from under operators.
DoorDash owns SevenRooms and Bbot. American Express owns Resy and Tock. Booking owns OpenTable. Thoma Bravo took Olo private. In each case an operator's guest data now sits inside a company that also competes for that guest's attention, and charges the operator for the privilege.
Independent ownership of the guest layer is not a marketing angle. It is the thing operators are quietly most worried about, and there is nobody credible offering it.
DoorDash acquires Bbot, then SevenRooms for $1.2B.
American Express completes Tock at $400M and consolidates it with Resy.
Thoma Bravo takes Olo private at roughly $2.0B equity value.
Operators are renting their guest relationship from competitors. Splitbite gives it back and keeps the network neutral.
Deployed, not theoretical.
Live venues
Running in production in Europe and the United States: Piazzetta degli Amici, Re.Kolta, Madre and Must in Bucharest, E3 by Entourage, and The Serving Spoon in Inglewood, California, a 2026 James Beard America's Classics award winner.
The graph is on
Cross-venue guest profiles are live today, not roadmap. A returning guest is recognised at a venue they have never scanned before.
Full stack shipped
Guest interface, voice concierge, staff and stock operations, invoice reconciliation and P&L, the Studio that shoots and finishes the venue's imagery, and the agent layer that acts inside third-party systems and runs campaigns end to end.
Three categories
One layer serving restaurants and bars, hotels, and salons and spas, which validates the cross-category claim rather than asserting it.
Let's talk about the layer.
We will walk you through the graph, the unit economics per venue, and the retention data behind the agent layer. Deck on request.