How do you stop losing phone bookings when the restaurant is packed?
To stop losing bookings mid-service, the phone and WhatsApp have to be answered there and then, even with a full dining room. The most direct way is an AI receptionist connected to your booking software, such as CoverManager or TheFork: it answers, checks real availability and gets the table booked. Then confirm every booking before service so no table is left empty.
David Zafra · · 9 min read ·
- In 46 digafood restaurants, 4 in 10 calls come in mid-service, while the team is serving tables.
- An empty table is worth about €78 in potential revenue, according to the calculation InfoHoreca publishes.
- An SMS reminder or a card guarantee cuts no-shows, according to CoverManager.
- Anything that is not on the menu, a group or a complaint is always decided by a person on the team.
- Before changing anything, count how many calls go unanswered over a week.
When do calls reach a restaurant?
Many come in exactly when nobody can pick up.
heydiga counted what time calls come in at 46 digafood restaurants. The period ran from 16 September to 15 October 2026, Madrid time. That makes 6,171 calls.
| Time slot | Calls | Share of total |
|---|---|---|
| Mid-service: 13:00–16:00 and 20:00–23:00 | 2,667 | 43% |
| Between services: 16:00–20:00 | 1,676 | 27% |
| Rest of the day | 1,828 | 30% |
The figure does not say those calls are lost: it says they arrive when the dining room is busy. There is no peak within service: only the 13:00 hour stands out, with 13.5% of calls. On average, each service hour gets 7.2% and each hour between services 6.8%, and some service hours get fewer than the hours between services. The problem is that, during service, whoever could answer is serving tables.
This sample is made up of restaurants that already use digafood. It does not represent all restaurants in Spain.
How much does an empty table cost?
About €78 in potential revenue per table, according to the calculation InfoHoreca publishes.
InfoHoreca starts from an average spend of €27 per person and 2.9 people per booking; the article does not say where those two figures come from. It is potential revenue, not profit: it does not deduct staff or ingredients (InfoHoreca, 8 July 2026 (opens in a new tab)).
In Spain, the average no-show rate was 3.3% in 2025 and 3.6% in 2024, according to TheFork (InfoHoreca, 26 February 2026 (opens in a new tab)). With that 3.3% and €78 per table, InfoHoreca estimates about €15,500 a year for a restaurant with 500 bookings a month. That is its estimate, not a TheFork figure.
Work out yours with your own numbers: bookings a month, times 3.3%, times the average bill for your tables. Then add the calls you don't answer, which are not in that sum. No reliable public figure has been found for how many there are. The one that matters is your restaurant's.
What cuts no-shows?
Confirming every booking, and in some cases asking for a guarantee.
CoverManager analysed 9,500 Spanish restaurants on Father's Day 2025 (InfoHoreca, 8 July 2026 (opens in a new tab)):
| Booking system | No-show rate |
|---|---|
| No protection | 1.92% |
| Automatic SMS reconfirmation | 1.52% |
| Card guarantee or prepayment | 0.66% |
Read it with care: it is a single day of high demand, and InfoHoreca warns that it cannot be extrapolated to the whole year. What it does show is the order: confirming helps, and the card helps more.
Asking for a card is the restaurant's decision. Confirmation can be automated without changing how you take bookings.
digafood confirms on the day and flags any changes. If a table comes free, it offers it to the waiting list (heydiga, digafood, accessed 29 October 2026).
Can an AI receptionist take bookings mid-service?
Yes, if it is connected to your booking software and your availability.
The AI receptionist answers several calls at once, with no queue and no hold music. It checks real availability and books before hanging up. On WhatsApp, it answers just the same.
An example. Friday, 21:10. A call comes in: a table for six at 21:30. The team is serving. The call is answered on the first ring, the sitting and the room are checked, and the booking is made in the software. The head waiter sees it on the usual screen. It comes with a note of what was said.
| The AI receptionist handles it | It goes to the team |
|---|---|
| Booking, changing and cancelling, according to availability | A group or an event: it takes the full request and the restaurant decides on the room, menu and time |
| Opening hours, menu, terrace, parking | A question that is not in what has been loaded |
| Allergens, using the menu the restaurant has loaded | A special request outside what has been loaded |
| Confirming the booking on the day | A customer who would rather talk to a person |
heydiga does not improvise: it never makes up a price, a time or a booking. If the question goes beyond what has been loaded, it passes the call to the team.
Does it work with CoverManager, TheFork or Restoo?
Yes: it reads from and writes to the software you already use.
- CoverManager. It reads availability by sitting, room and number of covers, opening hours and the customer record. It creates, changes and cancels bookings, and notes allergies, high chairs and celebrations (heydiga, CoverManager integration, accessed 29 October 2026).
- TheFork. It checks the published availability, whether it comes from the portal or the phone, and creates bookings by sitting and covers. According to heydiga's page, it does not offer a sitting the portal has just sold (heydiga, TheFork integration, accessed 29 October 2026).
- Restoo. It checks the floor plan, sittings, capacity and waiting list. It creates the booking on the right table and sitting, changes or cancels it, adds people to the waiting list and leaves allergies and notes for the floor (heydiga, Restoo integration, accessed 2 November 2026).
Access is authorised from your own CoverManager, TheFork or Restoo account, with a permission you can revoke whenever you like. Nothing to install, and nothing you already have set up gets touched.
How do you test it before signing up?
With your real booking software and made-up data. Note how each case ends: completed, pending confirmation, passed to the team or unavailable.
| Case | What should happen |
|---|---|
| A booking for four in a sitting with a free table | It is made in your software, with sitting and covers |
| A booking in a full sitting | It offers another sitting or the waiting list, without promising a table |
| A table that has just been booked through another channel | It does not offer it |
| A change to the number of covers on a booking | It changes it in the software |
| A large group or an event | It takes the details and passes them to the team |
| An allergen that is not on the loaded menu | It does not improvise: it passes it to the team |
| Two calls at once | It answers both, with no queue |
One rule for any provider, heydiga included: if the booking software does not respond, the booking cannot be treated as made. It has to stay pending and alert a person.
Where heydiga fits
The AI receptionist from digafood, heydiga's product for hospitality, answers your restaurant's calls and WhatsApp in a natural voice and works 24/7. It takes several calls at once, books into your software and passes to the team whatever it cannot resolve.
- To see how it writes into your system, look at the integrations with CoverManager, TheFork and Restoo.
- If you get messages while the restaurant is closed, see the after-hours calls use case.
- If your problem is calls in general, read how to stop missing customer calls.
Nothing to install, no change of booking software and no change to your number. The integration with your software comes with Diga Start: €99 a month excl. VAT, with 500 minutes. No lock-in. Conversational WhatsApp is added for €39 a month excl. VAT: with it, €138 a month.
Book a demo of the AI receptionist
A first step: count a week of calls
Before changing anything, note down over a normal week:
- How many calls are answered and how many are not, by time slot: lunch, afternoon and dinner.
- How many are bookings, how many are repeat questions and how many need judgement.
- How many bookings end in an empty table, and what they had in common: day, group, channel.
- How long the team takes to return a missed call.
With those numbers you know whether the problem is the phone, no-shows or both. And which part to automate first.
Frequently asked questions
Do I have to switch from CoverManager or TheFork?
No. The AI receptionist works on top of the software you already use, and you keep your number. If you ever change booking software, digafood connects to whichever one you use from then on.
Does it answer questions about allergens?
Yes, using the menu the restaurant has loaded into it. If the question is not covered by the menu, it does not improvise: it passes the call to the team.
Does it replace my head waiter?
No. It takes the repetitive calls and the ones that come in when nobody can pick up. The dining room, groups and decisions stay with the team.
Will the customer know they are talking to an AI?
Yes. The AI receptionist introduces itself as a virtual assistant from the very start, and passes the call on if the customer would rather speak to someone on the team.
What about group bookings?
It takes the full request and passes it to the restaurant, which decides on the room, menu and time.
Sources
- The TheFork and CoverManager figures are the ones InfoHoreca publishes (a secondary source): they could not be read on the TheFork or CoverManager websites.
- InfoHoreca, “La tasa del no-show cae al 3,3% en 2025 frente al 3,6% de 2024, según datos de TheFork”, 26 February 2026, infohoreca.com (opens in a new tab).
- InfoHoreca, “Las reservas fantasma cuestan miles de euros al año a los restaurantes”, 8 July 2026 (TheFork data and CoverManager analysis), infohoreca.com (opens in a new tab).
- Features, integrations and pricing: digafood, CoverManager, TheFork and pricing, accessed 29 October 2026; Restoo, accessed 2 November 2026.
- Calls by time slot: heydiga's own data, 46 restaurants on the digafood line, 16 September to 15 October 2026.
How it was written: written with the help of AI tools; reviewed and verified by the heydiga team.
