Restaurant Special Requests: How Heep Captures Every Allergy, Birthday & Accessibility Need
1 in 5 guests has a special request that decides whether they return. Heep captures it word for word, in the guest's own language, and briefs your team before service.
A special request is not a piece of information: it is a chain. It gets taken somewhere, it has to be classified, handed off, then delivered on the floor. And it always breaks at the same link: the handoff.
That is why we stopped treating the note as free text. Heep files it under a category, allergy, occasion or accessibility, because a category can be pulled up at opening. A free-text note cannot.
The trade-off is ours to own: classifying forces a decision. "Gluten-free by preference" and "severe peanut allergy" should not trigger the same thing, and the line between them is yours to draw.
Why 1 in 5 guests has a request that decides whether they return
Allergy. Birthday. Pushchair. Wheelchair. Gluten-free. No pork. One guest in five arrives with a special request, and whether they come back rides on that one. Captured and delivered, it builds loyalty. Forgotten, it loses the guest, usually in silence, sometimes on Google.
In the UK, an estimated 2 million people live with a food allergy (Allergy UK), and since Natasha's Law came into force, full allergen labelling is mandatory across all food businesses. On the occasion side, a birthday cover spends on average 1.5× more than a standard cover and brings 5 to 6 guests to the table.
What a forgotten request actually costs you
The real price of a missed detail is not on tonight's ticket. It is in the lost guest, the one-star review, and the word-of-mouth that follows.
Acknowledged, an occasion lifts the return rate by around 30%. Ignored, an allergy can trigger a kitchen return, a food safety complaint, or temporary closure. A forgotten birthday means 5 guests who do not come back: roughly £300 in monthly margin lost from one regular table.
How many special requests does your team miss every week? See how many requests you're losing
The handoff chain, link by link
Four links: noting, structuring, handing off, executing. The first and the last almost always hold. It is the two in the middle that snap.
The one that breaks most often has a name: the sticky note. Jotted down during a rush, poorly transcribed, never passed to the chef, it gets discovered at the table, in front of the guest.
Heep replaces it with a fixed sequence: request captured word for word, classified (allergy, occasion, accessibility), pinned to the guest profile, then surfaced to the team at opening. The free text stays, but it no longer carries the information: the tag does.
4 special requests in one service, 0 missed
Friday evening, central London. 80 covers, 17 reservations across the night, 4 special requests received between 2 pm and 6 pm: severe peanut allergy, 30th birthday, wheelchair, no-pork meal.
Before, one request in three was passed verbally and lost at the shift handover. This time all four landed on the guest profile, ready at opening, with a printed team brief. Four out of four delivered, 2 five-star Google reviews that week, no negative feedback.
At a larger scale, our Noto Paris case study (Moma Group): 1,600 reservations a month, with zero handoff failures.
3 steps to capture every request word for word
Capturing without distorting is the whole job. The sequence never changes.
Rephrasing in the guest's language
Heep plays the request back in the language it arrived in, and confirms what cannot be guessed: the spelling of an allergen, the birthday name, the type of wheelchair.
Classification
Severe allergy, occasion, accessibility: every request gets a tag, and the tag decides what happens next. The guest profile builds itself, message by message, with nothing retyped.
Handoff at the right moment
The note is already on the guest profile before service. Nothing depends on someone remembering.
What separates Heep from a generic AI
A general-purpose AI can reply. A restaurant AI has to rephrase, classify and hand off.
- Trade vocabulary. Heep distinguishes gluten, lactose, peanut, tree nuts, sesame. A generic AI approximates.
- Guest memory. On the second visit, the allergy is already on file. The guest does not repeat themselves, the manager does not ask again.
- Conditional escalation. Severe allergy, ambiguous request, sensitive occasion: Heep does not make the commitment itself, it hands over to your team.
- One thread per guest. Request via WhatsApp, SMS or after a missed call: one thread per guest, every note in the same place.
What is left is the decision nobody takes for you: which categories trigger an escalation, which stay a note on the profile, and who gets what at opening. Heep applies the rule. You write it.
Frequently asked questions
Can Heep handle a severe allergy safely?
Yes. When an allergy is identified as severe, Heep hands the conversation straight to your team rather than making the commitment itself. It is one of the defined escalation cases.
How does the team receive the information?
The request is written onto the booking in your reservation system, with the original conversation attached: your team sees it like any other service note, with no double entry and nothing depending on anyone's memory.
What if the request comes in a foreign language?
Heep replies in the guest's own language. A Japanese guest mentions a sesame allergy? The note arrives in English on the guest profile, without interpretation errors.
Is it compatible with my booking system?
Yes. Heep integrates with Zenchef, TheFork, SevenRooms, OpenTable, and others. The note arrives directly on the existing guest profile.
How long does it take to deploy Heep?
A few days. Initial briefing 30 minutes, calibration, then go live. No commitment.
