Three points of use
Three Food2Mood solutions for the digital menu, service and marketing
Each module solves its own business task, yet they all run on the same Food2Mood core — the guest taste profile and the tagged menu.
Recommendation engine
Personal recommendations
Analyses order history, tags the menu and builds the guest taste profile to return personal recommendations and upsells over API
Waiter assistant
Prompts for the floor staff
Shows the waiter which dishes, drinks and add-ons to offer a particular guest based on their tastes and current order
Taste audiences
Marketing personalisation
Matches dishes and offers against the taste profiles of the customer base and builds ready audiences for further communication
One platform
one taste profile
recommendations at every touchpoint
Modules connect separately and share one core — start with a single module and add the rest later.
Explore the modulesHow it fits together
Three solutions run
on one taste profile
Digital taste profile
A shared core for all three modules
The profile is fed by events from all three modules and returns recommendations to them — the more touchpoints are connected, the sharper the output in each of them.
Recommendation engine
The recommendation engine inside your app
Food2Mood analyses the purchase history of guests, tags the menu by taste attributes and builds a digital taste profile of the user. Trained ML models then decide which dishes, add-ons and combinations are most likely to suit a particular guest.
Analyses purchase history
Detects stable preferences, choice patterns and frequent ordering scenarios
Tags the menu and the users
Labels dishes and guests by tastes, ingredients, cuisines, formats and restrictions
Understands pairings and upsells
Finds which dishes, drinks and add-ons go together most often
Uses trained ML models
Builds recommendations from the taste profile, the cart and the order context
Guest data + menu markup + ML models = accurate personal recommendations
Recommendation engineProven in practice
A/B test results in a coffee chain app
to revenue from upsells compared with a conventional recommendation system
Measured over the first week of the test in the app. The comparison was not against having no recommendations, but against the standard ranking logic already in place.
Recommendation engineHow it works
From guest data to a personal recommendation
The app passes Food2Mood a guest id and the current cart; the engine matches them against the taste profile and the tagged menu and returns a personal ranking.
Identification
The app passes a user_id, and Food2Mood finds the taste profile and order history
Taste profile
The system uses purchase history and the preferences already known
Real-time recommendations
The app sends behaviour data, and the engine adjusts the ranking to the current situation
Business settings
The engine respects the rules and priorities set by the restaurant in the admin panel
Balance of interests
The ranking keeps dishes that suit the guest and serve the goals of the venue
The engine runs in two scenarios: it builds a personal ranking by user id, or recommends the next dish given the cart already assembled
Recommendation engineManaging recommendations
Managing and tuning recommendations
The restaurant sets the rules and priorities, and recommendations show the right dishes to the guests they actually suit.
Ideal order
Defines the target order structure the engine aims for when picking add-ons
High-margin dishes
Raises the priority of high-margin items among the relevant recommendations
Food pairings
Defines correct combinations of dishes, drinks and add-ons
Go-list and stop-list
Boosts the items that need to sell and excludes the unavailable or unwanted ones
Food2Mood balances relevance for the guest against the commercial goals of the restaurant
Waiter assistantWhat the waiter sees
The waiter knows the guest before the order starts
Before taking the order the waiter opens the table and sees tastes, restrictions and what to offer next.
Tastes and restrictions
Visible before the guest opens the menu
What to offer now
The prompt changes as the order comes together
Shift tasks
The manager sets a goal for the floor or one waiter
Recommendation quality stops depending on how long a waiter has worked here
Taste audiences
Building audiences for precise communication
Food2Mood analyses the menu and order history, builds a digital taste profile of guests and finds the most relevant audience for a particular offer.
Menu markup
Dishes get gastronomic attributes: ingredients, tastes, textures, serving format and consumption scenario
Guest taste profile
Order history turns into a digital profile: favourite ingredients, tastes and behaviour patterns
Audience search
The system matches the offer profile against the customer base and finds the most relevant guests
- 1
Menu
The venue dish catalogue
- 2
Markup
Dishes get taste, texture and serving attributes
- 3
Taste profile
Order history turns into the guest tastes
- 4
Matching
The dish is matched against the customer base
- 5
Audience
Guests the offer actually suits
Taste audiencesHow it works
Every message becomes a personal recommendation
The system analyses the dish, compares it with guest taste profiles, excludes the irrelevant audience and generates messages with different accents.
Choosing the offer
The marketer picks a dish, a new item or a promotion
Dish data
The system pulls the dish card from the tagged menu
Taste analysis
It determines tastes, ingredients, cuisine and dish attributes
Audience search
The system finds the guests the offer suits
Filtering
Guests with a mismatch or restrictions are excluded
Ready segment
The audience and taste segments are handed over for communication
Selection example at the audience search step
- · Likes prawns
- · Often orders soups
- · Asian cuisine
- · Likes spicy food
- · Orders seafood
- · Often picks first courses
- · Likes sweets
- · Orders meat
- · Excludes fish
Targeted communication instead of mass mailings that exhaust the customer base
Modules at a glanceWhat changes
What each module changes and for whom
The three modules share one profile, so their effects overlap. Here is exactly where.
| Engine | Waiter | Audiences | |
|---|---|---|---|
| Guest | Finds a suitable dish without reading the whole menu | Is served with tastes and restrictions already known | Only hears about things that actually suit them |
| Staff | — | Ready prompts and arguments; new hires learn the menu faster | — |
| Marketing | In-app reactions sharpen the taste profile | Floor-level preferences feed the same profile | One idea becomes a campaign with no manual segmenting |
| Business | Higher conversion and average check — inside your priorities | Service quality depends less on staff experience | More repeat visits, higher campaign conversion |
Guest
- Engine
- Finds a suitable dish without reading the whole menu
- Waiter
- Is served with tastes and restrictions already known
- Audiences
- Only hears about things that actually suit them
Staff
- Waiter
- Ready prompts and arguments; new hires learn the menu faster
Marketing
- Engine
- In-app reactions sharpen the taste profile
- Waiter
- Floor-level preferences feed the same profile
- Audiences
- One idea becomes a campaign with no manual segmenting
Business
- Engine
- Higher conversion and average check — inside your priorities
- Waiter
- Service quality depends less on staff experience
- Audiences
- More repeat visits, higher campaign conversion
CDP
To get more out of your data — connect the CDP tools
Recommendations get sharper as the data about the guest gets fuller. A CDP collects it from every channel into one profile and turns it into personal scenarios.
Data collection points
Collect guest data during service
Guest data
Order history, preferences, taste profile, restrictions
CDP
Merges all the information into one customer profile
Personal scenarios
Audience segmentation, triggered events, personalisation of site and apps
Revenue growth
Higher average check, more frequent visits, higher LTV
Where the data comes from
- POS: orders, receipts, basket composition
- Waiter assistant: recommendations, preferences
- Site and app: menu views, cart
- Delivery: online orders, purchase frequency
- Campaigns: opens, clicks, reactions
- Loyalty programme: bonuses, status
What you can do next
- Segmentation by behaviour and value
- Triggered scenarios based on events
- Templates for SMS, push and email
- Personalisation of the site and the app
- Managing bonuses and statuses
- Analytics of campaign performance
CDPSystem modules
Key modules of the CDP
Show the eight platform modules
Unified customer profile
Merges customer data from different sources into a single user card
Audience segmentation
Builds customer groups by shared traits and behaviour models
Campaign manager
Lets you run triggered omnichannel communication scenarios
Loyalty programme
Tiers, bonuses, certificates and deposits instead of scattered discounts
Personalisation
Adapts the site and the app to a particular customer segment
Message templates
An editor for SMS, email and push templates
Pixel
Tracks customer actions on the site and records events
Analytics
Shows how the modules perform and builds reports
Analytics closes the loop: it shows the result of every action and helps improve the next communication
Personal data operator
The company is listed in the operators register; guest data is processed under Russian law
Two registered programs
The recommendation engine and related modules are registered as in-house software
Moscow IT startup register
Food2Mood is listed in the city IT startup register and works with Russian chains
API integration
The engine plugs into your app and does not require replacing the restaurant POS
Access separation
Settings and data are split by role: chain admin, venue manager, waiter
Service data only
What leaves the perimeter is a guest id and menu markup, not personal details
Send a request
Apply to join the programme. We will get in touch, clarify the details and help pick the solutions that fit your organisation.
FAQ
A recommendation system for restaurants and food-service chains. It builds a digital taste profile of the guest, matches it against the tagged menu and suggests what to offer a particular person at a particular moment instead of the same menu for everyone.
Three. The recommendation engine returns personal ranking over API. The waiter module shows prompts to the floor staff. Taste audiences build segments from taste and behaviour patterns for marketing. All three share one core and can be connected separately.
Over API: the system takes a user id and returns an ordered menu — a ranked list of dishes for that guest. It plugs into your existing app, so there is no need to build a separate interface.
No. The engine works on top of your current stack: it receives the menu markup and order data and returns the ranking back to your app or to the waiter assistant.
Priority dishes, exclusions, food pairings, stop-list and go-list, a separate boost for high-margin items and the structure of the ideal order. The engine builds recommendations inside those rules, not against them.
A coffee chain: +5% to upsells and +5% to their share of app revenue in the first week of an A/B test. An author-cuisine restaurant: +10.6% to the average check. A ready-meal service: a pilot with a forecast of +10% to conversion. A federal retailer: a ready-meal pilot with a forecast of +2% to GTV at scale.
The waiter opens the table, sees the guest profile and ready prompts: what to recommend and what to add to the order. A manager can set a task — for the whole shift or for one employee — and track how it is going.
From order history, the current cart and behaviour in the app. This is matched against the menu markup — tastes, ingredients, cuisine and dish attributes — and the personal ranking is assembled from the overlaps.
The audience is collected by taste segments rather than by broad traits: the system picks guests for whom the offer is genuinely relevant and excludes the rest. Targeted communication instead of mass mailings that burn the base.
Send a request — we will discuss the tasks of your venue, pick a suitable module and pilot terms. An A/B test on your own audience is used to measure the effect.