Development

FoodTrack AI

A Telegram Mini App that counts calories from food photos, with AI nutrition reports and its own analytics system

Project
Calorie counting from food photos in Telegram
Service
Websites and apps
Format
Telegram Mini App · AI · analytics

Task

Build from scratch a food diary inside Telegram where nothing has to be weighed or looked up in food databases: you photograph your plate and AI works out the dish, its weight, calories and macros. And from day one, see which acquisition channels and prices bring in payments.

Solution

  • Food recognition from a photo, a label screenshot or a text description: name, weight, calories and macros, a goal-aware conclusion and a usefulness score from 1 to 10. Results can be corrected in text and recalculated.
  • An onboarding survey and a personal norm based on the Mifflin–St Jeor formula, adjusted for activity and goal: lose, maintain or gain weight.
  • AI reports for the day, week and month: a rating of nutrition and activity, with specific advice.
  • A daily dashboard with calorie and macro rings, history with search and favourites, activity and calories burned, weight and calorie charts.
  • A Telegram Stars subscription with a trial and promo codes; prices and the trial were A/B tested across six groups.
  • A referral programme: one-tap invites, and after a friend’s first payment both get a month of subscription.
  • Russian and English interface; photos are compressed on the phone, and iPhone HEIC images are converted.

TechnologyNext.jsReactTypeScriptTailwindSupabaseRedisOpenAIRecharts

Result

  • 3 ways to log food

    What it means

    A photo of the plate, a label screenshot or a few words: AI works out the dish, its weight and macros in seconds.

  • 6 price A/B test groups

    What it means

    Different subscription prices and trials were tested on live traffic, with results visible as revenue in the in-house analytics.

  • 5 in-house analytics reports

    What it means

    Overview, performance, retention, goals and funnels, with no third-party counters and attribution via the launch link.

Screens

  • FoodTrack AI: Daily progress: calories, macros and the meal feed
    Daily progress: calories, macros and the meal feed
  • FoodTrack AI: AI recognised the dish from a photo: weight, macros and a verdict
    AI recognised the dish from a photo: weight, macros and a verdict
  • FoodTrack AI: A daily AI report with advice
    A daily AI report with advice
  • FoodTrack AI: Weight trend and calories by day
    Weight trend and calories by day
  • FoodTrack AI: A personal plan after onboarding
    A personal plan after onboarding
  • FoodTrack AI: A Telegram Stars subscription
    A Telegram Stars subscription

In-house analytics

To see where users come from and what brings in payments, we built the app its own analytics system: events are sent straight from the mini app, and a separate dashboard turns them into reports.

  • Source, campaign, creative and A/B group travel in the mini app’s launch link and stay attached to the user.
  • Sessions, page views, onboarding steps, food logging, paywall views and payments, with platform and geo data.
  • Reports: overview, performance broken down by any dimension with revenue, cohort retention, goals and funnels.
  • Dashboard sign-in via Telegram, for admins only.
  • FoodTrack AI: Overview: users, sessions, purchases and revenue
    Overview: users, sessions, purchases and revenue
  • FoodTrack AI: The onboarding funnel step by step
    The onboarding funnel step by step
  • FoodTrack AI: Cohort retention
    Cohort retention
  • FoodTrack AI: Performance by A/B group and source
    Performance by A/B group and source

The screenshots show demo data.

What shall we build together?

Tell us what you want to launch or improve. We’ll discuss the task and suggest the next step.