Georgian Recipe App - WR134/Knowledge-Base GitHub Wiki
Georgian Recipe App — Product & Business Analysis
Summary
There is currently no modern, feature-rich recipe application in the Georgian language. The only existing attempt — Receptebi.ge Kulinaria — is a minimal no-code app last updated in 2023 with no meaningful functionality. International platforms like Tasty or Allrecipes do not support Georgian, have no local pricing data, and lack any awareness of the Georgian ingredient market. The space is completely open.
This app fills that gap as a bilingual (Georgian-first, English-toggle) recipe platform built for the Georgian market. It launches with 200 curated recipes — both Georgian and international — and goes well beyond a standard recipe listing. Every recipe includes full nutritional macros, cook time, difficulty level, cooking techniques, and approximate ingredient pricing in GEL. Users can see what a meal costs in total and per serving, helping with budgeting and meal planning.
One of the more distinctive features is a crowdsourced "where to buy" system. Users can submit and upvote locations for hard-to-find or niche ingredients — useful for anyone who has tried finding specialty items outside of standard grocery chains. This creates a community-driven ingredient map that gets more valuable as the user base grows.
The app also includes AI-powered recipe search in Georgian. Users can describe what they want in natural language — available ingredients, budget constraints, dietary preferences, time limits — and receive relevant recipe suggestions. This goes beyond keyword matching: the AI understands context and intent, making recipe discovery faster and more flexible.
Social features are built in from the start. Users can create accounts, save favorites, submit their own recipes, and leave ratings and comments on others. A clear distinction between verified (developer-curated) and community-submitted recipes maintains content quality. The platform supports community moderation through voting, reporting, and automated filtering.
Future possibilities include partnerships with well-known Georgian chefs (verified profiles with exclusive content), direct integration with Glovo and Wolt for ingredient purchasing, AI-generated weekly meal plans with macro targets, and expansion into neighboring markets (Armenia, Azerbaijan) or to the global Georgian food community through the English version.
From a business standpoint, the app runs on modern serverless infrastructure that costs under $100/month even at tens of thousands of active users. Revenue comes from multiple streams: in-app advertising, a low-cost premium subscription (5 GEL/month for ad-free experience and advanced AI features), affiliate commissions from grocery delivery referrals, and brand sponsorships from local food companies. The financial risk is low — the primary investment is development time, while infrastructure costs remain negligible until the app is already generating revenue.
Georgia has 3.7 million people, 80%+ smartphone penetration, and 95% Facebook adoption. The market is reachable, the tools to build this are accessible, and there is no competition.
Core Features & Functionality
Recipe Content
- Curated recipe library: 200 verified recipes at launch, covering Georgian classics (khachapuri, khinkali, lobio, pkhali, etc.) and popular international cuisine
- Bilingual content: Georgian-first with full English translation, user-toggleable
- Detailed recipe data per entry:
- Step-by-step cooking instructions with photos
- Full nutritional macros (calories, protein, carbs, fat per serving)
- Servings per person with adjustable portions
- Cook time and prep time
- Cooking techniques and difficulty level
- Complete ingredient list with quantities
Pricing & Availability
- Approximate ingredient pricing: Displayed in GEL as ranges (e.g., "chicken breast: 12–18 GEL/kg"), sourced from Geostat inflation data, periodic Glovo/Wolt scraping, and community corrections
- Total meal cost: Automatically calculated from ingredient prices — both total and per-serving
- Price disclaimer: Clearly communicated as approximate, with "last updated" timestamps
- Update cycle: Weekly to monthly, with community flagging for outdated prices
- "Where to buy" system: Crowdsourced locations for niche/hard-to-find ingredients (e.g., Mexican chili, specialty spices). Users submit locations, others upvote/downvote for accuracy. Includes "last confirmed" dates and automatic expiry for stale entries
Search & Discovery
- Standard search: Text-based recipe search with filters (cuisine type, cook time, difficulty, dietary tags)
- AI-powered natural language search: Users describe what they want in plain Georgian — "something easy with 2-3 ingredients I have" or "quick dinner under 30 minutes for 15 GEL" — and receive intelligent recommendations
- Ingredient-based matching: Select ingredients you have on hand, get matching recipes ranked by relevance
- Dietary filtering: Vegetarian, vegan, gluten-free, low-carb, and other dietary restriction tags
Social & Community Features
- User accounts and profiles
- Save favorite recipes to personal collections
- User-submitted recipes: Community members can create and share their own recipes with photos, instructions, and ingredient lists
- Ratings and reviews: Like, comment, and rate recipes
- "Verified" badge: Developer-curated recipes clearly distinguished from community submissions
- Community moderation: Upvote/downvote system, community reporting, automated AI filtering for spam and inappropriate content
- Recipe sharing: Direct sharing to social media and messaging apps
Future Features (Post-Launch)
- Chef partnerships: Verified profiles for recognizable Georgian chefs with exclusive recipes
- Grocery delivery integration: "Buy all ingredients" button linking to Glovo/Wolt carts
- Weekly meal planner: AI-generated meal plans based on budget, dietary goals, and macro targets
- Recipe substitution suggestions: "Make this recipe healthier" or "I don't have X, what can I use instead?"
- Personalized recommendations: Based on cooking history and preferences
Market Analysis
Georgian Digital Market Overview
| Metric | Value | Source |
|---|---|---|
| Population | ~3.7–3.9 million | Geostat, 2025 |
| Smartphone penetration | 80%+ | Geostat, 2024 |
| Internet users | ~3.12 million (81.9%) | DataReportal |
| Average monthly salary | Geostat, Q3 2025 | |
| Consumer food spend per capita/year | ~$1,330 USD | Statista |
| In-app advertising market (Georgia) | $26M (2025), projected $35M by 2030 | BYYD |
| Key smartphone age group | 25–34 years (35.9% of users) | Start.io |
| Urban population | ~62% | UN estimates |
Social Media Landscape
| Platform | Users in Georgia | % of Population | Primary Demographic |
|---|---|---|---|
| 3.7M | 95.2% | All ages, largest group 25–34 | |
| TikTok | 2.3M (adults) | 80.5% of adults | Skews younger, 56.1% male |
| 1.78M | 45.7% | 25–34 largest group, 58.4% female | |
| Messenger | 3.0M | 77.7% | Aligned with Facebook |
Sources: NapoleonCat, DataReportal Digital 2025: Georgia
Facebook's near-total penetration (95.2%) makes it the primary marketing channel. TikTok's 80.5% adult reach makes it the strongest channel for viral recipe content targeting younger demographics.
Grocery & Food Delivery Landscape
Major grocery chains:
- Nikora: 600+ stores, 35% of meat market, largest chain. Has a loyalty app but no online ordering
- Carrefour: 95 outlets, wide product range, uses Glovo for delivery
- Goodwill: First hypermarket chain, strong in ready-made meals and imports
- Others: Ori Nabiji (17% market share), Spar, Fresco, Agrohub, Ioli
Delivery platforms:
- Glovo: Active in Georgia, delivers from Carrefour, Goodwill, Fresco. Has a QCommerce API for grocery partners
- Wolt: Delivers from Nikora, Fresco, and others. Now part of Delivery Hero
- QuickShipper: Tbilisi-based company bridging Glovo and Wolt integrations
Competition Analysis
Direct competitors: Effectively none.
| Competitor | Status | Features | Threat Level |
|---|---|---|---|
| Receptebi.ge Kulinaria | Abandoned (v1.1, last updated 2023) | Basic recipe listings, built on GoodBarber (no-code). No macros, no pricing, no AI, no social features | Negligible |
| kulinaria.ge (website) | Active, web only | Recipe articles | Low — no app, no structured data |
| foodhub.ge (website) | Active, web only | Recipe listings | Low — no app, no structured data |
| gemrielia.ge, gurman.ge (websites) | Active, web only | Recipe articles | Low |
Indirect competitors:
| Competitor | Reach | Threat Level | Differentiation |
|---|---|---|---|
| Georgian food bloggers on TikTok/Instagram/Facebook | High reach, free content | Medium | They offer inspiration, not structured data. No macros, no pricing, no search, no ingredient matching |
| International recipe apps (Tasty, Allrecipes, Cookpad) | Available but not localized | Low | Not in Georgian, no local pricing, no Georgian recipes |
Key insight: The market gap is not just "no app exists" — it's that no platform combines Georgian-language recipes with structured nutritional data, local pricing, and intelligent search. This is a genuine first-mover opportunity.
Technical Architecture
Recommended Stack
| Layer | Technology | Role |
|---|---|---|
| Database | Supabase (PostgreSQL) | Primary data store — recipes, users, comments, prices, favorites |
| Auth | Supabase Auth | User registration, login, social auth (Google, Facebook) |
| Image Storage | Cloudflare R2 | Recipe photos, user avatars — zero egress fees |
| AI / Vector Search | Supabase pgvector + embedding model | Semantic recipe search, ingredient matching |
| LLM API | DeepSeek V3.2 (primary) / Gemini 3.0 Flash (fallback) | Natural language recipe queries |
| Image Processing | Sharp (Node.js) | Compress, resize, convert uploads to WebP |
| Backend | Node.js + Express | API layer, business logic, image upload pipeline |
| Frontend (Web) | React + Tailwind CSS | Web application |
| Frontend (Mobile) | React Native | iOS and Android apps |
| Hosting | Vercel or Railway | Backend and web frontend hosting |
Why Supabase (Primary Choice)
PostgreSQL is the natural fit for recipe data. Recipes have ingredients. Ingredients have prices at multiple stores. Users have favorites. Recipes have comments from users. This is textbook relational data handled natively with JOINs.
Advantages:
- Relational model: Natural for recipes → ingredients → prices → users → comments relationships
- PostgreSQL full-text search: Can support Georgian with custom text search configuration and trigram similarity (pg_trgm) for fuzzy matching
- pgvector extension: Built-in vector search for AI features — no separate service needed
- Built-in auth: Free, supports social login (Google, Facebook)
- Row Level Security: Secure user data without backend middleware
- Generous free tier: 50K MAU, 500MB database, 1GB storage on free plan
- Mature self-hosting: If cost becomes an issue, can migrate to self-hosted PostgreSQL
- React/React Native support: Official JS client works in both environments
Pricing:
| Scale | Plan | Cost |
|---|---|---|
| 0–50K MAU | Free tier (500MB DB, 1GB storage) | $0/month |
| 50K+ MAU | Pro ($25/month base) | ~$25–75/month |
| 100K+ MAU | Pro + usage overages | ~$75–200/month |
Alternative: Convex
Convex is a real-time backend platform with a document-based transactional database and excellent TypeScript developer experience. It is a viable alternative worth considering.
Convex Advantages:
- Reactive by default: Real-time updates (live comment counts, instant like feedback) with zero boilerplate. Best-in-class for social features
- TypeScript end-to-end: Backend functions and frontend hooks share types — fewer bugs
- Developer experience: Write a backend function, call it with a hook. No REST API boilerplate
- Built-in vector search: Native vector search for AI features
- Built-in file storage: Handles recipe photo uploads natively
- Self-hostable: Recently open-sourced
Convex Disadvantages:
- Georgian text search limitations: Built-in full text search uses Tantivy's SimpleTokenizer, optimized for Latin scripts. Georgian tokenization works (spaces between words) but no stemming — "ხინკალი" won't match "ხინკლის"
- Document model for relational data: Recipes → ingredients → prices requires manual "joins" in functions. Less natural than SQL for this data model
- Auth is BYOA: Requires Clerk or Auth0 integration (additional cost/complexity vs. Supabase's built-in auth)
- Function call pricing model: Social features (likes, comments, real-time updates) generate many function calls. Cost scales with interaction volume, not just user count
- Smaller ecosystem: Fewer community resources, tutorials, and third-party integrations compared to PostgreSQL/Supabase
Convex Pricing:
| Scale | Plan | Cost |
|---|---|---|
| 0–5K MAU | Free tier (1M function calls/month) | $0/month |
| 5K–20K MAU | Starter (pay-as-you-go) | ~$10–50/month |
| 20K–100K MAU | Pro ($25/member/month + usage) | ~$50–200/month |
Bottom Line: Supabase is recommended for this app due to the naturally relational data model, superior Georgian text search support, built-in auth, and more predictable pricing. Convex is the better choice only if real-time social features are the top UX priority and the team has prior Convex experience.
Image Storage Architecture
Recipe photos are the primary storage cost driver. Images must be stored separately from the database.
The problem with storing images in Supabase:
- Free tier: 1GB storage, 2GB bandwidth/month
- 200 recipes × 3 photos × 250KB = 150MB at launch
- A single user browsing 20 recipes downloads ~5MB. 400 users/day = 2GB bandwidth — free tier exhausted on day one
Solution: Cloudflare R2
| Free Tier | Beyond Free | |
|---|---|---|
| Storage | 10GB | $0.015/GB/month |
| Bandwidth/egress | Unlimited, $0 forever | $0 |
| Upload operations | 1M/month | $4.50/million |
| Read operations | 10M/month | $0.36/million |
Zero egress fees means no matter how many users view recipe photos, bandwidth costs nothing. At 50K MAU with ~15GB of images stored, total R2 cost is approximately $0.08/month.
Image upload pipeline:
User uploads photo
→ Node.js backend receives file
→ Sharp library processes:
- Resize to max 1200px width
- Convert to WebP (quality 80) → ~200-300KB
- Generate 400px thumbnail → ~30-50KB
→ Upload both versions to Cloudflare R2
→ Store CDN URLs in Supabase alongside recipe data
Storage projections:
| Content | Year 1 | Year 2 |
|---|---|---|
| Curated recipes (200-500) | ~200MB | ~500MB |
| User-submitted recipes | ~1.2GB | ~6GB |
| User avatars | ~500MB | ~2.5GB |
| Total | ~2GB | ~9GB |
| R2 cost | $0 (under 10GB free) | ~$0.14/month |
Georgian Language Technical Considerations
Georgian script (ქართული/Mkhedruli) presents specific technical challenges:
- Tokenization: Georgian uses spaces between words (unlike CJK languages), so standard whitespace-based tokenizers work for basic splitting
- No case distinction: Georgian script has no uppercase/lowercase, eliminating case-sensitivity issues
- Agglutinative morphology: Words have many prefixes and suffixes. Searching "ხინკალი" (khinkali) won't match "ხინკლის" (of khinkali) without stemming
- No standard stemmer exists: PostgreSQL doesn't ship with a Georgian stemmer. Mitigation: use trigram similarity (pg_trgm) for fuzzy matching, and vector search for semantic matching
- LLM comprehension: Gemini 3.0 Flash handles Georgian well (confirmed by testing). DeepSeek V3.2 and Qwen3 need to be tested for Georgian food terminology accuracy
AI Features & Cost Analysis
AI Feature Tiers
| Priority | Feature | Technical Approach | Complexity |
|---|---|---|---|
| P0 (Launch) | Ingredient-based matching: "I have X, Y, Z — what can I make?" | Vector search against recipe embeddings | Medium |
| P0 (Launch) | Natural language search: "quick Georgian dinner under 30 min" | LLM-powered query understanding + recipe matching | Medium |
| P1 (Post-launch) | Budget-based search: "I have 20 GEL, what can I cook?" | LLM + price data context | Medium |
| P1 | Dietary restriction filtering | Tag-based filtering (no AI needed) | Low |
| P2 (Growth) | Weekly meal planner with macro targets | Multi-step LLM reasoning + optimization | High |
| P2 | Recipe substitution suggestions | LLM feature | Medium |
| P3 (Future) | Personalized recommendations based on history | Recommendation engine | High |
LLM Options (2026 Pricing)
| Model | Input/1M tokens | Output/1M tokens | Cost per recipe query | Georgian support | Notes |
|---|---|---|---|---|---|
| DeepSeek V3.2-Exp | $0.028 | ~$0.11 | ~$0.0001 | Needs testing | MIT licensed, cache hits at $0.014/1M (90% discount) |
| Qwen3 (small, via SiliconFlow) | $0.05 | ~$0.05 | ~$0.0002 | 119 languages, likely includes Georgian | Apache 2.0, self-hostable |
| Gemini 2.0 Flash | $0.10 | $0.40 | ~$0.0004 | Good | Free tier: 1,000 requests/day |
| Gemini 3.0 Flash | $0.50 | $3.00 | ~$0.0025 | Excellent (tested) | Frontier quality, 25x more expensive than DeepSeek |
| Self-hosted Qwen3-8B | $0 (infra only) | $0 (infra only) | ~$0 | 119 languages | Runs at 25 tok/sec on consumer laptops (8-12GB VRAM). Server cost ~$50-100/month |
Sources: Google Gemini Pricing, DeepSeek Pricing, SiliconFlow, PricePerToken
Monthly AI Cost Projections
| MAU | Est. LLM queries/month | DeepSeek V3.2 | Gemini 2.0 Flash | Gemini 3.0 Flash |
|---|---|---|---|---|
| 10K | 50,000 | ~$5 | ~$20 | ~$125 |
| 50K | 250,000 | ~$25 | ~$100 | ~$625 |
| 100K | 500,000 | ~$50 | ~$200 | ~$1,250 |
Embedding costs (for vector search): Essentially $0 regardless of scale. OpenAI text-embedding-3-small costs $0.02/1M tokens. Alternatively, self-host Qwen3-Embedding-0.6B for zero API cost.
Recommended AI Strategy
- Vector search (embeddings) for all users — ingredient matching and basic semantic search. Cost: effectively $0
- DeepSeek V3.2 as default LLM — test with Georgian first. If accuracy is acceptable, this becomes the primary engine at negligible cost
- Gemini 3.0 Flash as fallback — use for complex queries or if DeepSeek underperforms with Georgian
- Self-hosted Qwen3-8B for development — zero-cost testing and iteration during development phase
Given 2026 pricing, AI costs are low enough that basic AI search can be offered to all users for free. Advanced features (meal planning, multi-step reasoning) can be gated behind premium if cost control is needed, but even at scale the costs remain manageable.
Monetization Strategy
Revenue Model: Hybrid
The app will use a hybrid monetization approach combining multiple revenue streams. Given Georgia's average monthly salary of ~$800 USD, aggressive subscription pricing is not viable. The strategy prioritizes non-subscription revenue while offering a low-cost premium tier.
Revenue Streams
1. In-App Advertising (Primary)
- Non-intrusive ad placements between recipe listings and on recipe detail pages
- Contextually relevant: food brands, grocery stores, kitchen products
- Georgian in-app ad market: $26M (2025), growing to $35M by 2030
- Estimated CPM: $2 (conservative for Georgian market)
2. Premium Subscription (Secondary)
- Price: 5 GEL/month (~$1.80 USD)
- Features: Ad-free experience, unlimited AI-powered search, meal planning, advanced macro tracking
- Target conversion: 2-3% of MAU (conservative for price-sensitive market)
3. Grocery Affiliate Revenue
- "Buy all ingredients" button linking to Glovo/Wolt carts
- Commission: estimated 3-5% of order value
- Requires partnership negotiation (post-launch, once user base provides leverage)
- Glovo's QCommerce API enables technical integration
4. Brand Sponsorships
- Local food brands (Nikora products, Georgian wine brands, spice companies) sponsor featured recipes
- Featured recipe placement: brand's products highlighted in curated recipes
- Estimated: 500-2,000 GEL/month per brand partnership
5. Chef Partnerships (Future)
- Verified chef profiles with exclusive recipe content
- Potential for cooking class upsells or premium chef content
- Primary value is marketing/credibility rather than direct revenue
What Stays Free
Core recipe browsing, search, favorites, ingredient pricing, "where to buy," and basic AI ingredient matching will remain free. The free tier must deliver enough value to drive adoption and word-of-mouth. Paywalling core recipe content would kill growth in a market conditioned to free online recipes.
Revenue & Cost Projections
Infrastructure Cost Model
| Component | 10K MAU | 50K MAU | 150K MAU |
|---|---|---|---|
| Supabase (database + auth) | $0 (free) | $25 | $75 |
| Cloudflare R2 (images) | $0 (free) | $0 (under 10GB) | $5 |
| Hosting (Vercel/Railway) | $0 (free) | $20 | $40 |
| AI API (DeepSeek V3.2) | $5 | $25 | $50 |
| OpenAI Embeddings | $1 | $3 | $5 |
| Domain + misc services | $10 | $10 | $10 |
| Total infrastructure | ~$16/month | ~$83/month | ~$185/month |
Optional marketing spend:
| 10K MAU | 50K MAU | 150K MAU | |
|---|---|---|---|
| Facebook/TikTok ads | $100-200 | $300-500 | $500-1,000 |
Revenue Projection — Conservative Scenario
Assumptions (deliberately conservative):
- Ad CPM: $2 (low end for Georgian market)
- DAU/MAU ratio: 25%
- Ad views per session: 3
- Sessions per DAU per day: 1.5
- Premium conversion rate: 2% (below industry average of 3-5%)
- Premium price: 5 GEL/month (~$1.80 USD)
- Affiliate commission: deferred until post-partnership (not included in Year 1)
- Brand sponsorships: conservative count, 1,000 GEL/month per brand
Year 1 — 10K MAU (achievable with active marketing)
| Revenue Source | Monthly |
|---|---|
| In-app ads (337K impressions x $2 CPM) | ~$20 |
| Premium subscriptions (200 users x $1.80) | ~$360 |
| Brand sponsorship (1 brand x 1,000 GEL) | ~$360 |
| Affiliate revenue | $0 (not yet established) |
| Total revenue | ~$740/month |
| Total costs (infra + $150 marketing) | ~$166/month |
| Net profit | ~$574/month (~1,600 GEL) |
Year 2 — 50K MAU (1.3% of smartphone users)
| Revenue Source | Monthly |
|---|---|
| In-app ads (1.7M impressions x $2 CPM) | ~$100 |
| Premium subscriptions (1,000 users x $1.80) | ~$1,800 |
| Brand sponsorships (2 brands x 1,000 GEL) | ~$720 |
| Affiliate grocery referrals | ~$200 |
| Total revenue | ~$2,820/month |
| Total costs (infra + $400 marketing) | ~$483/month |
| Net profit | ~$2,337/month (~6,500 GEL) |
Year 3 — 150K MAU (4% of smartphone users, optimistic but achievable)
| Revenue Source | Monthly |
|---|---|
| In-app ads (5M impressions x $2 CPM) | ~$300 |
| Premium subscriptions (3,000 users x $1.80) | ~$5,400 |
| Brand sponsorships (3 brands x 1,500 GEL) | ~$1,620 |
| Affiliate grocery referrals | ~$800 |
| Total revenue | ~$8,120/month |
| Total costs (infra + $750 marketing) | ~$935/month |
| Net profit | ~$7,185/month (~20,000 GEL) |
Key Financial Takeaways
- Break-even is nearly instant. Infrastructure costs are under $20/month at launch. Even minimal monetization covers this
- Margins are high. Modern serverless infrastructure keeps costs low regardless of scale. At 50K MAU, margins exceed 80%
- Premium subscriptions drive most revenue. Even at a conservative 2% conversion rate, subscriptions outperform ads significantly. A small increase in conversion rate (2% → 3%) would add ~$900/month at 50K MAU
- The revenue ceiling is real. Georgia's population caps total addressable users at ~3M smartphone users. At maximum penetration (10-15%), monthly revenue likely caps at $15-25K. This is a sustainable small business, not a venture-scale opportunity — unless expanded to neighboring markets (Armenia, Azerbaijan)
- Financial risk is minimal. Total investment to reach MVP is essentially developer time. Infrastructure costs remain negligible until the app is already generating revenue
Marketing & User Acquisition Strategy
Phase 1: Pre-Launch & First 1,000 Users
- Georgian Facebook food groups: Post curated recipes with a hook — "get full recipe + macros + pricing in the app." Multiple active food groups exist with engaged audiences. Free, organic
- TikTok recipe shorts: 30-second cooking videos of Georgian classics. One viral video can drive thousands of downloads. TikTok reaches 80.5% of Georgian adults
- Instagram food photography: High-quality finished dish photos linking to app. Target the 25–34 female demographic (58.4% of Georgian Instagram users)
- Personal network seeding: Get 50-100 friends and family using the app, leaving ratings and submitting recipes to populate initial community content
Phase 2: Growth to 10K Users
- Facebook ads (paid): Georgian Facebook ads are cheap compared to Western markets. Estimated CPM $1-3. A budget of $100-200/month can reach 50K-100K impressions
- Food blogger partnerships: Early access for 5-10 Georgian food bloggers/influencers. They review the app, post about it, gain credibility
- PR in Georgian media: "First Georgian-language recipe app" is a newsworthy story for local tech and lifestyle media
Phase 3: Scale to 50K+ Users
- Chef partnerships: A recognizable Georgian chef's verified profile drives press coverage and credibility. Even one partnership changes the marketing narrative
- Glovo/Wolt cross-promotion: If "buy ingredients" integration is built, delivery platforms may co-promote — it drives orders for them
- Word-of-mouth flywheel: The "where to buy" feature creates organic sharing — users share niche ingredient locations with friends, bringing them into the app
- App Store Optimization: Georgian-language keywords have virtually zero competition. Basic ASO will rank the app for any recipe-related Georgian search
Marketing Budget Recommendations
| Phase | Monthly Budget | Primary Channels |
|---|---|---|
| Pre-launch | $0 | Organic social, personal network |
| Year 1 (growth) | $100-200 | Facebook ads, blogger partnerships |
| Year 2 (scale) | $300-500 | Facebook + TikTok ads, chef partnerships |
| Year 3 (maturity) | $500-1,000 | Diversified: ads, partnerships, PR |
Risk Analysis
Market Risks
| Risk | Severity | Likelihood | Description | Mitigation |
|---|---|---|---|---|
| Small addressable market | High | Certain | Georgia's population of 3.7M caps total potential users. Even at 10% penetration, that's 370K users. Revenue ceiling is real. | Expand to Armenian and Azerbaijani markets. English version attracts Georgian food enthusiasts globally. Position as regional Caucasus recipe platform long-term. |
| Low willingness to pay | High | High | Average salary ~$800/month. Georgian users are price-sensitive and accustomed to free online recipes. | Keep core features free. Premium at 5 GEL/month (~$1.80) is impulse-purchase pricing. Primary revenue from ads and affiliates, not subscriptions. |
| Social media as indirect competition | Medium | High | Georgian food bloggers on TikTok/Instagram/Facebook post recipes for free with large followings. Users may not see the need for a dedicated app. | Differentiate on structured data: macros, pricing, "where to buy," AI search, organized collections. Social media cannot offer recipe search, nutritional data, or ingredient matching. |
| Copycat risk | Medium | Medium | A well-funded company (or Glovo itself) could clone the concept once market viability is proven. | First-mover advantage matters in a small market. Crowdsourced "where to buy" data is a defensible moat — it gets better with more users and cannot be replicated overnight. Build community loyalty. |
| Market timing | Low | Low | Economic downturn in Georgia could reduce consumer spending and ad budgets. | Recipe apps actually benefit from cost-conscious consumers — "cook at home for less" messaging becomes more relevant. |
Technical Risks
| Risk | Severity | Likelihood | Description | Mitigation |
|---|---|---|---|---|
| AI accuracy in Georgian | High | Medium | Georgian is a low-resource language. LLMs may misunderstand ingredient names, quantities, or cooking terminology. A recommendation for the wrong dish erodes trust. | Test extensively before launch with real Georgian food queries. Use vector search (language-agnostic) as primary search method. LLM as supplement. Fall back to Gemini 3.0 Flash (confirmed good Georgian support) if cheaper models fail. |
| Price data staleness | Medium | High | Weekly/monthly price updates may not reflect reality. A key ingredient doubles in price due to seasonal changes or supply issues, but the app shows old data. | Display "last updated" dates prominently. Use ranges, not exact prices. Clear disclaimer. Community corrections ("flag this price as outdated"). Treat pricing as "helpful estimate," not authoritative source. |
| Scraping fragility | Medium | Medium | Glovo/Wolt may change their website structure or APIs, breaking the price data pipeline. | Don't depend on scraping as the sole source. Use Geostat official data as baseline. Community corrections as fallback. Treat scraping as supplementary enhancement. |
| Georgian text search limitations | Medium | High | No standard stemmer exists for Georgian in any database engine. Basic text search won't handle morphological variations (e.g., "ხინკალი" vs "ხინკლის"). | Use PostgreSQL trigram similarity (pg_trgm) for fuzzy matching. Vector search handles semantic similarity regardless of morphology. Consider building a basic Georgian suffix-stripping function. |
| React to React Native migration | Medium | Medium | The planned approach of building a React web app first and converting to React Native involves significant frontend rewrite (~60-70%). UI components don't transfer directly. | Plan the architecture for cross-platform from the start. Share backend logic, API clients, and state management. Accept that UI layer will be rewritten. The Node.js backend transfers fully. |
Operational Risks
| Risk | Severity | Likelihood | Description | Mitigation |
|---|---|---|---|---|
| User-generated content quality | High | Medium | Bad recipes damage brand trust. An incorrect khachapuri recipe gets attributed to the platform. Spam and low-effort posts dilute content quality. | Clear visual distinction between "verified" and "community" recipes. Rating/voting system surfaces quality. Require minimum detail (photo, at least 3 ingredients, instructions) for submission. |
| Moderation burden | Medium | High | Team of three handling spam, inappropriate content, inaccurate recipes, and "where to buy" disputes across a growing platform. | Automated filters (profanity, spam detection, AI content screening). Community reporting + downvote thresholds trigger auto-hide. Invest in moderation tooling early. |
| "Where to buy" data decay | Medium | High | User-submitted store locations for niche ingredients go stale. Stores close, products rotate, seasonal items disappear. | Show "last confirmed" timestamps. Periodic prompts: "Is [product] still available at [location]?" Auto-expire entries after 3-6 months without reconfirmation. Incentivize updates (gamification, points). |
| Content creation bottleneck | Medium | Medium | Curating 200 quality recipes with accurate macros, pricing, and photos requires significant upfront effort. AI-assisted creation needs human verification. | Prioritize 50 flagship Georgian recipes for maximum quality. Use AI for macro calculations and initial drafts. Verify the most important recipes by actually cooking them. Scale remaining 150 with lighter verification. |
Business Model Risks
| Risk | Severity | Likelihood | Description | Mitigation |
|---|---|---|---|---|
| Ad revenue depends on volume | High | High | Georgian in-app ad CPMs are low (~$1-3). Meaningful ad revenue requires 50K+ MAU, which takes time to achieve. | Treat ads as supplementary. Focus on subscriptions and sponsorships for early revenue. Optimize ad placement and targeting as user base grows. |
| Affiliate deals aren't guaranteed | Medium | Medium | Glovo/Wolt may not offer affiliate programs, or commissions may be too small to matter. Partnership requires negotiation leverage (user base). | Build "buy ingredients" as UX value first — users benefit from convenience regardless of affiliate revenue. Monetize when there is leverage (proven user engagement with the feature). |
| Premium conversion uncertainty | Medium | Medium | 2-3% is industry average, but Georgian market may convert lower given price sensitivity and cultural expectation of free content. | Aggressively test premium value proposition. A/B test pricing: 3 GEL vs 5 GEL vs 7 GEL. Make AI features premium-only if needed to create clear value gap. Offer annual discount. |
| Revenue doesn't justify full-time work | Medium | High | At conservative projections, Year 1 revenue (~$740/month) doesn't support a three-person team full-time. | Treat as a side project until 30K+ MAU proves viability. Infrastructure costs are negligible, so financial risk is limited to time investment. |
Legal Risks
| Risk | Severity | Likelihood | Description | Mitigation |
|---|---|---|---|---|
| Recipe content copyright | Medium | Low | Scraping instructions verbatim from Georgian recipe websites could lead to claims. | Ingredient lists are factual (not copyrightable). Rewrite all instructions in original voice. Use AI to generate macro calculations. Use scraped data only as reference. |
| Price accuracy liability | Low | Medium | Users make purchasing decisions based on displayed prices that turn out to be significantly wrong. | Prominent disclaimer: "Approximate prices for reference only. Actual prices may vary." Display as ranges with last-updated dates. |
| Data privacy (GDPR equivalent) | Medium | Low | Processing user data, especially routing queries through external AI APIs. Georgia's data protection law aligns with EU standards. | Anonymize all AI API queries — don't send user identifiers. Clear privacy policy. Don't share personal data with third parties. Store minimal PII. |
Development Phases
Phase 1: Foundation & MVP
- Set up Supabase project (database schema, auth, RLS policies)
- Set up Cloudflare R2 for image storage
- Build Node.js backend (API layer, image processing pipeline)
- Build React web app: recipe browsing, search, user accounts, favorites
- Create and verify 50 flagship Georgian recipes (curated, with photos, macros, pricing)
- Implement basic text search with Georgian fuzzy matching
- Implement vector search for ingredient-based matching
- Deploy web app
Phase 2: AI & Content Expansion
- Integrate LLM-powered natural language search (DeepSeek V3.2 / Gemini 3.0 Flash)
- Expand recipe library to 200 curated recipes
- Implement ingredient pricing system (Geostat data + manual updates)
- Implement "where to buy" crowdsourced system with upvote/downvote
- Add dietary restriction tags and filtering
- Launch premium subscription (ad-free + AI features)
Phase 3: Social & Community
- User-submitted recipes with photo upload
- Comments and ratings system
- Community moderation tools (reporting, auto-hide, AI filtering)
- In-app advertising integration
- Recipe sharing to social media
Phase 4: React Native & Mobile Launch
- Build React Native mobile app (iOS + Android)
- Push notifications (new recipes, price alerts, community activity)
- Optimize mobile UX for kitchen use (larger buttons, voice-compatible, screen-awake mode)
- App Store / Google Play launch
Phase 5: Growth Features
- Chef partnership profiles with verified badges
- Glovo/Wolt "buy all ingredients" integration
- Budget-based recipe search ("I have 20 GEL")
- Weekly meal planner with macro targets
- Personalized recommendations
- Expand to English-speaking audience / neighboring markets
Summary
This app addresses a clear, uncontested market gap: Georgia has no modern, feature-rich recipe application in its own language. The combination of curated and user-generated recipes, nutritional data, local ingredient pricing, crowdsourced "where to buy" data, and AI-powered search in Georgian creates a product that no existing platform — local or international — currently offers.
The financial model is low-risk: infrastructure costs start near zero and remain under $200/month at scale. Revenue comes from multiple streams (ads, subscriptions, affiliates, sponsorships), and even conservative projections show profitability from launch. The primary risk is not financial but executional — building a quality product, maintaining fresh data, and growing to critical mass in a small but underserved market.
The technical stack (Supabase + Cloudflare R2 + React/React Native + modern LLM APIs) is mature, cost-effective, and well-suited to a three-person team. AI API costs in 2026 have dropped to a level where intelligent features can be offered to all users without significant cost burden.
The path forward is clear: build the MVP with 50 flagship recipes, validate with real users, expand content and features iteratively, and grow through Georgia's dominant social media channels — Facebook, TikTok, and Instagram — where food content already thrives but no dedicated platform captures it.