Skip to main content
Case StudiesProjectsAbout
Case StudiesProjectsAbout
Product · Consumer Social — Case Study

Cravia: Rate the Dish, Not the Restaurant

A dish-level food review app for India that scores individual menu items on taste, portion, and value — reframing the atomic unit of review from the restaurant to the dish to build a data moat no incumbent can easily copy.

7 min read·
Product StrategyConsumer SocialGamification0→1
Dish
Unit of review
not the restaurant
Gurugram
Launch market
one city deep, 10 defined areas
Pre-launch
Stage
feature-complete PWA in active development
My Role
Product definition, core loop, gamification & monetization model
Timeline
2026, active development
Team
Solo
Platform
Progressive Web App (Next.js, mobile-first)
Stack
Consumer social · food discovery
Status
Pre-launch · feature-complete PWA

Dish-level food reviews for India. "Tells you what to order, not just where to eat."

The Problem

Every major food platform in India — Zomato, Swiggy, Google Maps — reviews at the restaurant level. A restaurant sits at 4.2 stars, but that number blends the legendary biryani with the forgettable dessert, a birthday dinner with a rushed lunch, and dishes that left the menu months ago. For the one question a diner actually cares about — "what should I order here?" — a star rating gives essentially zero signal.

Meanwhile, the behavior already exists: Indians recommend dishes, not restaurants. "You HAVE to try the mutton sukka at XYZ" travels through WhatsApp groups and Instagram Stories every day. It's dish-specific, photo-forward, and social — but it's trapped in ephemeral, unsearchable channels.

◈Core Insight

A dish at a specific restaurant is a repeatable product — same recipe, same kitchen, same price. That makes it far more reliable to review than a "restaurant experience," which varies by table and time of day. When 47 people say a butter chicken scores 4.6 on taste but 3.2 on portion, that's genuinely actionable.

The Bet

●The one reframing that is the whole product

Make the dish — not the restaurant — the atomic unit of review. This single reframing creates a data asset no competitor has: a scored menu for every restaurant, with structured sub-ratings, tags, and photos per dish. Zomato can't easily copy it without cannibalizing its existing review model and UX.

Target User

Primary — The Urban Foodie (22–35). Eats out 3–6× a week, photographs meals, trusts friends over platform ratings, and will cross the city for a specific dish. Food is identity. Frustrated by the absence of reliable dish-specific info online.

Secondary — The Restaurant Owner. Wants to know which dishes are winning and losing — feedback no platform gives them today.

Tertiary — Admin / Moderator. Manages content quality and the coupon catalogue.

Core Product Decisions

◆Decision — Review structure

Structured micro-reviews over free text. Every review requires a photo + three sub-ratings (taste, portion, value) + at least one tag from a curated set of 20; text is optional. Structured data is aggregatable and comparable — free text isn't. Sub-ratings surface why a dish is loved (great taste, stingy portion), which restaurant-level stars can never express. Trade-off: higher submission friction. Requiring a photo on every review is a known risk — flagged as an assumption to validate, with an escape hatch (photo optional + bonus points) queued for V2. (Considered instead: free-text reviews, a single overall star rating.)

◆Decision — Cold start

A DishPoints economy to bootstrap the content flywheel. The product is worthless empty, so contribution is directly incentivized: 10 pts for a basic review, 25 pts for a full one (30+ chars, 5+ distinct words, not a near-duplicate). Seven-day streaks double points; milestone nudges fire at 250 and 450 points; points redeem for real restaurant coupons in a rewards marketplace. The 2.5× gap between basic and full reviews is a deliberate quality lever — it pays users to write better content, not just more. (Considered instead: manual seeding only, waiting for organic UGC.)

◆Decision — Second audience

Turn feedback into a B2B wedge. Verified restaurant owners get an analytics dashboard: 30-day review volume, top/bottom dishes, and per-dish sentiment with tag distribution. Same UGC data, second audience — and a future paid-SaaS path. (Considered instead: a consumer-only product, selling raw data.)

◆Decision — Go-to-market

Ship one city deep, not ten shallow. Launch Gurugram-only with 10 defined areas. City/area is denormalized down to the dish level, so expansion is a config + ingestion step, not a rebuild. Depth of content in one market beats thin coverage everywhere for a network-effect product. (Considered instead: a multi-city launch, pan-India from day one.)

The Core Loop

the core loop
Search a dish  →  Read structured, scored reviews  →  Order with confidence
      ↑                                                        ↓
Redeem coupon  ←  Earn DishPoints  ←  Write a photo review  ←  Loved it

Every review enriches the database, which makes search more useful, which pulls in more users, who write more reviews. Three reinforcing loops sit on top:

Content flywheel — reviews compound the data asset. Retention loop — streaks, badges (Newbie → Foodie → Critic → Legend), wishlist "save-for-later". Value loop — points → coupons closes the effort-to-reward gap.

Monetization

Three pillars, sequenced from consumer to B2B:

  • Premium subscription — ₹199/mo or ₹1,999/yr (dish comparison, advanced stats).
  • Coupon marketplace — partner / commission fees from featured restaurants.
  • Restaurant analytics SaaS — the owner dashboard as a paid product.

Competitive Positioning

Zomato · Swiggy · Google

  • Review unit: Restaurant
  • Sub-ratings: None
  • Photo per review: Optional / rare
  • Dish-level aggregates: No
  • Owner insight: Restaurant-level

Cravia

  • Review unit: Dish
  • Sub-ratings: Taste · Portion · Value
  • Photo per review: Required
  • Dish-level aggregates: Yes
  • Owner insight: Per-dish sentiment

Defensibility is the data asset. A dish-scored menu for every restaurant is expensive to build and compounds with scale. A competitor would have to break its own review model to match it.

Outcomes (Build State)

A fully functional PWA, not a prototype:

  • End-to-end discovery, review, wishlist, and rewards flows.
  • Complete gamification engine (points ledger, streaks, 7 badges, 4 levels) with atomic Firestore transactions.
  • Razorpay premium checkout with signature verification.
  • Restaurant claim → verified analytics dashboard.
  • Admin panel for moderation, coupons, dish requests, and user roles.
  • Production infra: PWA / offline, SEO with JSON-LD, dark mode, rate limiting, Sentry, CI/CD.

Known Risks & What I'd Validate Next

▲Honest risks

Cold start — empty restaurants deliver no value; needs concentrated seeding per city. Review fatigue — photo + 3 ratings + tags is a heavy ask if rewards feel thin. Coupon supply — the value loop dies without desirable partner coupons (a BD problem, not a product one). Thin premium — only one feature (dish comparison) is truly gated today; the upsell needs more meat.

Highest-leverage next bets:

  • WhatsApp / social share — the single biggest growth gap for India; recommendations already live there.
  • Typesense search — Firebase prefix-match misses restaurant names, mid-word, and typos.
  • Restaurant QR codes — turn claimed owners into a distribution channel at the table.
  • Push notifications — service worker already exists; wire streak reminders + wishlist activity.

What This Case Study Demonstrates

  • Framing a market by changing the unit of analysis (restaurant → dish) rather than competing feature-for-feature.
  • Designing an incentive system that solves cold-start and enforces quality with one lever.
  • Building a data moat where UGC serves both a consumer product and a B2B wedge.
  • Intellectual honesty — naming the unvalidated assumptions (photo friction, coupon supply, "does dish granularity matter enough?") instead of hiding them.
★Key Takeaway

The sharpest product lever isn't a better feature — it's a better unit of analysis. Reviewing the dish instead of the restaurant reframes the entire market, and the same reframing that makes the product useful (structured, dish-level data) is also what makes it defensible.

PreviousCaloVoice: Voice-First Calorie Tracking
NextFlipkart Cart API — Design

Built with intention, not templates. If you made it this far, thank you for reading thoughtfully.

© 2026 · Designed and built with care