How would you improve the recommendations for Zomato?
Product Improvement Interview Question
Dear readers,
Thank you for being part of our growing community. Here’s what’s new this today,
Product Management Interview Question:
Q: How would you improve the recommendations for Zomato?
Note: This post is for our Paid Subscribers, If you haven’t subscribed yet,
Before jumping into solutions, I want to make sure I understand the scope and constraints of this question clearly.
Step 1: Ask Clarifying Questions
Q: When we say “recommendations,” which surface are we talking about? The homepage feed, the search results page, the reorder widget, the “For You” section, or push notifications?
Let us take a broad view of all consumer-side recommendation surfaces inside the Zomato app, since they share an underlying signal layer even when the UI differs. We will prioritize the homepage feed since that is where first-session intent is set.
Q: What is the primary goal we are optimizing for: order conversion in the current session, long-term retention, or restaurant discovery for newer restaurants on the platform?
Let us optimize for order conversion and long-term retention together. A recommendation that converts today but erodes trust through irrelevance hurts retention. We want recommendations that are immediately useful and build the user’s confidence in the platform over time.
Q: Are we restricting ourselves to the India market, or taking a global view?
India-first. Zomato’s core business is domestic, and the product decisions in this answer should be grounded in Indian user behavior, infrastructure constraints, and competitive dynamics.
Q: Should I consider what Zomato has already shipped in personalization, or are we treating this as a greenfield problem?
Build on what already exists. Proposing a feature that Zomato already ships is disqualifying at the senior level. Map the current state first, then identify the gaps.
Q: Are we thinking about a single-user context, or should we also account for shared accounts such as a couple or family using one Zomato login?
Both matter, especially in India where shared accounts are extremely common. Consider solutions that work for both individual and multi-user contexts.
Q: Should the recommendation improvement focus only on restaurants, or does it also include dish-level and cuisine-level recommendations?
Focus primarily on restaurant-level recommendations on the homepage. Dish-level recommendations within a restaurant page are a separate problem surface and a follow-up scope.
Q: Are there any constraints on the engineering side, such as avoiding changes to the core recommendation model or staying within the mobile app only?
No hard constraints. We can propose both model-level and UI-level changes, and we can think across the mobile app, push notifications, and the underlying ranking infrastructure.
Q: Should I account for the tension between organic recommendations and sponsored restaurant placements that Zomato monetizes through its advertising platform?
Yes. That tension is real and any serious recommendation improvement strategy must address it rather than ignore it.
These clarifying questions are doing distinct work. The surface question narrows scope from a vague system to a specific screen. The shared-account question proactively names a structural challenge specific to India. The sponsored-placement question signals awareness of the business model tension. And the “already shipped” question signals that the candidate does product research before interviews, not during them.
Step 2: Describe the Product
Zomato is one of India’s leading online food delivery and restaurant discovery platforms, founded in 2008. It connects millions of consumers with restaurants for dining, takeout, and delivery across more than 800 Indian cities.
Core Services
Food delivery: On-demand ordering from restaurants with real-time tracking
Restaurant discovery: Search, menus, reviews, ratings, and photos for dining-out decisions
Table reservations: Online booking at partner restaurants
Zomato Gold: A premium membership offering free deliveries, exclusive discounts, and priority access, with over 2 million subscribers by 2025
Blinkit (sister brand): Quick-commerce grocery and essentials delivery
Key Metrics
Revenue of approximately Rs. 9,418 crore in FY25
24.1 million monthly transacting customers as of Q2 FY26, up 16% year on year
Roughly 58% market share in Indian food delivery, compared to Swiggy at approximately 34%
Active in 800+ Indian cities with approximately 247,000 average monthly active restaurant partners
Mission and Competitive Context
Zomato’s mission is “better food for more people.” Its primary competitors are Swiggy in food delivery, Magicpin in restaurant discovery, and global players like DoorDash and Uber Eats (whose India business Zomato acquired). The food-tech market in India is projected to grow from USD 30.8 billion in FY24 to USD 120 billion by FY32, making retention and engagement improvements highly strategic bets.
Step 3: Define the Goal
Product Goal: Improve recommendation relevance for end consumers in the Indian market so that users find the right restaurant faster, order more confidently, and return to the platform more frequently. The measurable outcome is an increase in order conversion rate on the homepage feed and an improvement in 30-day repeat order rate among high-frequency users.
This goal aligns with Zomato’s mission of “better food for more people” and directly supports its two most important business health signals: order frequency (which drives gross order value) and user retention (which reduces customer acquisition cost over time).
It also respects the platform side of the business: better recommendations surface more relevant restaurants to users, which increases visibility and order volume for restaurant partners, creating a healthy supply-side flywheel alongside the consumer-side benefit.
Step 4: User Segmentation
Zomato serves an enormous and diverse user base across 800 Indian cities. The recommendation problem looks very different depending on which user we are designing for.
