Scaling party orders through Gen-AI curated menus and smart restaurant recommendations
Large order planner
COMPANY
Zomato
DEVICES
RESPONSIVE ACROSS MOBILE
ROLE
LEAD PRODUCT DESIGNER
4 MONTHS
TIMELINE
Overview
Zomato is India’s go-to platform for everything food—whether it's ordering in, dining out, or discovering new favourites.
Large Order was our attempt to bring the chaos of party planning into a calm, curated experience. People usually called restaurants directly for big gatherings—trusting humans over apps for price, quantity, and delivery reliability. We set out to change that. Using Gen-AI, we built a planner that crafted smart, customizable menus based on group size, budget, and preferences.
The result? A record-breaking ₹80 crore in Large Order GMV (Gross Merchandise Value) in October alone—a 54% jump over September and 150% higher than the previous year. In key regions like Delhi NCR, orders tripled. The journey wasn't without hiccups, but it was a masterclass in designing for scale, navigating constraints, and building with empathy.
Context
Despite Zomato’s stronghold in food delivery, we noticed a missing piece: large group orders—birthdays, office lunches, festive gatherings—were still happening offline. Customers preferred calling restaurants directly, trusting humans over interfaces for price negotiations, customization, and delivery assurance. This offline behavior revealed a quiet opportunity: to bring the warmth and flexibility of these high-stakes orders into a digital format. Much like framing a mural with structure and balance, our aim was to design a flow that felt as dependable as a phone call—powered by Gen-AI, rooted in user needs, and capable of handling the chaos of feeding a crowd.
The opportunity
Many customers preferred placing large orders directly with restaurants. Why? It felt more human, more reliable. They could negotiate, ensure timely delivery, and get the reassurance of speaking to someone. We saw an opportunity to recreate that same sense of confidence—but through design and smart tech.
.png)
Bring offline trust online
Customers believed large orders were cheaper and more flexible offline. Online, they struggled to estimate quantity, compare options, or find good bulk deals. This wasn’t about discounts—it was about transparency. We had a chance to help users feel in control of both budget and portions.
.png)
Solve for value and clarity
Big orders come with big decisions. Most users didn’t want to scroll endlessly—they wanted someone (or something) to take charge. We saw a space for Gen-AI to act like a maître d’, taking in guest details and returning thoughtful, ready-to-go menu suggestions that felt “just right.”
.png)
Curate, don’t overwhelm
So, how might we reimagine the role of a digital concierge to curate, coordinate, and care for every detail of a large order?
.png)
.png)
.png)
.png)
.png)
So, this is what we're gonna do then
And so it begins.....
-
Before jumping into design, we built a functional demo using Streamlit — a quick, scrappy way to simulate the AI logic.
-
The demo let users input party details (like guest count, preferences, budget) and instantly see auto-generated, restaurant-style menus.
-
This helped us test the core value of Gen-AI curated carts — would the suggestions make sense? Feel complete? Match budgets?
-
It gave us early validation without needing to design or build complex UI flows.
-
More importantly, it helped align the team (design, product, engineering) around what the system should feel like, before we started designing.
-
Think of it as our “sketchbook phase” — where we tested the brush before painting the canvas.
a demo on streamlit showing test inputs and the output by AI
The complexity of curation
Curating party meals wasn’t just about picking a few popular dishes. The sheer range of possible combinations made the decision-making overwhelming for users:
-
Users often wanted a complete experience—not just a dish, but a thoughtful ensemble of mains, sides, desserts, and drinks.
-
Cuisine preferences varied drastically—Indian, Asian, Continental, American, even regional favourites like chaat or biryani.
-
Party moods and occasions were different: casual office lunches, birthday dinners, team parties, kitty parties, etc.
-
If we showed all possible permutations, the UI would’ve felt like a chaotic buffet rather than a curated platter.
How we solved it
We grounded our solution in real data. By studying large orders (₹10K+) placed between 12–2 PM in Gurgaon,
we found a pattern:
​
Functional meals like Burgers, Thalis/Bowls, Pizza, and Wraps made up 50%+ of these orders​
​
This insight helped us filter down the clutter and focus on what works, keeping the user experience simple and goal-oriented
Explorations of menu items
Let the games begin
Putting things to action
Once the key features were locked in, we mapped out a sample flow to bring clarity to the user journey and ensure all moving parts worked together seamlessly. Wireframing at this stage helped us visualize the planner end-to-end, identify edge cases early, and align the ensemble on how the experience would actually come to life.
Entry point
The scheduling feature here should be extended to up to 7 days
Cool illustration (animated) showing the menu curation process - showing users preferences being turned into a menu
When a user selects a restaurant
-> Cuisine is pre-selected
-> User can still choose more cuisines but cannot remove the preferred restaurant’s cuisine
If user skips menu from same restaurant twice - we show menu from some other restaurant (This is only in the case of user choosing a res preference)
When the Curtain Rose... and Fell
After two months of designing, refining, and developing what we thought was a solid solution, we rolled out the feature to a small test group. But during testing, cracks began to show. The Gen-AI-powered search was too heavy—latency shot up and each query was costing way more than expected. Despite all the approvals and excitement, we had to make the tough call to roll the feature back. It was a hard pause, but a necessary one—time to go back to the drawing board and rethink how to make this work at scale.
THE GRAND RE-ENTRY
With Diwali around the corner, we reimagined the large order experience to capture festive group dining—smarter, faster, and actually shippable.
Diwali Reloaded
Feature Party SKUs
Spotlighted special party-friendly dishes from both regular and scheduling-only restaurant partners
Scheduling Support
Enabled advanced order scheduling for better coordination during busy festive days.
Large Order Deals
Added a filter to highlight restaurants offering exclusive deals and discounts on bulk orders
Curated menus
Revamped the Gen-AI planner to suggest cached carts, ensuring lower latency and faster load times
.png)
Impact
October 2024 was our brightest festival yet — we hit a record ₹80 crore in Large Order GMV, with an extra ₹10 crore just in the few days till Bhai Dooj!
During the 23-day Diwali window (Dussehra to Bhai Dooj):
-
India’s LO GMV grew 49% YoY (₹70cr in ’24 vs ₹47cr in ’23)
-
Delhi NCR alone grew 33% YoY (₹28cr in ’24 vs ₹18cr in ’23)
Contribution to the Bigger Picture
-
India LO GMV contribution rose to 2.6% (vs 2.1% YoY)
-
Delhi NCR reached 5.4% (up from 4.3% YoY)
-
On Oct 30: ₹8.5 crore GMV, ₹4100 AOV, 7.45% platform contribution
The Large Order Planner helped Zomato tap into the festive party market, driving a massive surge in high-value group orders. It turned into a key growth lever during Diwali, proving the potential of well-curated, scheduled large orders.
My Learnings
What building a Large Order Planner taught me about product, people, and letting go
Detach from your masterpiece
I learned not to take design changes personally. When the product faced setbacks, stepping back helped me see the larger picture and iterate better.
Paint the picture, don’t just describe it
I realized that when there are a lot of stakeholders involved, decision making becomes harder—quick wireframes and flows in meetings made alignment faster and clearer.
Effort ≠ Outcome
Even the most thoughtful products can fall short. This taught me to embrace experimentation and see value in learnings, not just launches.
Design for what's real, not what's ideal
This was a big project spanning across multiple teams including Product, Data, Research, Business and of course, design. It made me more confident in presenting my thoughts and ideas out loud in meetings and giving design a seat at the table.