Car Recommendation Model:
Addressing the Supply-Demand Gap at Zoomcar
Car Recommendation Model:
Addressing the Supply-Demand Gap at Zoomcar
Zoomcar is a car-sharing platform in India with over 25,000 self-drive cars available in more than 38 cities. It allows users to rent cars by the hour, day, week, or month through a mobile app and website. The platform has over 20,000 hosts offering various car options.
Timeline
June 2024
PLATFORM
Zoomcar Host App, OneApp
Problem Statement
Problem Statement
The platform currently faces a supply-demand mismatch in certain areas. For instance in the Bengaluru city, Indiranagar experiences a daily demand of 8–10 bookings for Maruti Suzuki Swift, while the supply is only 4–6 cars per day. This results in lost revenue for Zoomcar and missed earning opportunities for hosts.
Solving for lack of awareness of adding a quality car on the platform rather than the less performing cars.
The platform currently faces a supply-demand mismatch in certain areas. For instance in the Bengaluru city, Indiranagar experiences a daily demand of 8–10 bookings for Maruti Suzuki Swift, while the supply is only 4–6 cars per day. This results in lost revenue for Zoomcar and missed earning opportunities for hosts.
Solving for lack of awareness of adding a quality car on the platform rather than the less performing cars.
Goal
Goal
The goal of this project is to bridge the demand-supply gap by recommending another vehicle additions to existing high-quality hosts. We aim to create a car recommendation model, in collaboration with partners like Cars24 and Spinny, that curates a list of top cars based on location and budget.
The goal of this project is to bridge the demand-supply gap by recommending another vehicle additions to existing high-quality hosts. We aim to create a car recommendation model, in collaboration with partners like Cars24 and Spinny, that curates a list of top cars based on location and budget.
Research and Data
Research and Data
To understand the willingness of our hosts to add another car to the platform, Zoomcar conducted a Return User (RTU) survey. Here’s a summary of the findings:
Total Respondents: 451 (Single Car Hosts: 353, Multi-car Hosts: 98)
Single car hosts willing to add another car: 131 (37%)
Multi-car hosts willing to add another car: 39 (40%)
To understand the willingness of our hosts to add another car to the platform, Zoomcar conducted a Return User (RTU) survey. Here’s a summary of the findings:
Total Respondents: 451 (Single Car Hosts: 353, Multi-car Hosts: 98)
Single car hosts willing to add another car: 131 (37%)
Multi-car hosts willing to add another car: 39 (40%)
Approximately 37.8% of all respondents showed interest in adding another car to Zoomcar.
Approximately 37.8% of all respondents showed interest in adding another car to Zoomcar.
Why Focus on Existing Hosts?
Why Focus on Existing Hosts?
Trust in the platform is essential for the success of our car recommendations. Once a host is comfortable with the earnings and reliability of the platform, they are more likely to add another car. Additionally, partners like Spinny and Cars24 prefer working with professional hosts who make sound financial decisions when buying cars.
Trust in the platform is essential for the success of our car recommendations. Once a host is comfortable with the earnings and reliability of the platform, they are more likely to add another car. Additionally, partners like Spinny and Cars24 prefer working with professional hosts who make sound financial decisions when buying cars.
UX Audit
UX Audit
A few months back before this problem statement got priority, we did UX audit and identified several pain points. The current model is almost broken, which led us to brainstorm ways to improve it.
A few months back before this problem statement got priority, we did UX audit and identified several pain points. The current model is almost broken, which led us to brainstorm ways to improve it.
Ideation
Ideation
We brainstormed extensively on screens and requirements, drawing from survey insights and audits. Mapping the user journey for hosts adding new cars, we identified key ingresses for different use cases throughout the application:
Homepage: Action cards and marketing banners for top-rated hosts with specified bookings.
‘Your Cars’ Page: Repository of all cars a host has on the platform, with settings and a direct ‘Add a New Car’ flow linked to the recommendation model.
Profile & Earnings Pages: ‘Add a New Car’ option in the profile settings and ‘Earn more!’ marketing on the earnings page.
We brainstormed extensively on screens and requirements, drawing from survey insights and audits. Mapping the user journey for hosts adding new cars, we identified key ingresses for different use cases throughout the application:
Homepage: Action cards and marketing banners for top-rated hosts with specified bookings.
‘Your Cars’ Page: Repository of all cars a host has on the platform, with settings and a direct ‘Add a New Car’ flow linked to the recommendation model.
Profile & Earnings Pages: ‘Add a New Car’ option in the profile settings and ‘Earn more!’ marketing on the earnings page.
Brainstorming sessions
Brainstorming sessions
In our team meetings with Product managers, we identified the right hosts using benchmarks: ratings above 4.5, more than five bookings, and past the D60 milestone.
We plan to recommend cars by:
Internal Data: Using vehicle search conversion rates and search duration mix.
External Data: Comparing national purchase behavior and city-specific Google search trends.
Develop a risk exposure metric based on credit scores and current debt. Additionally, create revenue and earnings calculators projecting net earnings after vehicle EMI deductions.
✅ Recommendations by city regions (e.g., East Bangalore, West Bangalore).
✅ Focus on recommending used cars, not new ones.
✅ Consider hosts’ vehicle preferences, not just credit scores.
✅ Simplify model recommendations to Car Make + Transmission Type + Ignition Type + Year.
In our team meetings with Product managers, we identified the right hosts using benchmarks: ratings above 4.5, more than five bookings, and past the D60 milestone.
We plan to recommend cars by:
Internal Data: Using vehicle search conversion rates and search duration mix.
External Data: Comparing national purchase behavior and city-specific Google search trends.
Develop a risk exposure metric based on credit scores and current debt. Additionally, create revenue and earnings calculators projecting net earnings after vehicle EMI deductions.
✅ Recommendations by city regions (e.g., East Bangalore, West Bangalore).
✅ Focus on recommending used cars, not new ones.
✅ Consider hosts’ vehicle preferences, not just credit scores.
✅ Simplify model recommendations to Car Make + Transmission Type + Ignition Type + Year.
Design iterations on figma
Design iterations on figma
UX/UI Design Variation
UX/UI Design Variation
After a lot of discussions and we came up with the following flow for the car recommendation model!
After a lot of discussions and we came up with the following flow for the car recommendation model!
UX Feedback
UX Feedback
Into further discussions with the stakeholders, managers and team, these points were concluded to include following approaches:
Category-Based Recommendation Approach:
Organise cars into categories like Hatchbacks, Sedans, Compact SUVs, and SUVs, allowing for simplified decision-making and targeted recommendations. However, it may limit customization and overlook specific preferences or regional nuances.E-commerce Style Filterable Recommendation Approach:
Allowing hosts to filter car recommendations based on various criteria such as price and car age. It offers high customisation and adapts to diverse host preferences, but the extensive choices and filters might overwhelm some users.
Into further discussions with the stakeholders, managers and team, these points were concluded to include following approaches:
Category-Based Recommendation Approach:
Organise cars into categories like Hatchbacks, Sedans, Compact SUVs, and SUVs, allowing for simplified decision-making and targeted recommendations. However, it may limit customization and overlook specific preferences or regional nuances.E-commerce Style Filterable Recommendation Approach:
Allowing hosts to filter car recommendations based on various criteria such as price and car age. It offers high customisation and adapts to diverse host preferences, but the extensive choices and filters might overwhelm some users.
UI Feedback
UI Feedback
Earnings Calculator: Simplify the UI and set default sharing days to reduce user input.
High Demand Areas Map: Make the map more visual and highlight the economic benefits of choosing high-demand areas.
Top Performing Clusters: Cluster top-performing cars to make them more appealing.
Car Card: Make monthly earnings more visible on the car card as a motivation.
Earnings Calculator: Simplify the UI and set default sharing days to reduce user input.
High Demand Areas Map: Make the map more visual and highlight the economic benefits of choosing high-demand areas.
Top Performing Clusters: Cluster top-performing cars to make them more appealing.
Car Card: Make monthly earnings more visible on the car card as a motivation.
The car card UI
The car card UI
Final Screens
Final Screens
We have divided the hosts into 2 categories:
FTU: First Time Users, Hosts who have zero car on the platform.
RTU: Return Users, Hosts who have one or multiple cars on platform.
Let’s talk about the journey of a FTU:
We have divided the hosts into 2 categories:
FTU: First Time Users, Hosts who have zero car on the platform.
RTU: Return Users, Hosts who have one or multiple cars on platform.
Let’s talk about the journey of a FTU:
Based on the feedback, we’ve completely redesigned the input flow, making the map more visual and highlighting the economic benefits of high-demand areas.
Based on the feedback, we’ve completely redesigned the input flow, making the map more visual and highlighting the economic benefits of high-demand areas.
The redesigned car recommendation screens feature car cards that emphasised estimated earnings, categories for easy decision making and include high-demand tags to address supply and demand, and provide a glimpse of the onboarding process for the recommended car.
The redesigned car recommendation screens feature car cards that emphasised estimated earnings, categories for easy decision making and include high-demand tags to address supply and demand, and provide a glimpse of the onboarding process for the recommended car.
RTU have the similar flow, however a bit straightforward Landing page and primary CTA to add a new car, we have removed the introductory benefit section about Zoomcar as they already have trust on the platform as a return user.
RTU have the similar flow, however a bit straightforward Landing page and primary CTA to add a new car, we have removed the introductory benefit section about Zoomcar as they already have trust on the platform as a return user.
And folks, that’s how we designed the recommendation model to address the supply demand gap.
And folks, that’s how we designed the recommendation model to address the supply demand gap.
Conclusion
Conclusion
As a Product Design Intern at Zoomcar, This was my first primary project along with my colleague Mayur Ranka (Product Designer, LinkedIn) and Sukrit Mukherjee (Product Manager).
The project had a really interesting discussions within the design team, stakeholders, developers and more to design the right approach for our hosts on the platform in UX construct meetings.
I learned to balance between users needs as well as business requirements, and it won’t be possible with great minds and designers around me.
We made sure that we connect with our partners such as Cars24, and hear their perspective too.
One of the biggest learning through this project was to keep getting feedbacks at the early stages of design, it does takes multiple iterations and opens up different ways to solve a problem.
As a Product Design Intern at Zoomcar, This was my first primary project along with my colleague Mayur Ranka (Product Designer, LinkedIn) and Sukrit Mukherjee (Product Manager).
The project had a really interesting discussions within the design team, stakeholders, developers and more to design the right approach for our hosts on the platform in UX construct meetings.
I learned to balance between users needs as well as business requirements, and it won’t be possible with great minds and designers around me.
We made sure that we connect with our partners such as Cars24, and hear their perspective too.
One of the biggest learning through this project was to keep getting feedbacks at the early stages of design, it does takes multiple iterations and opens up different ways to solve a problem.
Thank You.
Found this project interesting and want to discuss more?
Connect with me at
Connect with me at