Internship
A step towards tackling toxicity in online gaming.
I created this Car Recommendation Project as part of my internship with Wongdoody, a well-known design firm. Throughout this experience, I had the chance to work under the supervision of a mentor and contribute to a project that aimed to revolutionize the car-buying experience for car dealerships.
Given Brief
Vehicle recommender for Auto Dealer
Client: ■■ based in ■■
Several Auto Dealers across ■■ are affiliated with our ■■, and they regularly bid and purchase vehicles through our online auction platform. The users buy cars through our platform and resell them through their stores. However, the users have been finding it difficult to find the cars they want from the massive catalogue of cars we have available.
We need to design a recommendation system for the auto dealers to help them find the cars they want much easier. We will use AI & Machine Learning to study the usage patterns to create curated lists of cars to each user.
Our vision:
Landing page with multiple individual carousels. Each carousel will implement its own methodology.
Example carousels:
'Because you bought Ford F150'
'Because you missed out on Honda Civic 2009'
'Because you live in ■■'
'Because you like Tesla Model S'
Smart search where users could input text like 'Cars like Volkswagen Polo', 'Honda sedans below $12,000','Muscle cars manufactured before 1990', etc."
With the help of my mentor, who provided valuable guidance and support, I worked closely throughout the project. Together, we set out on a journey to develop a car recommendation system that would revolutionize the way auto dealerships discover and select vehicles.
In the following sections, we will go into more detail about the project, starting with the initial thoughts and questions that guided our approach. We will then explore the intricacies of understanding dealerships, the process of building the recommendation system, and the subsequent design and testing phases. Let's dive in and discover the exciting details of this project.
Objective
Our goal was to develop a Car Recommendation System that would simplify the car selection process, empower dealers with personalized recommendations, and improve their overall sales performance. The problem we wanted to solve was the
Inefficiency and complexity of searching for and selecting cars from the extensive catalogue available on the online auction platform.
Traditional methods of browsing through thousands of listings were time-consuming and often resulted in suboptimal choices. To overcome these challenges, I set out to create a recommendation system that uses AI and machine learning to provide tailored suggestions to car dealers.
The scope of our solutions was limited to the recommendation system. The auction and financing solutions were already established by the client. Thus, our role was strictly limited to the recommendation system.
When developing the Car Recommendation System, several questions and considerations arose. These initial thoughts helped us establish the foundation for our approach to an effective solution. Some of the "first thoughts" questions were:
- How would the recommendation system work?
- Would the recommendation depend on past buying habits or the current popularity of different cars?
- Would people's buying habits and interests come into suggestions?
- Would suggestions be seasonal / depending on the time of the year?
- Would the location of a dealership impact car needs?
- What existing auction platforms or ways do dealerships use to get cars?
- What impacts dealerships' car buying decisions?
- How much does a car's current popularity come into play?
These initial thoughts helped us lay the foundation for our approach towards an effective solution.
It was essential for me to have a comprehensive understanding of the dealerships involved in the car buying and auction process. By learning about their operations, motivations, and decision-making factors, I could create a customized Car Recommendation System that truly meets their needs. I accomplished this by combining secondary research and interviews with car sellers. Here are the key insights gained during this phase:
How do dealerships work
In the United States, direct manufacturer auto sales are banned in many states by franchise laws that require that new cars be sold only through dealerships. Some manufacturers have opened showrooms, but customers must order their new cars online. A car dealership can be either a franchised dealership, which sells new and used cars, or a used car dealership, which sells only used cars. Car dealers order their inventory based on their understanding of the market, how well certain models have sold in the past, and feedback from consumers.
New cars
Car dealerships can either be franchised to obtain new cars directly from the manufacturer or contract with multiple manufacturers. They also receive assistance from manufacturers in financing the cars with interest, which is referred to as flooring. Auctions are also a good source for dealerships to find cars, as well as new cars that have been returned to the manufacturers.
Used Cars
Dealerships can acquire used cars through trade-ins and auctions. Trade-ins account for up to one-third of the used cars on their lots, while auctions account for another one-third.
We investigated the factors that influence dealerships' car purchasing decisions. This included considering variables such as market demand, customer preferences, pricing, vehicle specifications, and regional trends. By incorporating these factors into the recommendation system, we aimed to provide dealerships with targeted suggestions that meet their specific needs.
Car Popularity
Popularity of a car depends on the trust the public has in a brand as well as media trends in case of new launches. Understanding these shifts in market is how dealership order cars.
Location of a Dealership
Factors such as demographics, economic conditions, and lifestyle preferences influenced the types of cars that would resonate with customers in a specific area. By accounting for these regional variations, we could provide dealerships with recommendations that were tailored to their target market.
Add-ons and Spare Parts
Car dealerships make money from a variety of sources, including car sales, add-ons, and parts for car repairs. Add-ons can be a significant source of profit, and dealerships may make more money from cars that have a general trend of customers adding more add-ons.
Suggestion Systems
The car recommendation system was developed by taking inspiration from popular platforms such as entertainment websites, game platforms, e-commerce platforms, and social media. These platforms use user data to understand their preferences, interests, and behaviors, enabling them to offer personalized recommendations. The system aims to incorporate similar principles while adapting them to the specific needs of car dealerships.
The system takes cues from platforms like Netflix, Amazon, and Flipkart, recognizing the importance of understanding user preferences based on their past searches, purchases, and engagement. By leveraging user data, the system aims to create a recommendation system that can anticipate the needs and desires of dealership professionals.
Similar to Netflix's approach, the system understands that dealing with human tastes and preferences poses a significant challenge. Members may visit the platform without a clear idea of what they're looking for, and it's the system's job to provide them with relevant suggestions quickly and effectively.
Within each row of recommendations, the system prioritizes the strongest suggestions on the left, ensuring that the most relevant options are easily visible and accessible. Additionally, the system prioritizes the strongest recommendations at the top of each row, allowing users to quickly identify their preferred choices.
High vs Low Investment
When recommending high-investment products like cars, it is important to understand the user's perspective and address their concerns. Unlike low-investment products, where users are more likely to give a service a second chance, a poor experience in the car-buying process could have significant consequences for the user's finances and overall satisfaction. Therefore, the goal is to minimize the possibility of poor decisions by providing users with the necessary information and easing their decision-making process.
While the aim is to assist users in making informed buying decisions, it is important to find a balance between providing comprehensive information and avoiding overwhelming the user with excessive clutter upfront. Too much information at once can lead to decision paralysis. Therefore, the approach involves delivering the benefits of the recommendation system without overwhelming the user.
Similar to how Netflix displays a thumbnail and a short trailer on hover, users are provided with concise summaries and key features of recommended cars, giving them a quick glimpse of each option.
However, a high-investment product like a car requires more than just a glimpse. To strike a balance between providing comprehensive details and avoiding clutter, interactive elements that allow users to access additional information when desired are useful.
How can we give all the essential information without overwhelming the user?
Car Score
CarScore is a comprehensive rating system that goes beyond traditional factors to evaluate the suitability of a car for a particular dealership. The algorithm considers key elements such as price, features, specifications, popularity, color, commission rate, add-ons, and upgrades. It also takes into account the dealership's location and its surrounding factors, such as the local economy, nearby landmarks, and local buying trends. By considering these regional influences, the algorithm ensures that the recommended cars align with the preferences and needs of the dealership's target market. The integration of location-based data allows CarScore to provide a more tailored and accurate recommendation for each dealership. It enables the system to account for the unique characteristics and demands of different geographical areas.
For example, a dealership located in a suburban area with a high demand for budget-friendly cars may receive higher CarScores for affordable options. On the other hand, a dealership situated in a more affluent neighborhood might receive higher scores for luxurious and high-end vehicles.
Car Score provides essential information at a glance, so users don't get overwhelmed. Instead, they get a good idea of what to expect. If they want to learn more, they can look at how the score was calculated.
During our ideation sessions, everyone had the opportunity to share their thoughts, experiences, and innovative concepts. We encouraged an open and inclusive environment where no idea was dismissed without consideration. This allowed for a diverse range of perspectives and solutions to emerge.
Once we had a comprehensive list of ideas, we embarked on a process of evaluation and selection. Each team member had the opportunity to review and rank the ideas based on their potential impact, feasibility, and alignment with our project goals. Through a voting system, we identified the top four ideas that garnered the most support and had the greatest potential for addressing toxic behavior effectively.
During our ideation sessions, everyone had the opportunity to share their thoughts, experiences, and innovative concepts. We encouraged an open and inclusive environment where no idea was dismissed without consideration. This allowed for a diverse range of perspectives and solutions to emerge.
Once we had a comprehensive list of ideas, we embarked on a process of evaluation and selection. Each team member had the opportunity to review and rank the ideas based on their potential impact, feasibility, and alignment with our project goals. Through a voting system, we identified the top four ideas that garnered the most support and had the greatest potential for addressing toxic behavior effectively.
After a thorough discussion with my mentor, we prioritized usability and incorporated proven ideas. Using our learnings from well implemented recommendation systems and combining them with car score to create a user-friendly interface.
More CaseStudies
Gaming Addiction
4/24/23
Carbon - Tackling Toxicity
4/8/23