Inspiration
What it does
Prior to pickup, customers are given incentivized upgrade rental options based on their distance from nearby Sixt stations.
Example: A customer booked a red, medium-sized, automatic transmission SUV for 30€ per day at a Sixt station 22km away from where they are currently located. Prior to pickup, the customer will be given a sorted list of cars very similar to what they requested located at a closer Sixt station for 32-36€ per day.
How we built it
Our solution's implementation relies on enforcing constraints on a list of available cars, such that it ranks potential upsell options from the available cars based on:
- How similar it is to the booked vehicle based on a Cosine Similarity
- How close the user is to the sixth stations based on the Haversine Distance between them.
- Additional weighting based on a distribution to prioritize a certain profit margin
Since the solution is purely algorithmic, it can be executed on demand with no latency.
Challenges we ran into
Our biggest challenge was preparing mock data that closely mimics the sample vehicle catalogue that SIxt provided (including the vehicle properties and attributes). In the end, we were able to create a database generator which creates mock vehicles based on attribute permutations.
Accomplishments that we're proud of
Throughout the entire development process, we put ourselves in the shoes of both Sixt and the customer.
From the Sixt side, we prioritize the ease of development and cost savings. Since our solution is built between the booking and the pickup process using only existing vehicle information, it should be able to be integrated into the current pipeline without any changes to the prior rental process. Additionally, since the solution is purely algorithmic (and does not rely on LLMs), it does not require significant computing power that would lead to high expenditure costs.
What we learned
The challenge that the Sixt team has presented may be a straightforward problem to solve, but the solution planning can lead to various complexities, as there are many ways we can think of to tackle the problem. We decided to think of a simple solution that can enhance the services Sixt has already offered, because we believe that the technology Sixt has provided is already great as itself. In the end, we are really proud of our solution, providing not only potential benefit to the business in terms of customer experience, but also not disrupting the existing pipeline that has already been in place for Sixt services to run in.
What's next for Lapis Legit Sixt Project
Though our solution is not of what something that is very complex, it can be extended in many ways. Firstly, using a real-world database from Sixt would be beneficial into understanding more of how the algorithm runs on real data. Secondly, this solution can be improved or extended in many ways, such as cross-platform improvement, further UI/UX expansion, as well as possibly integrating our solution to the Sixt pipeline.
Built With
- flask
- next.js
- python
- react
- typescript
Log in or sign up for Devpost to join the conversation.