(Won) Ride Austin Data Hackathon

Providing Insights on new riders and creating KPIs for marketing and sales teams to use.

Back in May 2017 I participated in my first hackathon. It was put on by Ride Austin a local ride share company and Galvanize a learning community for technology. Needless to say, I was very hesitant about participating because I doubted I would be able to add value. I doubted if my abilities were going to be on par with the super experienced data scientists that were obviously going to be there. But I embraced the challenge and went. At the very least I would learn, I will get a chance to work on a different type of problem, and interact with other data scientists. 

Competition Rules

There were four teams and each team had a particular set of questions to answer in 8 hours. All the questions were descriptive and didn’t involve any inferential statistics or predictive modeling. I chose the team that answered questions around new riders. Specifically what is the average percent of new riders and what is the weekly percentage of new riders.

The Data

The data was hosted on data.world. Data.world is a web tool where people can securely share data, findings and visualizations. The dataset consisted of two CSV files when combined contained over 1.6 million rows and 60 columns of data. For my analysis I leveraged the rider ID, created on and the total fare columns. Links to the datasets are here and here.

Crunch Time

When the competition started, I sat back and listed to the different approaches. To my surprise, I was able to take charge, come up with a plan, execute and even teach others. I developed the methodology to answer the specific questions Ride Austin wanted. I then worked on answering those questions in Python and pandas. I presented the findings at the end and my team won the hackathon! We won because provided the best business value. We answered the questions and even added some additional business insights based on frequented drop off/pick up locations. 

I would highly recommend people participate in hackathons. It’s an awesome experience to work side by side with strangers to solve problems.

If you are interested in the code I developed you can find it here.

Findings

We found that on average ~ 25% of the rides were done by new riders. The way I identified new riders in this dataset is looking at the first time a rider_id was used. A new rider is defined as someone using the platform for the first time.

This graph shows the raw number of new riders vs total rides per week:

This graph shows by week the percentage of total rides done were done by new riders:

Further Insights

Another thing the dataset provided was the ability to do an analysis on revenue. Revenue is often used as an indicator, but it can hide things. For example, revenue can go up quarter over quarter, but what if new customer revenue is dropping? Since we are looking at new riders, we should break that revenue down to just look at new rider revenue. Looking at these KPIs means that marketing and sales teams can use them to evaluate whether certain strategies are effective at bringing in new customers and how their behavior changes.  We can also use this data to determine the times of year when new customers are the most valuable. (Since Ride Austin is in Austin, the time around South by Southwest is very valuable).

This graphic shows how revenue can be broken down to different KPIs:

The mean number of transactions for a new rider during their first week of service is ~ 1.9 and the mean value per transaction is $15.99. 

This graph shows the average number of transactions for new riders by week:

This graph shows the average fare for a new rider during their first week of service:

Finally this graph shows the average value of new customers by week:

Final Thoughts

If I succumbed to my fears I would have never participated in an awesome event. Hopefully this post encourages someone to take a plunge and showcase your skills. At the very least you will learn something, but you can surprise yourself if you just focus on ways to provide value.