In October 2022, the Data Analytics & Data Science team at diconium data Berlin hosted an exciting hackathon to strengthen organizational collaboration, creativity, technical knowledge, and, most importantly: to have fun! Here we would like to share our experiences in the organization and execution of the hackathon.
A hackathon offers the opportunity to develop data business ideas in a short but intense time window and to become creative in teams away from the daily work routine. It fosters teamwork by promoting cross-functional and cross-locational collaboration under high positive stress. The high degree of required collaboration leads to fruitful exchanges and knowledge sharing. It is a great chance to explore new technologies and boosts the morale of everyone involved. Also, hackathons are a great source of new business cases: remember that Facebook’s iconic Like button got its footing at a hackathon.
Nevertheless, planning a hackathon requires significant resources in terms of time, monetary costs, and other resources to produce while having no predictable return on investment. Before you move forward with the idea of an in-house hackathon, make sure to be supported by the company’s management. Fortunately, this is not a problem at diconium; we felt encouraged and supported to go all the way.
The challenge: Anomaly detection in time series data
Our task for the participating teams was to create a toolbox for detecting and understanding anomalies in time series data. The groups were provided with a dataset containing 10,000 observations and were tasked with developing a real-time monitoring visualization that shows anomaly detection using a predictive model. For a final presentation, we expected 1) well-documented code, preferably as Python Notebook or R Script, which enabled us to verify the solution, 2) a fully functional monitoring visualization that flags anomalies in the online dataset and 3) they should provide an overview of the business case for the model.
To develop this idea, we had previously conducted a workshop together with product owners from our growth team. Different cases were brainstormed, discussed, and presented by small groups. In the end, a decision was made by voting. This ideation workshop already felt like a small hackathon; it was fun collaborating with colleagues outside the daily business and thinking about use cases to develop innovative data products.
Prepare and provide a dataset
To make the hackathon challenge concrete, we had to find a suitable dataset that met the task’s requirements, was easy to evaluate and had no public solutions for the problem to avoid cheating. We also discussed the balance of skill sets needed and the amount of work in the given time. In the end, we went well with a fake dataset connected to a real-world problem concerning electric resistance welding (ERW).
Ten other things to keep in mind when organizing a hackathon
Even after getting support from management, defining the challenge, finding a dataset to work with, and finding a date to execute the hackathon, there are a lot of things to be organized:
- Invite people to join – we also included a brief questionnaire to set up diverse cross-functional and cross-locational teams. Send reminders, because with the everyday work, people might forget about the deadlines.
- Create a timetable for the hackathon day, including introductions, working sessions, a lunch break, a final presentation, a jury session, and a winner’s announcement. And, of course, the possibility of having a cold drink together after this intense day.
- Make the final presentation open to the whole company, so those who could not participate are also involved in the event.
- Provide a code of conduct to clarify legal topics and remind everyone to be kind, professional, respectful, and to have fun.
- Mentoring: we had two mentors to support the teams on the day of the hackathon in terms of content and time management.
- Set up a jury to join the final presentation and choose the winning team. We found three director-level judges from different locations with different backgrounds and leading different core teams, which enriches the discussion to select the winning team.
- The organizer needs to define and communicate the evaluation criteria to the participants and the judges.
- Prices: Although a hackathon is about collaboration and having fun, little giveaways are nice to encourage a bit of competition. diconium provided a fancy water bottle for every attendee and a voucher for every member of the winning team.
- Remember to block rooms for the groups and presentations and enable hybrid work possibilities for attendees who can’t participate in person.
- Take care of food, snacks, and catering: a good mood sometimes needs a little sugar!
How long does it take to organize a hackathon?
We don’t want to scare anyone off, but to be honest: hackathons take a long time to plan. All in all, from the first idea to the execution of the hackathon, it took us almost eight months – with most of the work being done in the three months following the hackathon. Please consider that this has been our first time, and everything has been organized in addition to everyday work.
How it went – the results of the hackathon
Fifteen participants came together from four locations and were divided into four teams. It was great to see how everybody came up with different approaches focusing on the technical part, models, dashboards, or business views in their presentations. It was interesting to observe the various forms of collaboration, and it was impressive to see what each team produced in such a short time.
We are excited to announce the winner: Congratulations to “Pimp my Pipe”, a solution by Ceren Gündogdu (Consultant Data Analytics – Berlin), Hoai-Nam Nguyen (Data Scientist – Hamburg), Pedro Silva (Machine Learning Engineer – Stuttgart) and Sebastian Humberg (Machine Learning Engineer – Hamburg), who provided “the most well-rounded overall package”, according to the committee. In the end, the unanimous question was: “When is the next hackathon?”
Perhaps next time, one could consider extending the hackathon to a second day because the time to work in groups was very tight. In the beginning, more general information about how and why we do it could have helped because only some were familiar with the concept. And another minor learning is to mention prices straight with the invitation to gain even more attendees.
To summarize: the goal of having fun with developing data products in cross-functional and cross-locational collaboration away from the daily work routine was met 100%. We are happy to have spent such a productive and creative day together.
So finally, and representative for everyone involved: Thank you, Maria Victoria Coppola Harus (Intermediate Consultant Data Analytics – Berlin) for being our Chief of Hackathon!