Hackathons at Wayfair
Hackathons usually give people an opportunity to play with some unique ideas, advanced technologies, and even start diving into completely new areas which are small beginnings of big new initiatives. For example, Wayfair is one of the industry leaders in 3D commerce today thanks to what started as a hackathon project a few years ago. Another type of hackathon projects are ones that are solving problems that are never making it to the drawing board, but hackathons bring them quickly to near-production and get support for finalizing them. They are also not unknown at Wayfair and we also followed the path of delivering tangible results working closely with our stakeholders, all the way experimenting with new technologies.
Our goal – simplify a process that is currently tedious and manual.
The title of our project was “Lightning Disputes” and aimed to speed up the process of handling chargeback disputes. After an order is finalized, the customer may request their money back for a variety of reasons. The request where the customer disputes the charge on their account is called a chargeback. This process is time-consuming, so we realized that, as a tech company, we could improve it by automating certain steps on behalf of operations agents. Such automation would save time, reduce errors, and make the job less mundane, all of which allow agents to process more requests at a higher quality.
Two hackathon days
This hackathon, like many other things in 2020, was unique because it was all remote. No lunches in front of the same whiteboard or coffee runs for the entire team. However, our team culture has always been a lot of team discussions, a lot of pairing, and sharing knowledge and effort. After initial group brainstorming, we broke the problem into a few pieces and split into pairs. During the two days, we were changing the pairs and tasks as necessary, all the time basically having an active video call. Also, one of the key factors of our success was the constant feedback loop with our end users (operation agents) throughout those two days.
Experimenting with new technologies
The tech stack of my team consists mainly of Java on Spring Boot and we are using Wayfair’s on-prem data centers to deploy on internal k8s clusters. For this hackathon, however, we decided to go fully cloud-native and serverless, so we ended up using different GCP’s managed services, including Cloud Run and BigQuery. Some of them were first for our team and it was another layer of excitement and innovation that kept us glued to our laptops. Luckily, the sandboxes prepared by the hackathon prep team were all up and running, so we could just focus on learning, and skip the tedious wiring setup.
Outcome – before vs. after
Quick reminder of before: agents jumping through a variety of tools, portals, and data sources to get all the data points and put them all in a single document; a process that was taking a lot of time to collect the data, to verify (and re-verify) and was producing quite some stress.
The picture after: agents entering an order ID and getting a near-ready document in real-time. Quick and smooth.
The hackathon version was already connected to production data thanks to us using Google credentials as a method of verification. Hence, our product was immediately able to handle production data. Thankfully, the hackathon judges recognized the value and innovation direction of this project and decided that we’re co-winners getting this amazingly hip hippo as a trophy!
Next steps
Luckily, we managed to get this project near to full production usage during the hackathon, and got some great feedback from our customer agents such as:
- "Team, we can't express how excited we are to start using this tool you are building!"
- "Overall, this tool represents yet another large step in our modernization of disputes handling at Wayfair."
So it was just natural to quickly follow-up on tightening up some straightforward issues that we intentionally skipped during the hackathon (like data access management and performance optimization). Disputes are a known challenge for the entire fraud & payments industry, so I am sure there will be plenty of iterations on this product and in this space.
At the end, huge thanks to the team that made this happen: Liliana Ziolek, Jonida Ndreu, Issam Ouardi, Nebojša Jogrić, and special shout out to our stakeholders: Niko Mikac, Brad Weil, Dylan Ruiz and Matt Port.