Making millions through fantasy sports may sound like a pipe dream, but one Indian company is working to make this dream a reality.
Dream11 is a gaming app that revolves around building fantasy sports teams and battling it out with friends and strangers in a virtual playing field. The app hosts an array of sports – cricket, football, hockey and more – and has caught the attention of millions of users all over the country.
HackerRank’s Senior Director of Marketing, Aadil Bandukwala, recently sat down with Dream11’s CTO Amit Sharma to talk about the dream teams, tools and technologies behind the popular app – watch the full conversation below, or read on for takeaways.
- Performance is the DNA of Dream11’s engineering teams
- Personalization and enhancing social gaming are key areas of focus for the company in 2022.
- The tech team is divided into four pods, or “dream teams.”
- Dream11 is big on homegrown solutions.
The theme of the year is personalization
Personalisation has been a key driver of success in businesses for many years now – Coca Cola’s “Share a Coke” campaign, Snapchat’s Bitmoji and Grammarly’s weekly personalized reports have all been incredibly successful initiatives.
At Dream11, Amit said they’re focusing on this in 2022. “We’ve dabbled with personalization. For example, a couple of years ago, we introduced a “For You” page on our app where users are recommended contests based on previous ones they participated in. It’s enjoyed a lot of success so far; 85% of user interaction with contests actually begins in this section.”
“This year, we want to optimize for personalization everywhere in the app – from the homepage to the payments screen.”
“Another key area of focus for us this year is integrating our social and gameplay aspects. The app contains groups, connected networks, and chat options but we’re going beyond providing these basic social features. We’re going to provide users abilities like joining contests from a group and reading friends’ scores to make the app more engaging.”
Social gaming saw a meteoric rise during the pandemic, and is predicted to make a long lasting impact on the gaming industry. At the fringes of social gaming lie metaverses, blockchain, NFTs and the play-to-earn business model which are all piquing the curiosity of business leaders around the world today.
The “Dream Teams” – and Their Tools
Big goals necessitate a competent and organized team equipped with the right tools to achieve them.
“Our entire team is divided into four pods, or dream teams,” said Sharma. “Each team is responsible for a certain area – there is one for gameplay, there’s one for social, there’s one for gamification.”
What are the tools and technologies that these dream teams are equipped with to create? “We do some standardization when it comes to our programming languages and frameworks – more than 90% of our services are written on vert.x.” said Sharma, “Because we’re a lean team of 350 engineers, standardization helps. Some of the “behind the scenes” services I talked about earlier use Cassandra, some use Ignite, some use MySQL and Aurora, and some use Aerospike, and elasticache is the caching layer for it.”
He talked about the journey they made from a monolithic architecture (where all software is unified and sits in one place) to a microservice-based one – “A few years ago, we realized that we need to develop deep expertise in distributed systems and create a complete micro services architecture instead of the monolithic one we had in order to serve millions of users concurrently.”
Amit expanded on the services, “When you open the app, the homepage that loads is powered by a service, we have a “contest” service that lists all the contests associated with the user’s account, we have a “leaderboard” service and behind the scenes, we have automated services like authentication that complete the user flow. Each of these services are individually and carefully designed.”
Alongside attention to detail and creative flexibility through industry-first technologies, Sharma said that performance is the DNA of the engineering team. The performance testing framework they use is “Torque”, a homegrown solution made up of Jenkins, Scala, Spark and more.
In-house vs. Outsource
Being made in-house, Torque is not an outlier in Dream11’s tech stack – in fact, they have many homegrown solutions that they’ve built from the ground-up.
Sharma recalled when they decided to focus on building solutions in-house rather than look for suitable third party solutions. “Every startup goes through a journey. In the beginning, you don’t have the in-house expertise to build solutions yourself, so you just pick anything off the shelf that you can afford and fit it into your stack,” said Sharma.”It becomes different when you start to scale. Our scale is one of the highest in the country right now. When we leveraged third-party vendors and solutions, we observed many bottlenecks.”
It’s a common problem that isn’t talked about enough – Third party solutions that serve hundreds and hundreds of customers tend to have a common set of features that they offer everybody or groups of customers. “If there’s a use-case or need that’s very specific to you, it’s highly unlikely that they’ll incorporate it in their product or service unless you’re a huge and influential brand,” said Sharma.
Developing solutions in-house comes with its own set of challenges, too. For example, this approach begets higher costs including training if your employees don’t have the requisite skills. For the last couple of years, outsourcing has enjoyed the spotlight as freelance workers grow in numbers and more businesses embrace the gig economy. According to Upwork, 50 percent of freelancers provide highly skilled services such as computer programming, IT, and business consulting.
DataAware is Dream11’s data solution built from scratch; it’s a funnel analytics tool they use to analyze and understand user behavior. “In today’s world, a data lake with a simple SQL interface is not enough,” said Sharma. “We’re also a data-driven company; one of our company values is data obsession so it makes sense that we built our own data platform.”
They’ve also built a customized machine learning model. “The data science team is able to accurately predict the traffic expected using our homegrown ML model,” said Sharma.