Agility and an understanding of the business environment has helped a team from the UK to win the first Xperts Analytics Hackathon, held in London. The Hackathon brought together teams of insurance data analysts from across Europe to compete in a real-world insurance simulation contest. Throughout the game, the winners kept a well-adjusted portfolio that balanced the number of customers and the level of profitability at all times. “We reacted fast to the changing market conditions as the new dataset was introduced every 15 minutes,” said a member of the winning team. “We adjusted the commission level and would take a chunk out of our nearest competitor.”
Eighteen teams, comprising 70 pricing strategists, competed in the peer-to-peer learning event gaining hands-on experience of advanced technologies in a highly dynamic and competitive environment.
Using the cloud-based V8.5 of the Earnix Analytics Platform, the teams had to work out the best pricing strategy and react quickly to other competitors in the market using Earnix pricing execution module. The online capabilities as well as having one tool for all pricing activities helped the teams to understand the environment, see what their competitors are doing, enhance their models and deploy new strategies into the market in a matter of minutes.
Teams began with a ‘warm up’ first round and once the competition began the number of quotes released into the market were increased in subsequent rounds. The dataset with cumulative results was updated every 15 minutes throughout the competition. More information about the contest was released with each round. The teams realised they had to react to each new piece of information, gaining an understanding of the winning quote in each round and what defined it.
The first bonus challenge was a reverse engineering exercise that could be solved both by traditional GLM methods or using machine learning. Teams were required to estimate the minimum price that the aggregator was willing to offer. Using any statistical method, contestants trained and tested a model on a data with a minimum price for each quote. The winning team used the gradient boosting method (GBM) to achieve the lowest sum of absolute errors evaluated on the validation dataset.
The second bonus challenge was Natural Language Processing, an analysis based on text mining, requiring teams to solve a credit scoring problem. The teams could use any software or technique. The winners used h2o in R, achieving the highest AUC. The goal of this bonus challenge was to expose the users to cutting edge technologies and give them a glimpse into the work the Earnix research department is doing.
As the day progressed, the competition intensified, with teams focused on improving their standings, which were monitored in real time in a league table. Not only did the table show the performance in terms of customer numbers and profit margins, but teams could also assess the velocity at which they published quotes compared with their peers.
Five key pillars combined to produce the winning strategy in the Hackathon:
- Agility in getting quotes to market quickly and removing dependencies;
- Understanding of the business environment and what competitors were doing;
- Being proactive and realizing that a business strategy might have to change depending on market conditions;
- Being innovative and using the latest techniques and technologies such as machine learning and natural language processing; and
- A simplified and faster process – the use of a single platform enables strategists to manage all aspects of their pricing in a straightforward and flexible way. Pricing and its impact on customer behavior can be monitored in real time and the pricing execution model can be adapted rapidly.
Most of the teams understood that they needed to deploy different strategies during the Hackathon to improve their results; everyone changed their pricing more than 15 times during the contest. While at the beginning of the Hackathon teams guessed the best quote, by the end of the competition when full information of the market was revealed, the teams converged on the same pricing level.
“The Hackathon gave us a glimpse of the future of the insurance industry. There are definitely things we have learned today that we can take back to our office,” said one competitor. “The Hackathon has given us all a bit more confidence to use the software to its full extent. During the day we learned what we were doing wrong and could remedy it immediately. We could figure out what other contestants were doing and try to copy and build on that to climb the league table.”
To learn more about best practices in analytics and how your analysts can develop winning strategies, download the Best Practices
in Pricing Report.