We built a model for an online trading platform to
accurately forecast how customers would react
on any given day’s trading the next day.
An online trading platform that allows customers to place bets on financial instruments.
In the world of online trading, market fluctuations and economic news events have the potential to drive variations not only in price but also in customer engagement. Not all days and not all news events are the same. While some news stories can cause a big influx of business, quarterly announcements from the likes of the European Central Bank make little difference to the flow of new business. On days of large price movements, there are greater levels of activity. The company wanted to understand when and how customers respond to market events, to enable it to deliver more targeted marketing.
Reviewing data from the last four years, Faculty observed that large spikes in new customer leads coincided with highly significant news events (such as the Brexit vote, the Trump victory and the flash crash in late 2016). By looking at historical data and focusing on the actual impact of market volatility, we created a model linking the price movements for a particular day with corresponding business behaviour. The model explained well what was happening on any given day.
We were able to improve this model further by making it dynamic, such that it updated itself over time. As the model constantly updated with new data and discarded older, redundant information, it was possible to predict with far greater accuracy how customers would respond to market events.
While the events themselves remain impossible to predict, the model was able to accurately forecast the level of response that any given day’s trading would generate the next day. This is a good tool for many different business aspects. Marketing is one example; in the era of digital marketing these insights could be used to create more relevant and timely campaigns at short notice.