A web-based startup.
The startup has built a large email list to offer the chance to those on the list to change energy suppliers. However, it did not have a clear idea of the value of this list and, specifically, how many people would be likely to make a purchase in future.
Faculty analysed customer spending intentions by first tracking the number of interactions. We found that more than three-quarters (76%) of all purchases were made by customers within two weeks of signing up to the email list. We then used machine learning classifiers to uncover the spending intentions of the remaining 24% of customers on the list.
We based our predictions on a set of variables, including the way that people interact with emails, such as emails received, clicks and open rate. Our testing program determined the thresholds beyond which additional emails had no effect at all in encouraging people to buy.
We then helped the company to develop an email communication policy based on the results of this modelling to target customers more effectively. New customers are emailed frequently over the first 14 days of signing up, then less frequently until they start to show behaviour that indicates that they are thinking of purchasing (which might include visits to the website and the opening of emails), at which point email frequency is increased to drive conversion.
Our model was able to predict with an accuracy of 92% whether a customer would make a purchase after the initial two weeks. Using it, the company can market more effectively and has been able to identify 10,000 customers who have a high chance of conversion.