Consumer Behavior

It’s always valuable for businesses to have some insights into which customers are most likely to complete a sale so they know who to devote their time to. This was a question for an Australian business that sells investment properties. After spending considerable time with each client they were faced with people who wanted, ‘to think about it’ and then disappeared and people who kept indefinitely postponing a buying decision because the time was never quite right. The biggest impediment though, was customers who were not able to get a loan to finance the purchase.

This business wanted to know if there was a way to predict which customers were likely to end up closing on an investment property based on simple information that could be gathered naturally in the first five minutes of a conversation. With that in mind, we used machine learning and data mining to look at 5 customer attributes: whether they were employed, what their occupation was, how long they had been in that occupation (less than 1 year, more than 1 year, more than 2 years, more than 5 years more than 10 years), relationship status and whether they still had children at home.

Not surprisingly, we found people who weren’t working (unemployed or retired) never closed on an investment property. However, we found that 95% of people who had been in their occupation more than 10 years closed on an investment property. Interestingly, we also found evidence that people in certain occupations including engineering and nursing are very likely to close, while people in other occupations including hospitality, childcare and retail are very unlikely to close. The chances of these people closing or not closing based on occupation does not appear to be influenced by other attributes at all. We suspect some, but not all, of the explanation is because occupation is often a good proxy for income which in turn is a major determinant of ability to obtain an investment loan.

While an encouraging start, we are cautious because this analysis was not conducted on a large data set and we have suggested revisiting this analysis at a later date when this business has had a chance to compile more data.