In business, we are often so focussed on getting customers in the front door, we miss the ones wandering out the side exit. This ‘churn’ can be a significant cost to busy brokers

So, earlier this year, AFG’s Marketing Team and the Analytics and Insights Team launched an ambitious project to design and build a new AI solution. The system would identify residential home loan clients who might be thinking about changing their current situation and help brokers to more effectively develop their communications and contact timing.

Our role at AFG is to help brokers help their customers.

Our role at AFG is to help brokers help their customers. That means giving them the tools and insight they need to talk to their customers at the right time with the correct information. As a data-rich organisation and business that processes nearly 1 in 11 mortgage transactions in Australia, we have one of the best insights into customer behaviour in the industry.

A huge opportunity for brokers

The task has involved completely re-engineering the Red Alert platform early warning system, using machine learning to ingest and process hundreds of thousands of data points and to find patterns that can predict future behaviour. Those patterns are then compared to existing customer data points to identify which customers are the most likely to be making a change in the immediate future. The results are automatically displayed for brokers so they can reach out to the customer or use the SMART Marketing Automation tool to trigger a scheduled communication.

We are excited to report the new system has made a huge leap forward in testing, reaching accuracy of predictions to 37 per cent.

AFG’s Head of Analytics and Insights Alex Maund believes it can get even better, aiming for 60 per cent accuracy, as the algorithm is being rolled out for live testing.

The new system ensures AFG brokers are talking to the right clients at the right time about the right issues – this is often called ‘Next Best Conversation’. On the flipside, it avoids pestering contented clients with scattergun ‘Can I help?’ emails.

Artificial Intelligence to drive better customer outcomes

AFG’s SMART marketing platform already has a sophisticated automated marketing program that triggers different email communications to brokers customers at different times in their lifecycle. Whether it be an anniversary, birthday or end of a fixed interest period. AFG Red Alerts is what really sets our Marketing-offering apart in this space.

Traditional email automation based on fixed triggers works well in some situations, but using it in combination with predictive analytics is cutting edge technology for brokers.

The result is not only improved broker efficiency, but better service for customers, who will receive an almost serendipitous contact from their broker right at the moment they are likely to be thinking about new options.

Another significant improvement is that the system is now able to identify not only who is likely to change their loan arrangements, but it also suggests why. For example, a new homeowner looking for their next property or an existing price conscious homeowner looking for a better rate.

Being able to deliver the ‘why’ helps brokers get ahead of the situation and has been a critical focus for the AFG Analytics and Insights team.

A very human start to a complex problem

Despite what science fiction movies may have you believe, often the first stage in developing a sophisticated machine learning model is decidedly old school for such a high-tech endeavour.

The Analytics and Insights Team began by tapping into the knowledge base of AFG staff, many of whom are former brokers, to brainstorm as many triggers as possible that may lead clients to reassess their residential home loan arrangements.

The workshops produced hundreds of Post-It note hypotheses covering everything from the end of a fixed-rate term, to a family windfall, becoming empty nesters or children starting high school.

Kanban pain points

These hypotheses were converted to data points, written into an algorithm and run against a decade of in-house customer data to see which assumptions held and which didn’t. Predictions about which customers would stay and which ones would churn were compared to actual outcomes to gauge accuracy.

With a bit of tweaking the algorithm was correctly identifying more than a third of all borrowers who would eventually leave their broker.

Along the way, there were some important learnings, the main one being: outside of those intuitive moments in time, like customers coming to an end of a fixed-term, gut instinct is no match for a machine learning algorithm when it comes to predicting customer behaviour.

Overall, the Analytics and Insights Team found that – contrary to popular opinion – there was no single demographic or data point that reliably indicates a customer is more likely to churn.

Instead, it’s ‘micro-groups’ of customers that meet several key variables.

The complex correlations are impossible for brokers to track, but easy for an algorithm to spot.

In identifying these key drivers of change, AFG plans to add an extra layer of automation to the Red Alert system to make customer service even more seamless for brokers.

Putting AI into action for brokers

Once a customer is flagged, the primary driver for change will trigger an automated communication, offering information on a range of alternative solutions – often called ‘Next Best Action’.

It is important to note that this new system is designed to augment the service that brokers provide their customers, not replace them. It will make the brokers more efficient, allowing them to have more meaningful conversations with their customers, and help brokers anticipate their evolving needs – well beyond when their home loan was originally settled.

The Red Alert system also enhances competition and choice for customers by suggesting credit options at an appropriate time. This helps brokers to continue to act in their clients best interests.

Leave a Reply