Keeping customers has always been essential to any business.
Time and effort spent on retaining customers is always worthwhile, as long as you are speaking to the right customers. Missing a key conversation with those customers means they may have decided to go elsewhere.
It is simply too late – and you are stuck wishing you’d got to them sooner.
If only you had known that you needed to pick up the phone, if only you’d known the price was just too high, if only… but what if there was a way to catch these before it is too late?
Robotic renewals
Acturis empowers efficiency, not least at renewal. The Acturis AutoRenewals service can automate anything from simple processing to complex renewal flows, saving vast amounts of time and effort.
The benefits are clear; the average number of renewals managed per handler for brokers e-trading SME business on Acturis was three times higher for those using Acturis AutoRenewals than those processing renewals manually. This efficiency gain came at no cost to retention rates. Retention rates were in fact around 5% higher for AutoRenewals policies.
However, the lack of personal touch could mean some customers fall away at renewal when the relationship could have been salvaged. Can the huge benefits of AutoRenewals be achieved while still getting in touch with customers that need a bit more attention to keep their business?
Room for improvement
The data science team at Acturis decided to look at the data to see if there was a way we could apply machine learning to help solve the problem.
For those not familiar with the technology, machine learning can evaluate past data and discover complex historical patterns. The machine learning can then be used on new, unseen data. It allows us to predict outcomes for new situations using lessons from the past.
However, machine learning is only as good as the data behind it. To understand which data will make the difference, we first have to understand the problem.
When we analysed the data, the first lesson we learned was just how complex customer behaviour was at renewal, especially during the pandemic. However, as we progressed with our analysis we discovered that 60% of lapsed policies processed by Acturis AutoRenewals could be classified as potentially preventable: the loss was down to service levels, price or lack of contact.
Clearly there are improvements to be made to maximise retention on top of automation.
AutoRenewals Retention
It may seem like spending time manually renewing all your business would lead to a better outcome. But that simply isn’t true. Acturis data clearly shows that automated renewal books have a better retention rate than purely manually renewed business.
Brokers using Acturis AutoRenewals across e-traded SME business have retention rates 4% higher on average for 2021. Acturis is confident combining the Renewals AI service with AutoRenewals will drive this difference yet further.
AI – empowering brokers
With this analysis, we started to prototype and develop some machine learning models. The customer behaviour was complex and large numbers of prototypes were needed to find and train an effective AI.
The current machine learning model has a set of over 50 data inputs, most of which are transformed and aggregated from several data points themselves.
Despite the complexity, we have been impressed by the ability of the machine learning model to identify potential flight risks at renewal. We tested our model by showing the AI hundreds of thousands of 2021 renewals it had not previously seen. We then compared the actual result of the renewal to the prediction generated by the AI.
The prediction was given as a score between 1 and 100; 1 being most likely to lapse and 100 most likely to renew. The results in the table below show how well the AI could categorise renewals.
The devil’s in the details
At this stage, we were confident the model could help brokers identify the renewals that needed more attention. But any broker needs to know why that renewal needed more attention – otherwise their hands are tied. For the model to be truly transformative, it had to give brokers the information required to act on the AI prediction.
To do this, we needed to explain how the AI had come to its prediction.
This involved developing further models to best describe the decisions happening within the multi-layered neural network model we had trained. The results of these models then need to be converted into plain language.
The end result is very powerful; an AI that can predict the result of a renewal and detail the most important factors that went into that prediction. This gives brokers the detail they need to fully utilise the power of AI.
Explaining AI
The ability to explain how complex machine learning models arrive at their predictions is essential to build trust in the technology. Luckily this is an area where there have been rapid enhancements in capability in recent times.
The Acturis data science team has combined several advances in the field, along with various other methods, to develop the renewal assistant AI as being inherently open in what drives its predictions.
This allows us to present the prediction along with clear, business-driven factors behind the prediction in plain language, empowering confident decisions.
Reinforcing strong relationships
As we investigated renewals and focussed on the problem of increasing retention rates, we considered if there was something that could be done with the strong renewals; those the AI model was classifying as highly likely renewals.
Could those strong customer relationships be improved in a similar way to the teetering ones? What if you could deepen those strong relationships?
We had already been developing a cross-selling AI that identifies the most appropriate and likely classes of insurance required by the customer. It seemed like a natural fit; as a customer is coming up for renewal and touch points are already in place, start a conversation on other covers the customer may need. This makes even more sense when the existing business with the customer is in a strong place.
The Acturis data science team are excited by the potential of these services, both on their own and when deployed in conjunction to drive added value to the renewal conversations happening between broker and client.
If you would like to talk further about our renewals or cross-sell AI enhancements for AutoRenewals, or AI and machine learning at Acturis in general, please contact DataScience@acturis.com.
If you would like to talk to our AutoRenewals Team about automating your renewal processes, please contact us at AutoRenewals@acturis.com.
Gordon Jenkins is the Acturis Data Science Product Manager and has been working with data at Acturis since he joined us back in May 2008.