Having worked in an insurance company for almost 13 years, it always surprised me just how much we relied on the trusty Claims Loss Ratio (CLR). For those that are not familiar, this is an underwriting metric and its calculated quite simply by totalling the claims paid out in a period, and dividing by the income taken in the same period. It's calculated at policy level to allow aggregation at various different points such as individual, client, product, sales agent etc.
In the company I worked for we considered that below 70% we were making a profit, between 70-100% unlikely to be profitable and above 100% obviously making a loss. The problem with this, certainly for long established insurance companies, is they tend to have huge variety in terms of customer demographic, which in turn tends to differences in customer behaviour. Depending on the customer behaviour the point at which you make profit could be very different and this is not evident using just the CLR calculation.
To illustrate this, lets consider a retention scenario for two customers. Customer A has a 75% CLR and Customer B has a 65% CLR. With the above assumptions its a very easy choice, we make more effort to retain Customer B.
However if we don't just rely on CLR and dig a bit deeper in the data that's readily available. We can see that Customer A only claims online, customer B submits paper claims. Customer A is paid by direct credit, Customer B still receives cheques. Customer A self-serves, Customer B has multiple contacts for every claim: phones calls, emails and live-chats, all of which have to be serviced by an agent. Customer A is a lot cheaper to do business with, therefore can be profitable with a higher CLR than customer B. The retention choice is no longer such an easy one.
This is where a Service Cost Ratio can really make a difference in understanding which customers are actually profitable. Most businesses track every customer interaction, and If we assign a cost for each of those interactions then we can calculate the cost of service over a period of time in the same way we calculate claims. If we use the same period of time then we can add the two together to give a combined cost ratio.
Back to the illustration:
Customer A had a 75% CLR so lets say with their behaviours they have a 10% SCR giving a combined cost ratio of 85%.
Customer B has a 65% CLR and with their behaviours they have a 30% SCR giving a combined cost ratio of 95%
With the additional data we flip which customer we are more interested in retaining, its just a case of joining up the information that is already being tracked.
In deploying a Service Cost Ratio you can affect many aspects of the business. As mentioned at the start a business can aggregate at any level it wishes. Take the example of sales agent and influencing their behaviour. A business could create a commission structure with a rule aligned to the agent's aggregated SCR. This will encourage the sales agent to do everything they can to make sure it's as cheap as possible to do continued business with that customer. They would be wise to show the customer how easy it is to claim online, how easy it is to self-serve, and they will make sure they have all the details for the customer, and not just the bare minimum to set up the policy. They would be motivated to make this change because they will be paid a better rate of commission for doing so. This change in behaviour for the sales agent means that new business will be more cost effective because they will be easier to serve than legacy customers. Legacy customers could be dealt with at the point of renewal with a fixed price in return for demonstrating a change in their servicing behaviours.
The customer also wins; not only do they get a more efficient service but as we can now make a profit at a higher CLR, more attractive propositions can be created with better pay back rates for the customer. More attractive propositions should lead to an increase in new business.
Contact us today to find out how we can help your business.
Lawrence.Neale@telosanalytics.co.uk
