Why Do We Need to Predict What We Already Know?
Updated: Feb 9, 2020
Over the last 20 years a vast amount of the insurance industry resources have been allocated to Big Data projects and the hiring of skilled data scientists. However, the question still remains; Do we need predictive analytics if we already have a wealth of customer data that tells us exactly what the customer wants?
Data science has become a major part of the insurance industry over the last 20 years. Insurance companies all over the globe have been investing heavily in Data Science departments who are tasked with handcrafting cutting edge predictive models to make sense of customer data.
Why do we need to predict the customer's actions if we already have a clear idea of what they want based on our historical data?
This question has arisen due to the nature of data science and predictive modelling. Put simply, it's assumed that the more we know about a customer, the less necessary it is to pry for more data.
For example, let's consider the popular navigation app, Waze. Waze earns our preferences by carefully monitoring our behaviour. We enter our work address in the morning and drive to work. In the evening, when we're ready to leave work, we put in our home address.
By repeating this pattern only five times Waze learns that when we get in the car in the morning, we want to go to work. It doesn't need advanced prediction modelling to figure out our next move because we've already told it what our next move will be.
“I would say, a lot of the value that we’re getting from machine learning is actually happening kind of beneath the surface. It is things like improved search results, improved product recommendations for customers, improved forecasting for inventory management, and literally hundreds of other things beneath the surface.” Jeff Bezos
Similarly, Amazon analyses our browsing history to decide what to advertise to us. Google analyses our search engine behaviour so that the search engine becomes more adept at delivering us exactly what we want. We train the model by telling Google what we want.
Big Data is closely connected to Artificial Intelligence because the more data we have, the more intelligent our systems become. When we have a wealth of clean data, there's less need for predictive models that fill in the gaps in our knowledge.
No Need To Ask Again
Insurance companies have huge troves of historical data due to the tens or possibly even hundreds of millions of policies sold over the years. This data is invaluable when it comes to predicting the wants and needs of new and returning customers. For example, let's consider an American insurance company based in New York.
The company may have sold millions of insurance policies since they were founded and provided countless more insurance quotes. If a new customer comes in looking for home insurance, the company can draw on the millions of policies of people in a similar situation (the type of building, the value of contents, the profession of the client, etc).
If a returning customer approaches the company looking for a new policy, they already have their preferences on record and can offer a suitable policy based on this data. Even if the customer has acquired new preferences, these preferences can be fed into the collective company data to help inform new policies.
Put simply, every piece of data, whether it's from new or returning customers will contribute to improving the experience for new customers. The system remembers everything and the model improves each time it gets new information.