Share this content
Thinkstock/iStock

Magic numbers: How customer analytics is saving financial services

by
28th Jan 2014
Share this content

Why is customer analytics so important to your business? Below, Oliver Werneyer, VP of data at global reinsurers Swiss Re addresses this question in relation to his own work and explains why he advises on a customer-focused approach as a better way to grow a business in the economic hardships of 2014.

MyC. Why have customer analytics become so important to the retail financial services sector?

OW. The retail insurance sector became very competitive after the financial crisis due to consolidation. Smaller upstarts at that time have disappeared which means the number of players in the market has not increased significantly over the last couple of years, and neither has the market share.

People are still as reluctant to buy insurance as they were before. Overall, the pie has not increased. The only way to get more of the pie is by having a good advertising campaign - but insurance is not just about the number of policies you sell, but also the quality of your business.

Analytics capability really help insurers not only to differentiate themselves, but spend their money more wisely. Analytics helps commercialise concepts such as propensity to buy – the people most likely to actually buy something, and also the propensity to lapse.

Also, it can help to hone your marketing campaigns and message. These days the concept of one insurer or product for everybody is outdated. People look for something specific, something tailored. Customer analytics really help you understand who you are doing this for.

MyC. What are the biggest benefits of a customer analytics strategy?

OW. One of the tools We advise our clients on is predictive underwriting and others include the propensity modelling  I referred to earlier.

The commercial successes of our clients’ campaigns have helped people spend their marketing budget more intelligently, their profit has gone up although the total number of policies sold have not increased from the year before. Because the quality of that business is so much better the value of the business is so much more and you can see that in the profit that is coming through and in improved prospects.

MyC. How have you used the data to build deeper relationships with your customers?

OW. We help banks make use of their transactional data and some health information to build a model which has the analytical capability to go through all their customers' information and tell them which ones are most likely to be healthy and offer them insurance without having to go through and or many pesky underwriting questions up front.

It's not going to make it cheaper, but it makes it more convenient to buy. Within the bank environment that's much more important than price because the bank only has limited time to chat to customers.

MyC. Could you tell us a little about how you have used social media analytics as a reliable fraud and risk management tool?

OW. What we’ve found is that public or social media data that’s publicly available is decreasing because services are now starting from a default privacy position. To build reliable models is difficult. Where we see great potential in the future is in doing cross checking to confirm information provided by policyholders for non-disclosure or for fraud.

Let's look at a hypothetical situation where someone says they’re a non-smoker but, at some point in the future, we would be able to find out, from the information that's publicly available, that they like two or three smoking brands on Facebook. We would not say that he definitely is a smoker and challenge him on his answer, but could recognise that he's more likely to be a smoker than a non-smoker. We could then possibly ask him to do a cotinine test to confirm his actual smoker status. For somebody who indicated that they are not a smoker and, in the future sometime, we find no such public evidence out there, we wouldn't send them for a test.

Just as another example, non-disclosure - somebody might say "I travel a lot", but then you see through publicly available information that they haven't actually left the country in the last three or four years, so, they were probably very active – but not anymore. Is it because a change in lifestyle, injury, disability or something innocent like a baby being born into the family recently? It might warrant further investigation. In future we won't be making assumptions about people, but asking them to answer questions more specifically where we might find contradicting information that is available publically.

But because a lot of this data isn't really publicly available we've not yet found a model that's particularly reliable in projecting that.

MyC. So how can you get around that?

OW. Unless you buy the data, which you cannot do [on Facebook], we focus on other models that you build through other data bases and get insights from those. For example, credit unions - big companies in US and the UK, whose job is to collect all data that's publicly available and create a dataset of people and sell this data to interested parties. We try to buy that information to apply insights - but then you’re using it on an aggregate level not a personal level.

MyC. Are there any other challenges related to social media analytics?

OW. Sentiment is really difficult to analyse because it can switch depending on the question asked. For example, if you do some sentiment analysis on social media content, critical illness comes up as a negative sentiment. But if you were to look at the comments, people might be talking about buying a critical illness policy - which is more of a positive sentiment.

It's really difficult to have masses of data and know what the sentiment is. A lot of social media is picture-based and we have no way of analysing picture content and the sentiment it's trying to convey.

MyC. Can you share any advice regarding how to tackle the massive amount of structured and unstructured raw data so that you can generate business insights?

OW. I would say that the key is not to forget the reason you are generating business insights. People must not lose sight of why it is that they are analysing the data and what they are looking to get out of it.

They might start taking too much data thinking they will get more insight. You can buy a database and get much more insight than before but only ever increase the business insight by 1 or 2% at a possibly significant cost.

Secondly, there’s the human element. The more data you include and the more complex it becomes the more you rely on systems and the more difficult it is to apply sensible logic to it so be sure the data you add does not create unnecessary confusion.

One example, and I don't know if you've read Freakanomics, but in there they have a lot of scenarios showing correlations between things. There was one case where they showed that the number of slices of pizza sold was correlated to an increase of crime in New York, for youths between ages 16-21. There might be a statistical correlation there but it probably doesn't make much sense to pursue. You wouldn't go out asking how much pizza people buy in a certain area and start arresting kids in an effort to prevent crime in the area. The more complex and comprehensive data you have the more erroneous some of these insights are going to be and you'll feel obligated to follow up on them.

The last part is don't lose sight of the implications. Even if you have this amazing data that you buy and say okay, anyone who likes this Facebook page and has a dog this colour this is the kind of person that they are.

If you can't ask the person for that data or to answer that question, all those insights aren't going to help with anything. That's got a lot to do with customer journey. People often develop things, think they're amazing and developed something new and innovative, and mess up the customer journey.

Oliver Werneyer will be speaking at the Customer Analytics and Insights in Retail Financial Services conference, being held on February 26th - 27th in London.  

Tags:

Replies (0)

Please login or register to join the discussion.

There are currently no replies, be the first to post a reply.