Analytics - it's not rocket science!

You don't need to create to innovate

Any child of the 70’s in the UK will remember with fondness the frequent appearances on television of the science presenter James Burke. Burke established his reputation as a reporter on the BBC science series Tomorrow's World and he was the main presenter of the BBC's coverage of the first moon landing in 1969.

Connections

In the late 70s Burke presented a documentary series called "Connections” (check out episodes on Vimeo and YouTube) which explored an "Alternative View of Change".  It rejected the conventional linear view of historical progress and innovation and explored the reality that change comes about as a result of a web of interconnected events.  Think of it as a sort of historical “six degrees of separation” game.

Burke explored how innovation comes from connections. In one episode, Burke showed the connections that led from Napoleon to man landing on the moon – via bottled food, tin cans, air conditioning, refrigeration, thermos flasks and ultimately the Saturn V rocket!

In my career I have been fortunate to work with a whole range of leading organisations across many industries, including food manufacturers, pharmaceutical giants, banks, insurance companies, media, hi-tech, publishing, government, automotive, and retail. One of the most interesting aspects of my role is to make connections between delivering analytics in one industry or function and using the same techniques and approaches in another.

Some of the industries I have worked in are more resistant to the possibility that another industry could possibly do something better using a different approach – Investment Banking springs to mind – but many are interested in exploring the opportunities that until that point had not even been considered. In many cases they are not, as many of Burke’s examples were, “rocket science”.

Let’s explore two techniques that are common, well tested and have been used for years in retail and clinical trials that our teams are now applying to people analytics and the insurance broker market.

Survival of YOUR Fittest

Survival analysis is a technique where the outcome you are looking to predict is the time it takes for a specific event to occur. The event could be death, occurrence of a disease, marriage, divorce, etc. For years, this technique has been used in clinical trials and the development of drugs.

When one of our teams at Slalom was discussing the problem of staff attrition with one of our large media clients, we thought that this could be a great application of Survival Analysis. The event in this example being someone leaving the company. Using this approach, we not only identified the core drivers for attrition, we also presented a list of all their employees, sorted by their risk of leaving. Their reaction?

“Err yeah – that looks about right – the first two people on your list have actually just left…” 

We are now looking at applying this technique to a massive opportunity in reducing client churn in Insurance.

Cross about Cross-Sell

Secondly and one of the most interesting connections we have made, is applying machine learning techniques that are prevalent in retail, the likes of Amazon being the most obvious, to help identify cross-selling opportunities to the London Insurance Broker market (by the way, Amazon still doesn’t have this cracked - how many times after buying something, say a new TV, does it suggest you might want a new… TV?)

The market is massive, making up over 30% of the City’s economy and equivalent to all three of its closest rivals combined (Bermuda, Singapore and Zurich). But Insurance is steeped in traditional ways of working going back over 300 years and London’s dominant position is under threat. The insurance industry has remained largely unchanged and threats loom in the form of emerging markets, expanding competitors and new entrants. Can the London firms still rely on personal relationships and antiquated processes?

We didn’t think so, and neither did one of our Insurance clients. We showed them that the vast majority of their clients had just one product. Using collaborative filtering and association rules modelling, we built them a model that identified millions of pounds of incremental premiums and showed them which products had the highest probability of being cross-sold. 

So you see, it needn’t be bleeding edge, just something tried and tested in one area that has never been tried in yours. What connections could you make by thinking outside your industry?