Most analytics can be described as looking in the rear view mirror. At best you have a good view of what you have done and where you've been in the past. 

You need to look in that rear view mirror frequently  to check on threats (fast oncoming traffic), and anticipate potential problems before changing lanes and direction.

Of course you should spend most time looking where you're going aka "predictive analytics". Trouble is, driving an enterprise is far more complex than driving a vehicle.   

There is often no clear mirror of the past. The data is incomplete, redundant, plain wrong.Even worse, it is often inaccessible (unstructured, semi-structured & often external)

Insurance companies are full of big data and analytics technology.

Before predicting the future, here is a small challenge  for such organisations.

One of your fraud investigators, applying common sense and a nose for dodgy claimants, spots a suspicious motor accident claims notification form (CNF). 

In the free text field it states " I was parked on the ACME garage forecourt, Tysley Road, Birmingham when vehicle XXX reversed directly into me". 

Elsewhere in the structured fields of the CNF it states:-

  •  5 Occupants
    • 1 Driver
    • 3 Adult passengers
    • 1 minor
  • Time of incident 01.00
  • Solicitor ( Law Firm) Mike Daly Law Partners
  • The driver made the initial claim and the other occupants followed with separate claims.

Here's the challenge.

1) How long will it take to analyse all CNFs (without setting up SQL queries)  on the UK Ministry of Justice (MoJ) portal to find how many instances of:-

  • CNF with 5 occupants at Tysley Road, Birmingham or mentioned ACME
  • Between specified dates
  • Mentioned ACME
  • With any of named occupants

Then second search 

against each of named occupants in any CNF in same period

Third Search

All CNFs in which Mike Daly Law Partners the Law Firm in that period

Just yesterday I did just that with 360Retrieve and it is the means of uncovering insurance fraud rings- in this case fraudulent whiplash claims. 

Current document management & case management systems, and analytics apps, are too rigid and do not allow the fraud investigator to follow a hunch and uncover costly fraud.  The humdrum low unit value but high volume whiplash type claims. 

So before I start predicting I had better be able to analyse the past and present. I had better know where I stand on the Business Analytics Maturity Model.

If I am far enough advanced I can not only plan predictive analytics but also maybe even automated decisioning.  No need to look what's ahead on the road- let the vehicle drive me!