Analytics: The Power of the Law of Association

Analytics of late, is on the main stage for several organisations across industries. I have seen several implementations. As a consultant evaluated the cost benefit (investment to business value) and viewed it donning the hat of an information system auditor. This article is written from that experience. 

The business value is in the design of association of underlying data. This must also factor in, an organisation’s role based access control architecture to make sure information is available on ‘need to know’ basis. To achieve this, the designer (of analytics content) must understand the business domain and have a clear distinction of transaction data model as opposed to analytics data model.  

To illustrate the importance design, I will use a simple example of progressive association of data that with each association increases the potency of information.  

Let us start with a data point, a 9-digit number ‘123-456-789’. A sample size of 30, will exhibit a pattern. This pattern will help to generate an infinite series. This makes a very exciting coding algo for a novice.  To a person in analytics, an exercise in futility, meaning nothing.  

Association 1:  Metadata

We will add an association to the number, the metadata ‘Social Security Number’ (SSN).  

SSN: 123-456-789

The 9-digit number jumps to life, yet lacks the firepower. Let us do a few more such associations.

Association 2:  Data Point + Metadata

The next data point we will associate is ‘Name’ with the metadata ‘Customer Name’.

CUSTOMER NAME:  XYZC

SSN: 123-456-789

Association 3:  Data Point + Metadata

The third data point with metadata we will associate will be ‘Postal Code’. The next level of transformation will look

CUSTOMER NAME:  XYZC

SSN: 123-456-789

POSTAL CODE: NNNNN

 Association 4:  Data Point + Metadata

Finally, we will add ‘CREDITSCORE’ metadata. The next level of transformation will look

CUSTOMER NAME:  XYZC

SSN: 123-456-789

POSTAL CODE: NNNNN

CREDITSCORE: NNN

 A simple, four level related associations has transformed a meaningless number to a powerhouse of information on a unique customer, localizing on place of residence with credit rating. This information can support decision-making. It is clear I have used an obvious example for illustration.

 Now, going back to design, the start point will be the question on who in the organisation is the end user. In a financial institution with a significant credit portfolio, the person in risk will have a perspective different from that of a credit officer. The head of credit may use it for information only.  The chief credit officer plans for expanding the portfolio, while the chief risk officer is focused on concentration risk. In all the above cases, we are talking about the same underlying transaction data. This is the challenge that analytics must consider from visual content to the drill downs. A daunting task, indeed.

An extensible design and data model is a prerequisite for using analytics productively in an organisation. The swanky dashboards and colourful content can come next. Off the shelf, products will not always be the best solution. It is best designed bespoke