NICTA's Dr Rami Mukhtar says much of the problem with Big Data analytics is to do with human input, not only are we bending output to suit our agenda - we may not be even asking the right questions in the first place.
Moving the data industry on from simple data warehousing to Big Data analytics has seen a boom of interest; an 'insatiable hunger across enterprise for actionable insights."
C level executives now see the potential of Big Data, and are looking to data scientists to provide the 'golden nuggets', the opportunities to drive business forward.
For example, insurance companies now know that customer churn is directly related to the number of policies a customer owns with the brand. Therefore, try to sign up customers to as many policies as possible.
But Mukhtar believes that these kinds of correlations are very dangerous for business when clashing with the human factor. It can see executives only taking note of correlations data when it backs up their own ideas.
Even worse for IT on the shopfloor, if you produce data that doesn't produce the correct outcome, guess who is shown the door? It certainly isn't the executive. So how can you ensure Big Data analytics are used correctly?
Mukhtar said it requires a fundamental reshaping of how the business structure works, it needs to be used to measure the impacts of aspect on the business. Data has no bias, so ask it questions accordingly. See analytics is a revenue boosting/cost saving machine.
Ask the why. What is our objective? Is it to cut costs or generate profit? Analytics are more useful in business when they produce a simple number. It is the configuration and deployment of this 'machine' that proves most difficult for organisations. And this testing needs to be done before it is deployed to the front line.
Computers need binary inputs, parameters to work by. So ensure that you are working to this methodology too.
For example, in a supermarket chain, instead of pursuing 'human' C-Level objectives such as 'how can we reduce staff costs, and minimise the wait time for customers?' it's better to use data analytics to look at numbers, such as how can we produce more profit in six months. When dealing with big data, you would need to define to a number what a long 'wait time' is. Staff costs needs to be defined as shift rosters, for example.
"How can we parameterise business?" Mukhtar said.