How to use Big Data Analytics to improve performance

How to use Big Data Analytics to improve performance

Big data is a buzzword; the definition of a ‘buzzword’ is: something that everyone talks about; different people understand different things, everyone thinks they should be doing it and the funny thing is that most of them do but they don’t know they do!

The Friday before last, I was lucky to participate in the Teradata London Roadshow about big data analytics; it was recommended to me by Vodera - a DigitalMR partner in technology related projects.

A former executive of Telefonica, who was a speaker, borrowed the following analogy from Dan Ariely: “Big data is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it...”

Data Scientists

Big data seems to require data scientists to be part of the equation. The conference speakers dealt a lot with what makes up a good data scientist; apart from them being able to work with machine learning algorithms, everyone seemed to agree that they need to be curious and creative before anything else. Academically speaking, the route to being a data scientist has to start from mathematics and whether through statistics or computer science to lead in business science or engineering.

Predictive Models

Business intelligence is a good practice and still has its place in driving business performance but without predictive analytics, an organisation of today cannot expect spectacular results. Predictive modelling is about running machine learning algorithms such as CHAID, Neural networks, Random Forests, Multivariate Adaptive Regression Splines or SOFM on data to find plausible and usable models. Incidentally, a combination of machine learning algorithms and computational linguistic tools are used for eListen the multilingual social media listening solution of DigitalMR.

Apparently, words like ‘innovation’ and ‘visionary organisations’ are closely linked to the use of predictive analytics nowadays. However, sometimes it is not enough to run only one predictive model to get meaningful results, it may be required to run another predictive model on the first’s predicted results; case in point the last Obama campaign used predictive modelling to identify swing voters (impartial /unpredictable political party voters). That, on its own would not have been enough since they did not know what would make the voters vote for Obama; they ran a second predictive model among the swing voters to discover what approach to use for whom. The result was that a subset of them were not worth going after - as no positive result could be expected, another subset, preferred to be reached face to face and for others, a telephone call would suffice. Well, Obama won the election and even though it is not conclusive that the specific double predictive model was the reason, it is almost sure that it had contributed to the success.

Data in Motion

The notion of ‘data in motion’ was mentioned a few times during the day. It is a very exciting idea to consider the power of predictive models combined with real-time data and real-time action. Imagine a consumer walking down a Supermarket aisle (such as Tesco’s), who according to the loyalty (card) data are likely to buy a six pack of Coke; the moment they pass by the carbonated soft-drinks shelf, their mobile phone beeps. An SMS was automatically sent with a 20% discount code if they buy two six packs of Coke or Pepsi for that matter. This is what TIBCO calls “complex event processing”. Here is how they describe the opportunity on their website:

“…extract actionable real-time intelligence from high volumes of fast-moving data – enabling you to rapidly capture, analyse, and act on the trends, opportunities, and risks significant to your business.”

Data in motion is “hot data”: that, with the help of predictive models, separates the signal from the noise, to enable the instantaneous conversion of it to an actionable insight.

In Closing 

This is the era dominated by social media; the consumer of today - as we hear constantly – has a lot of power that can impact brand success. Gone are the days of one-way communication from the brand to the consumers.

One of the big data streams that may/should flow through the doors of every organisation today is data from social media. Along these lines, active web listening or otherwise known as ‘social media listening’ can be used to respond to queries and fix operational issues as they are reported by customers online. What cannot be ignored however, is that social media data should also be used to analyse the topics discussed and look at the sentiment for individual brands (within popular topics). Social media listening is an invaluable tool for brands to better understand the purchasing behaviours of the consumer of today and to identify influencers for customer advocacy.


Share this article: