Data analytics nears adulthood but has it grown up yet?
Here's how things work in the real world...
The problem with business analytics technology is that it has just about emerged from puberty and adolescence but nobody has given it a bank account and introduced it to the opposite sex yet.
Analytics has been around in one form or another ever since Henry Ford decided to catalogue his motorcar assembly line measurements a century ago. So far the computerisation of analysis engines has arguably had a less tangible business impact.
In a world where every large organisation has its own enterprise resource planning (ERP) system, or software designed to shoulder customer relationship management (CRM) or some form of business process management (BPM), how do we start to look for real competitive advantage?
Back to the future
“We have managed to use the BPM and ERP systems to identify where invoices and product orders were being unnecessarily slowed down, leading to a shortfall in profits,” says the business analyst with a six-figure salary and a smooth line in management consultancy speak.
This is good, but it does not necessarily realign the firm’s market strategy. Fine-tuning is all very well, but deeper analytics for decisive strategic business direction is needed.
Time to introduce "descriptive analytics" and "predictive analytics”. Descriptive analytics has often been used to drive business intelligence (BI). It uses historical data to rationalise and explain what has already happened, while predictive analytics evaluates patterns and relationships to identify what might happen in future.
“The advent of real-time data virtualisation, aggregation and processing has enabled automated, actionable insights through predictive modelling,” says SAP UK sales director Rob Coyne.
“This is already leading to the creation of advanced, almost-neural systems, which can learn complex patterns amid large data sets to predict the probability that an entity will exhibit behaviours that are of interest to the business. It is not confined to structured data."
Of course SAP thinks it has the inside track on analytics because it has engineered its HANA high-performance analytical appliance specifically for an Intel chipset to perform in-memory tasks faster than most.
The product can shoulder applications that consume massive volumes of data in near real time. So is this where competitive advantage through analytics will be brought to bear?
Well, it is not all just about speed and power. A company has to know what to do with all this data.
Intel IT Labs engineering specialists Moty Fania and John David Miller asserted in a recent paper that advances in parallel computing enable us to handle big data sets differently.
“It is becoming standard practice to capture and store information well before its value is completely understood,” they say.
“Developing the organisational skill to mine and process big data to perform predictive and descriptive analytics will be a key driver of performance in the future.”
Intel confirms that it is in the early stages of developing the capacity for better analytics and anticipates growth for these tools in R&D, cyber security, design, manufacturing, operations and human resource management.
“In the past, most companies could only try to aggregate data for analysis, or take samples and try to extrapolate meaning from them,” say Fania and Miller.
Both men agree that while skills in statistics, mathematics and machine learning will be important, the skills required to align the data to the business and find positive outcomes are what will make analytics truly shine.
We are entering a new phase of computing where competitive advantage will be gained or lost based on the quality of data and the ability to analyse it.
This trend could accelerate the adoption of high performance computing (HPC) infrastructures and workloads in the general business market. Your typical corporate IT shop will be asked to do more and it will be asked to do it faster than ever before.
Speed and power are important, business-cognizant machine learning HPC-related skills are important, and so is the underlying hardware and software architecture. Taken as one, this is a big ask.
So how do we juggle all these factors as software developers-cum-database management operational strategists, or whatever new hybrid job title will describe the new über data programmer?
Clive Longbottom, director and founder of analyst house Quocirca, thinks the market still has a way to go.
“Sure, we have moved from in-the-mirror business reporting on what has already happened to of-the-moment business intelligence giving insights into what is happening now. The predictive side is emerging, but slowly. It runs the risk of being by-passed by its own big-data bandwagon,” he says.
“Current business analytics is fine if everything you want is held within formal data sets in SQL-style databases, but it still struggles with less formal data, and particularly with areas such as social networks and external, web-based information.
“Some vendors, such as SAP, IBM, EMC and Teradata, are beginning to focus on all these areas as a singularity. Once everything can be included, we can call it true business analytics.”
Automation will shoulder some of the back-breaking donkey work
So there is a data management proviso here. We need to be able to lasso and corral all the data in the known universe into our analytics data suction pipe. This is another big ask, so a degree of automation will shoulder some of the back-breaking donkey work.
Nick Patience of Recommind, an unstructured-data analytics company, says that although human intuition is still of the utmost importance in this process, we can also use software capable of categorising any data in context.
So-called “supervised learning and automatic categorisation” technology can help to identify the relationships between data entities, such as people, titles, instances, dates and departments, he says.
“This software also provides a broad overview of the kind of information that resides on a company’s systems. It avoids the need to build a team of people to go through the information that businesses have stored on their repositories, map it out and find the relationships between entities.
“Instead you can use the experts in your company to teach the technology to focus on what matters to your business – what it needs to compute.”
The awkward age
Going back to our earlier analogy, perhaps this is why business analytics has been accused of being a bit spotty and premature, often bursting into implementation and deployment before information has been properly collated.
It is down to us pesky humans after all and our inability to complete complex OLAP-based analysis tasks that we claimed we were capable of. This is what leads data engineers to use (often irrelevant) pre-canned reports and templates to try to carry out their work.
We need to move towards the new era of in-memory visual analytics tool, according to Guy Cuthbert, managing director at Atheon Analytics, a data-visualisation vendor.
“Visual analytics rapidly and clearly highlights the exceptional areas of performance – the good and the bad, the areas that need to be improved or demonstrate opportunities for growth.
“Exploiting visual analytics can give access to sales patterns and promotional performance to uncover sales opportunities and improve supply-chain performance,” he says.
The trouble is, things get more complicated before they become straightforward. Part of the issue is that data analytics has to cover multiple sources of data because most companies do not store all their information and data within one app such as ERP, CRM, HR and finance. They all provide different versions of the truth.
Brad Peters at Birst, a provider of BI solutions, argues that bringing this information together into one coherent whole is what makes a difference here.
“The coherency point is a good one to make. It helps that data quality argument be made,” he says.
“Cloud-based BI delivers a new level of flexibility and agility. With less reliance on IT, line-of-business teams can roll out advanced analytics that gives business analysts (and other employees) the ability to discover insights and efficiencies from their data.
“I see this whole industry developing to become a standard office tool in the future. Whereas in the past people used spreadsheets for simple analysis, cloud BI could replace that use case.”
Keep it simple
So where and when will business analytics actually be able to deliver competitive advantage?
The answer lies in IT that is “less artistic and more pragmatic,” says Davy Nys of Pentaho, a business analytics company. We need to look at ways to simplify and shorten the entire data workflow, he adds.
“This includes using development tools to automate tasks, like MapReduce jobs and scripting. This will enable IT to be much more responsive to business change," he says.
“But as we now ramp up business analytics, we caution end-users against 'going rogue' and implementing business analytics solutions without involving IT.
“If anything, data management is becoming more complex as variety, volume and the need for speed increase. If you thought rogue spreadsheets were dangerous, just wait for the havoc people will wreak with rogue analytics.”
Are we at a tipping point for data analytics then? It certainly seems so. Could the new toolsets being developed today be aligning to the perfect storm of in-memory processing, cloud-based delivery and unstructured big data management? The vendors would have us say yes.
Did Henry Ford teach us anything after all? He said quality means doing it right when nobody is looking, so let’s get analytics 2.0 right before we shout about it. ®
Sponsored: Benefits from the lessons learned in HPC