Practical steps towards democratising analytics
A panel explores the competitive advantages to be had by spreading the knowledge and how to go about it, as John Leonard reports
The democratisation of analytics via self-service tools and data integration was the subject of a Computing web seminar entitled Practical steps towards democratising analytics today, which sought to tease out the practical aspects of becoming a data-driven business.
The arrival of big data integration, discovery and visualisation tools in the past couple of years together with the ability to store and analyse large datasets rapidly - sometimes presenting results in real time - is changing the balance of power. Davy Nys (pictured, second left), VP EMEA and APAC at analytics vendor Pentaho spoke of the pendulum swinging historically between IT (as with datawarehouses and BI tools) and the business (spreadsheets and dashboards). Currently he said it is swinging towards the business.
Trevor Attridge (pictured, second right), global director of data and technology at marketing firm MEC, said that ideally the balance of power should be halfway between the two, with IT handling the data governance and integration side as well as selecting the tools, and the business mining that data to improve processes and make them more efficient.
"The business cares about speed to value and IT cares about governance and security. We want the pendulum to sit firmly in the middle. Business can concentrate on leveraging the data and IT can deal with security," he said.
There is also a middle ground where IT, the business and possibly external analysts can work together, blending data in different ways to test scenarios and create a competitive edge, he suggested.
Previously, of course, data mining and analytics was very much within IT's remit, but Attridge rejected the suggestion that IT is going to become peripheral as top-down BI systems become less important.
"The role is changing but it is not diminishing," he said. "As data grows exponentially and becomes more prevalent through the business then the role for IT will continue to increase."
A culture of honest representation
A member of the audience wondered whether the tools are really there to allow non-technical staff to mine and interrogate data to provide useful information. Nys said yes they are.
"The technologies are there. Right now there are no technological limitations with everything from real-time processing, through to predictive data mining algorithms that can help you understand behaviour, next best action and so forth - technology is not the block," he said. "Fear of the facts", however, is an impediment.
"If you have your Excel file you can manipulate the numbers and you can make yourself look good or your department look good. That's a lot more comfortable than when all of a sudden you have to face the truth," he said.
Facts generated through analytics may be interpreted in different ways according to the context, but the important thing is that they are trusted and treated with respect.
"At some point you have to say, 'if we really want to start to base decisions on this we have to make sure we're consistent'," said Nys.
Rather than decision-makers jumping on a single pretty picture that just happens to support their views, or hiding behind data protection as a way of shoring up their authority, the democratisation process needs to be "industrialised", he said, with architectures, policies and governance in place to create a "culture of honest representation".
Attridge agreed, going on to say that having "facts" does not mean everyone has to agree, because facts and outcomes are not the same thing.
"It's probably a positive thing not a bad thing: if you're arriving at the same conclusion, fantastic. But you have different scenarios and outcomes using prescriptive and predictive analytics and you need a mechanism of precedents, how you learn from it, how you feed the information into the process going forward."
Knowledge or non-knowledge? That's the distinction
More and more workers will be working "inside" the new standalone tools in real-time, Nys said, rather than referring to business applications on an annual or monthly basis. He drew a distinction between knowledge workers, who have the power to use the tools to drive decisions, and non-knowledge workers who can refer to ready-built models passively via a dashboard. The sort of information used by the latter, he said, should be distributed without too many restrictions. But access to the information used by decision-makers and the sort of functionality used to drive outcomes is a different matter.
Attridge agreed: "You have to be very careful about who you give that data to because it could be counter-productive," he said. "If you're going to invest in this kind of technology you have to train people about what the expectation is around the data. It goes back to the culture of who is in control and who can make informed decisions... Here's what's expected of you and here are the tools we can give you."
However, Nys cautioned against replicating the old data warehouse model by implementing too deterministic a role-based architecture.
"The beauty of where we are now is that there are a lot of unknowns," he said. "We have the flexibility now to ask the questions we don't know we are going to ask. In the architecture you have to build enough flexibility."
Business as usual, just better
Attridge said when rolling out analytics it's best to start with quick wins that will be popular with the board, such as making an existing business process more efficient, then to move on to the more esoteric stuff once it has proven its worth.
"By undertaking analytics project you look at inputs and outputs that feed the project and that in itself can give you tangible efficiencies and savings. So, it's business as usual, just better."
The step change comes when you have informed users with confidence in the data who can work in pursuing new opportunities: "Start low and then incremental gains," he advised.
Nys gave an example of a hospital where access to KPIs on their unit, their department and even their shifts was rolled out from the administrators to the doctors and nurses.
"You saw instantly the impact on the organisation. All of a sudden you have a level of vindication, nurses were competing to get the best time for an operation and so on - that was an amazing example of how having the information was motivating to them to do their job," he said.
Despite uncertainty about ROI among some audience members, Attridge said the cost of investing in analytics now will be as nothing compared with failing to do so, because using data to generate knowledge is now a key differentiator.
"Look at the new businesses coming online: AirBnB, Uber - they have barely any physical assets. If you are in insurance for example, where will that be in 10 years time? How important will data be to the organisation? With a bit of smart thinking, the costs question will disappear."
You will shortly be able to view this web seminar online. Register now