Don’t throw money blindly at BDA projects: Gartner analyst
By Edwin Yapp April 28, 2016
- A venture capitalist-type view is needed in data science
- Invest wisely for innovation, but be prepared to fail
NO-ONE would argue that data analytics is becoming more significant in the business world today, but enterprises should not throw money blindly at such projects as this may be counterproductive, according to a Gartner analyst.
Ian Bertram (pic), research managing vice president at Gartner, said that companies need to view data analytics as innovation projects that have risks attached to them, and should not expect every project to yield high returns.
Arguing that more money thrown at data analytics projects is not necessarily a good thing, he said such a move could even potentially stifle the very thing such projects are trying to achieve in the first place.
“The worst thing you can do is to manage data science projects the same way you would traditional management,” he told a media briefing in Kuala Lumpur recently.
Bertram argued that if a project had millions of dollars in investment, typical management thinking would expect high a return on investment (ROI) on that project.
Such expectations could potentially interfere with the project and put a lot of pressure on the project stakeholders to produce something for the investment.
“You can’t expect that every [data science] project has a definite return,” he said, adding that such projects should be viewed in the same manner venture capitalists (VCs) view their investments.
VCs would typically invest various sums in a breadth of different ventures so that the risk of these investments is spread out. They also typically expect up to 70% to 80% of their projects to not pan out.
Gartner itself believes that analytics should be viewed as three distinct functionality pillars: The Information Portal, Analytics Workbench and Data Science Laboratory (click diagram below to enlarge).
The Information Portal typically consists of ongoing data warehousing and business intelligence infrastructure, and is what an enterprise depends on to function on a daily basis.
The Information Portal has data that is trusted by design, and applications typically consist of dashboards and reporting tools. In a bank for example, this would typically house customer data, credit information, and other core banking records.
The Analytics Workbench is about data discovery and comprises some level of verified data that may not entirely reside in an enterprise’s corporate database.
Typically, this would be social media data such as Facebook and Twitter postings about an organisation, which has a relational context to the enterprise but is not necessarily core to the company’s function.
Finally, the Data Science Laboratory comprises advanced analytics and data mining functionalities, which normally work with unverified data.
An example of this kind of data could come from information from sensors and other Internet of Things (IoT) devices, which are not used on a daily basis but are important for advanced data science discovery projects, according to Gartner.
Bertram said that it is this Data Science pillar that companies need to invest in, but in doing so, such projects should be treated with the same mindset that VCs view their investments, where a company invests in many small projects and expect only a few to succeed.
“People need to take a leap into advanced data analytics without huge investments,” he said, adding that data science projects are very experimental in nature and cannot be hamstrung by the constant pressure of ROI concerns.
Addressing the right issues
When asked how many companies are even thinking about data analytics in Gartner’s three-pronged approach, Bertram acknowledged that not many enterprises see analytics this way.
But he argued that this does not mean that they cannot or shouldn’t organise themselves along these lines.
“In terms of the maturity of the three pillars, I would say that most large enterprises already have some kind of Information Portal and have invested in business intelligence and data warehousing, so I would give it a 3/5 scale.
“For the Analytics Workbench, I would rate it at 2/5, while a Data Science Laboratory would be about 1/5,” he added.
That said, Bertram argued that the main impediment for any enterprise to organise itself along these lines isn’t because technology disallows it, but is a question of softer issues such as people, culture, and skill sets.
He said organisations should begin by evaluating how important their data is to them, and understanding where it fits into their business strategy. Organisations should not just be collecting data and reporting on it, he cautioned.
Asked what needs to be done to advance such a mindset, Bertram said enterprises should be investing in people and skillsets, and fostering a culture of change instead of focusing on technology per se.
There needs to be a change in how people view data.
“It’s not about ripping and replacing, but about modernisation and how to develop their capabilities, moving away from descriptive and diagnostics analytics towards a predictive and prescriptive one,” he added.
Become data-driven or become extinct, companies told
Big data analytics: Companies still struggling with the basics
The emerging role of the chief data officer
For more technology news and the latest updates, follow us on Twitter, LinkedIn or Like us on Facebook.