The importance of data analytics in business today is well documented. It is no longer just a buzzword, it is an accepted fact of life, and those failing to use the advanced data analytics software and methodologies that generate data insights are almost certain to lose ground on any competitors who are. However, many companies still make fatal errors when attempting to put together a data project. Indeed, Gartner predicts that in 2017, 60% of them will fail, languishing in the piloting and experimentation phases before being cast aside. We sat down with Aki Matsushima, Lead Data Scientist at Direct Line, Charlie Ewen, CIO at the Met Office, and Duncan Bain, Senior Data Scientist at Scottish Power to discuss what they felt were the key factors to ensuring this doesn’t happen.
Understand What Data You Need
The first thing you need to do is understand what it is you want from your project and what data will actually be relevant. There is a temptation to go hell for leather collecting all data possible, but this is wasteful in terms of resources and time. It is also likely that by unleashing this tidal wave of data, you will be unable to gain any real insight because there will be too much noise drowning it out, rendering analysis impossible.
Charlie Ewen argues that ‘The biggest mistake that companies can make is the logic that more data equals more insight. There is a balance to be struck between simply acquiring as much data as possible with understanding the data that you have in terms of quality, veracity, accuracy and so on. Whilst there is certainly validity in 'Big Data' machine learning techniques, even they are more effective when data can be characterized in the kinds of dimensions mentioned above.’
Get Senior Buy-In
There is still some debate around whether a top-down approach to a data project is better than bottom-up. Ultimately, though, without C-suite buy-in, it is nigh on impossible to get anything off the ground. The C-suite is essential for getting the necessary tools and staff in place, and they are also best placed to emphasize the importance of using data as the bedrock of decision making throughout the entire company. Duncan Bain explains that, ’Understanding simple, high volume processes is a good way to quickly deliver value based on data analytics. Demonstrating value brings people to the table at all levels of the organization.’
Getting C-suite buy-in is no easy task, however. In a recent survey of 2,165 data professionals commissioned by KPMG and conducted by Forrester Consulting, 49% of respondents said their C-level executives don't fully support their organizations' data and analytics strategies. Bain advises that ‘Being able to visualize complex processes in a way that simplifies understanding and facilitates action is key to C-level sponsors who often have limited time to devote to any individual project.’ Data visualization presents data as a story, providing a narrative that draws an impactful response from the user and reinforces it with numerical evidence. In this way, senior management will have a better understanding of what the data is revealing and how powerful it is.
Charlie Ewen, meanwhile, believes that taking a more incremental approach to demonstrate value is important. He notes that ’there is a degree to which belief needs proof. It is tempting to try to apply data-centric approaches to large business problems in order to demonstrate the principal. Without long-term committed sponsorship, this is unlikely to work as resources and commitment from senior stakeholders will not be significant enough, for long enough, to show results. A better strategy is often to take on a meaningful but constrained pilot and demonstrate a small benefit for which there is a scalable methodology.’ Aki Matsushima agrees, adding that, ’Organizations looking to take a data-centric approach shouldn't boil the ocean, and focus on delivering value. They should start with a concrete use-case that's high ease and value, i.e. the low-hanging fruit, to actually experience delivery of a data-centric project, then build on its success by capitalizing on the buy-in and skills built from the experience.’
Instill A Data-Centric Approach
A data-centric approach across the organization is not to be underestimated. It is a central plank in exploiting data and analytics to its full, as noted in a recent study by MIT Sloan Management Review and SAS ‘The Analytics Mandate’, which reported that an ‘analytics culture’ is the driving factor in achieving competitive advantage from data.
Hiring the right people is obviously important, and a highly skilled data scientist will be a tremendous part of ensuring the success of a data project. Unfortunately, they are in short supply. Even more important in leveraging your data to its full potential is creating an analytics culture that aligns the entire staff behind its data goals. This requires data democratization, providing every employee throughout the company with access to the data. Duncan notes that ‘Helping process owners understand, interpret and work with the data that their processes use and generate is going to be a core requirement for the future success of all businesses. This is especially true for large organizations like ours which both generate and consume huge quantities of data.’ By empowering everyone at every layer of the company with the right information at the right time, employees can make better decisions. They can also identify patterns of customer behavior and potential procedural improvements that only they may be in the position to see. It is also now easier than ever, with Aki saying, ‘The combination of open source technologies, cloud computing, and MOOCs means that everyone can become a Data Scientist. In the last few years, there has been an explosion in the volume of good quality resources available in all of these areas.’
She does, however, urges caution, continuing, ‘It's tempting to make ‘building a single version of the truth that is accessible by everybody’ a goal in itself, but is this a realistic, tangible goal? In a well-established organization which has gone through a history of acquisitions and divestments like Direct Line Group, it's a higher priority to do enough to democratize data as a means to a particular use case, but keeping in mind a blueprint for a healthy, working end-state.’
Charlie Ewen is similarly aware of the challenge companies face. He argues that ‘The Met Office sees this (data democratization) as an important priority both inside and outside of the organization. The challenges in doing so are significant and involve making data discoverable, available and useful. Whilst fully supporting Open Data programs (the Met Office releases more open data than anyone else in Government), this is only part of the challenge. At extreme scale, making data available and useful can be difficult and expensive - especially when timeliness (velocity) is combined with scale (verbosity) and complexity (variety).’
Manage Your Expectations
Gartner research has found that over half of analytics projects fail either because they aren’t completed within budget or on schedule or because they fail to deliver on over-optimistic features and benefits that were agreed on at their outset. It is important to be realistic in what you expect from your data project and to know that there will be challenges and setbacks in its implementation. If your data initiative does not immediately provide ROI, it’s not the end of the world, and certainly shouldn’t be abandoned.
We sat down with Aki, Charlie, and Duncan ahead of their presentation at the upcoming the upcoming Predictive Analytics Summit, taking place in London on March 30-31, where they will be joined by a number of other industry-leading experts in the field from companies leading in their data efforts.