During the conclusion of this year’s annual MIT Chief Data Officer and Information Quality Symposium (MITCDOIQ) held July 12-14 in Cambridge, MA, the organizers noted that the focus of the event was at last shifting from the rise and role of the CDO to a focus on the key, strategic business initiatives undertaken by the CDO and their organization. But this shift was evident right from the start of the event – one featured session called this the process of “Shifting from Survival to Success.”
Front and center were key topics around the business value of data: building out the data infrastructure and democratizing data, the data privacy requirements of GDPR, the value and use of machine learning and advanced analytics to drive business initiatives and achieve better business outcomes, and adaption to the accelerating pace of change. One panel of CDO’s – including Christina Clark of General Electric (GE), Mark Ramsey of Glaxo Smith Kline, and Venkat Varadachary of American Express – addressed a number of these points in detail.
Managing Data for Business Value and Growth
Though the CDO role emerged out of the financial crisis and the need for broader controls around regulatory requirements and data governance, that role has changed and most new CDO’s are addressing the need for business value and revenue growth. This doesn’t mean that the foundation and fundamentals of data governance and compliance are forgotten or dropped. In fact, those aspects are critical to this shift as business growth is dependent on an organization’s ability to structure and manage data for speed and flexibility. The approaches taken to succeed may vary considerably though.
One participant noted the example of Google absorbing YouTube. Google took 2 years to build the infrastructure and information needed to effectively monetize the volumes of data acquired. In the case of financial services firms, that approach is rare, particularly with mergers. In those cases, the customer support and client relationship aspects are critical and it is more important to leave the systems in place and develop an approach to span across those systems, even where they have similar data.
The participants particularly noted that the benefits of this data-driven business approach are large, but that there is a need for a clear vision of what the organization wants to do and address. This includes establishing a scorecard with quantifiable metrics for business value.
Some of these may be as straightforward as: identifying how many different revenue-generating use cases have been deployed; or determining how much faster deployment is with new approaches to data delivery. Often, it’s the critical data elements, perhaps no more than 50 per subject area, that provide the key metrics around operational value (and the impact to business processes that will break if incorrect) tied in with the costs to acquire, manage, and consume such data.
I was on hand with colleagues at MITCDOIQ to demonstrate how our Trillium data quality software can help organization’s data governance initiatives
The CDO of Glaxo Smith Kline noted a goal to change the time to discover new drugs from 8 years to 1 year, and transform the pharmaceutical industry by leveraging sensor-based and genomics data. Many steps are needed to reach that business goal including standardizing internal data and being able to connect the internal data to new external sources.
Democratization of Data
For individuals in an organization to be effective, they need trusted data at hand to move forward with speed and efficiency. Data scientists are just a part of that equation. Sales, marketing, operations, and others in the lines of business all need data. Getting data into the hands of employees, even if imperfect, is valuable – it creates incentives sooner as people can see the data issues and work to solve them. Helping solve the problems for these people in accessing and using data not only democratizes the data, but provides them the ability to act in a more agile way with faster time-to-value.
At the same time, it is important to remember that data must be served in a manner that is consumable to these varied users. Some will want visualizations and dashboards, some need alerts and notifications for faster action, some need data in Excel, and some could care less about visualization and want access via tools such as Python.
To achieve this democratization, it’s important then to understand what data people want access to, how it may be delivered and consumed, and how individuals can accelerate this process. Shifting the cultural mindset to a process of collecting and accessing data rather than modeling and structuring the data first helps to more readily identify where the business challenges are and how data may be applied to solve the issues and drive value.
Barbara Latulippe, CDO of Dell, reiterated many of the panels themes in her MITCDOIQ presentation on “Governance and Stewardship in the Big Data Era.” She noted that the data scientists in her organization were struggling to find data. In one case, it took 35 phone calls by a data scientist to determine all the context around the data! Democratizing data means it is critical to make the data easy to find, easy to understand, and easy to determine trust and quality.
One metric for Dell is simply reducing the time needed to find and consume data for prescriptive value with a goal to move from 70% of a data scientist’s available time spent in finding data (a statistic regularly reported) to 30% of their time. Achieving this requires data governance, echoing the earlier panel’s comments that governance is foundational to success in this area. Dell’s approach follows a Lean Data Governance model, a practice that Trillium Software has noted in the past, including: starting small, showing success, visualizing results, and breaking down silos by showing others “what’s in it for them.”
Finding Data Skills, Building Data Literacy
On the final day of MITCDOIQ, Natalie Evans Harris, VP of Ecosystem Development at The Impact Lab, discussed the perceived issue in finding individuals with the data skills needed to help organizations achieve business value and growth. She noted that this is often a “signaling” problem.
The focus by organizations on finding the “data scientist” who can understand and communicate with the business while finding and accessing data, testing hypotheses, building algorithms and models, and ramping these up into ongoing executable frameworks is misguided. What organizations need to focus on is bringing teams together with the mix of skills that can empower all involved to move the organization forward. This is the approach noted by Booz, Allen, Hamilton in their Field Guide to Data Science.
It’s important to remember that the range of skills needed to work effectively with data exist in many individuals and consider whether we are really looking for specialists or trying to take advantage of competencies (e.g. biologists, linguists, etc. can provide data science) and blend those with the subject matter experts who understand the business, understand business opportunities, and can present ideas in a manner that makes sense in the organization.
My own topic at MITCDOIQ, Finding Relevance in the Big Data World, touched on an aspect of data literacy, specifically how to approach the challenge of considering what data is important, i.e. relevant, for a given business initiative. Wolf Ruzicka, the Chairman of the Board at EastBanc Technologies, noted in his blog “Grow A Data Tree Out Of The “Big Data” Swamp” of June 1, 2017. “If you don’t know what you want to get out of the data, how can you know what data you need – and what insight you’re looking for?”
A fundamental step then in bringing data into the mainstream is ensuring that the individuals working with the data to establish a goal (whether generating new revenue, meeting compliance goals such as GDPR, or reducing operational costs). Only with a business goal in mind can you test hypotheses, evaluate and measure data, and determine whether the data is fit for purpose. The results must be documented in a way that they can be communicated out through a repeatable data governance process. Such a process should start small, but it provides an approach to build a practice, show success, and build business value while democratizing and measuring the data used and highlighting which data has value for which business purpose.
As Harris noted, it’s important to address change management services and processes, particularly to understand how people can use, interpret, and understand their data and their dashboards. This means not only thinking about data literacy, but building data literacy!
From MITCDOIQ: Lessons Learned in Dell’s CDO/Data Governance Journey
As the CDO panel noted, having both data governance and data science teams together in the organization helps ensure that regulatory obligations are met while building for growth. It’s the underlying foundation needed to achieve success at data-driven initiatives. And it’s hard to get people bought in fully, and requires culture change, but that is part of the CDO’s work. This shift is evidenced in even at MITCDOIQ in its topics – no longer is the focus on creating a CDO office but on sharing the stories of organizational change and the adoption of fundamental data-driven processes and data literacy.
Discover the new rules for how data is moved, manipulated, and cleansed – download our new eBook The New Rules for Your Data Landscape today!