If you think that buying into big data just means developing some analytics to solve a problem, think again. To create truly meaningful big data solutions, you need to dig into, and sometimes transform, the processes, operations, people, and culture on which your organization runs.
To explore this broader view of big data, GE’s recent Data Forecast event in San Francisco brought together the following panelists for a discussion focused on the changing culture of big data:
- Matt Denesuk - Chief Data Officer, GE Software
- Michael Jordan - Professor of Electrical Engineering / Computer Science, UC Berkeley
- Jaya Kolhatkar - Vice President, Engineering, @WalmartLabs
- Santhosh Nair - Vice President, Strategy & Business Development, Wind River Systems
Below are five highlights drawn from the evening:
#1 Resources and culture shift required
Nair referred to an American Management Association report stating the two biggest roadblocks to organizations embracing data culture are insufficient resources and unsupportive company culture. Data-driven decision making is new to companies that have traditionally made decisions based on intuition and experience. Accepting a fact-based approach requires a culture shift throughout every level of an organization.
#2 Regulated = reticent industries
Unregulated industries, such as retail and high tech, are more open to embracing data-driven culture. Regulated industries like those in the aerospace, medical, defense, and energy sectors tend to behave in a cautious way, tailored to regulation. Even if regulations themselves don’t necessarily prevent fostering data culture, regulated organizations have inertia when it comes to taking risks due to the high cost of mistakes and learning opportunities in their domains.
#3 Don't just collect it...
Even though there is a general, industry-wide acceptance of the value of big data, many companies don’t know how to monetize it. Kolhatkar pointed out that even in the retail space, getting results out of data takes time and patience. A big part of it is asking the right questions. What problems are you looking for data to solve? What kind of data do you need to collect in order to solve them? What’s meaningful? What are the areas where data can bring the highest value? Businesses need to do more than just accumulate data. They need to invest in it. Nair emphasized the role that organizations like the Industrial Internet Consortium could play in cross-pollinating the best practices of fostering data culture among industries.
#4 Losing touch with context is dangerous
Jordan warned about falling into a dangerously narrow data mindset. Data is used to solve problems of inference. Data structures themselves are not the end goal, yet sometimes data scientists make this mistake. It is absolutely vital for data culture to embrace and understand statistics, error bars, sampling, in order to reasonably apply the conclusions of a small set of data to an overall population. Otherwise, catastrophic mistakes could result simply by “following the data.” Denesuk noted how in GE, data is usually vetted through experts who understand “physics-based” models of processes, in order to make sure data jives with reality. Kolhatkar underscored how critical it is for executives to instill discipline in data science organizations by asking probing questions and continuously monitoring results to create a more effective structure for data science initiatives.
#5 Data scientists evolving into artisans
The role of the data scientist is still evolving. Jordan made the analogy to early bridge builders, who needed to understand not only physics, but also materials, architecture, and engineering in order to build successful bridges. Data scientists are currently taking more of an artisanal approach, operating at the intersection of computer science, statistics, and domain expertise. The exact level to which they specialize or take an end-to-end approach is still developing.
Watch the complete discussion