What's The Mindset For Innovation With Data Science Function?

By Kimberly Cook |Email | Dec 4, 2018 | 7005 Views

Principles to nurture a healthy and innovative Data Science function

As organizations turn to digital transformation strategies, they are also increasingly forming teams around the practice of Data Science. Currently, the main challenge for many CIOs, CDOs, and other Chief Data Scientists consist in positioning the Data Science function precisely where an organization needs it to improve its present and future activities. This implies embedding Data Science teams to fully engage with the business and adapting the operational backbone of the company (e.g. techniques, processes, infrastructures, culture, legal).

There is no formal description of Data Science. Its mission to "understand and analyze actual phenomena with data" can vary greatly from the industry to academia. With experiences in both worlds, my definition of Data Science is biased toward a combination of skills in mathematics, programming, and communication with the application of the scientific method to specific domains of knowledge. I like to summarize the practice into four different strategic areas:

  • Datasets typically cover data governance, strategic data sources, and infrastructures.
  • Skillsets are about measuring analytics readiness, managing talent, spreading an evidence-based culture (e.g. creating a shared language), applying Data Science processes, and designing the organizational framework of Data Science teams.
  • Toolsets cover the selection of the proper Data Science tools and the application of the best practices throughout the organization.
  • Mindset that assembles the animating principles that support the ethos of a Data Science function to deliver value and innovate at the source of a digital transformation.

I believe the Mindset is the moving force that transforms the investments in Datasets, Skillsets, and Toolsets into economic and cultural impact. Let me try to introduce why and how.

For more than 10 years now, I have had the privilege to collaborate in dream teams that bring Data Science to the avant-garde of their domains. At MIT SENSEable City Lab (MIT) we pioneered in techniques that analyze digital traces of human activity for urban innovation, Bestiario develops unique environments that democratize data manipulation and visualization and BBVA Data & Analytics (D&A) derives knowledge from financial data to transform the banking industry (e.g. risk modeling, customer advisory, process automation). I consider these institutions exemplar for their talented professionals, their open-ended challenges, and their access to large quantities of data. Each has a different structure, finance mechanism, and function, but all share a mission to spearhead the digital strategy of their sponsors or clients.

What I experienced persistently at MIT, Bestiario and D&A is a distinctive capacity to deliver tangible results, think differently and break the status quo. I experienced a mindset hardwired for complexity and unpredictability shaped with a leadership that sets the context and inspires people more than manages them.

Based on my observations and learnings, I have been developing a set of principles that I use as a blueprint for how to nurture the mindset for innovation with Data Science. These principles are not ranked in order of importance and are designed to complement one another. I value:

1. Culture over technology
2. Polymathy over expertise
3. Ambidexterity over disruptiveness
4. Stories over sprints
5. Ethics over profits
While there is value in the items on the right, I value the items on the left more. Of course, these principles will ebb and flow to varying degrees and should not be considered universal rules.

1. Culture over technology
At the very heart of the ethos at MIT, D&A and Bestiario is a commitment to always be learning, to seek peer recognition and to help others do the same. For instance, a good part of the D&A team dedicates time outside of working hours to teach. This culture provides a return of investment on knowledge acquisition and knowledge sharing across communities of practice and also in contacts, collaborations, branding that increase the levels of commitment, cohesion, devotion, and motivation. In other words, Data Science exists beyond techniques.

In a world of technological push, 'techno-prophets' and media echo chambers, it is easy to get sucked into the technology trap with its sense of urgency to constantly explore new tools that only lead to distractions and short-term cosmetic results. Similar to other creative teams, it is not the technology that attracts or retain talent in Data Science. What drives people is an open and healthy culture that makes sure people feel fulfilled, challenged and supported in their jobs.

2. Polymathy over expertise
What fascinated me most from my experience at MIT is that its research activity was not bound by the methodologies of a single field. It was characterized by an 'omni-disciplinary' approach - inspired by MIT Media Lab - that spoke the language of designers, engineers, physicists, biologists, social scientists, and even artists. Similarly, Bestiario's CEO Jose Aguirre sets the context for engineers to develop scientific muscles and think like designers.

Data Science encompasses a diverse set of disciplines and does not work in isolation. In order to collaborate with others, specialists need to learn beyond their domain of expertise. Some organizations might fall into the trap that a specialized Ph.D. is necessary to do Data Science or that it is a discipline of engineering. In fact, the practice requests a developed sense of curiosity to understand the language of other disciplines and a strong appetite for collaborative learning. These characteristics of a 'polymath' or 'generalized specialist' prepare scientific teams to engage with people with different roles in an organization from product management, design, marketing, legal, communication, engineering, finance, and more. Recently, new hybrid names like 'Product Scientist' or 'AI Designer' have emerged in the industry highlighting the need to connect various disciplines with the scientific method. In parallel, views from interdisciplinary domains like Science and Technology Studies (STS) on algorithmic accountability, data bias or 'thick data' are becoming part of the core themes of Data Science.

A generalized specialist brings the advantage of interdisciplinary knowledge, which fosters creativity and a firmer understanding of what society, an organization or a business needs. A team of generalized specialists brings a better overall perspective for deep, complex and unconventional areas than a team of experts can.

3. Ambidexterity over disruptiveness
At the moment of positioning their Data Science function, some organizations associate the novelty of the practice with their innovation agenda. They expect new value creation with disruptive concepts. Unfortunately, this proactive approach often results in some levels of disconnection from the organization realities with teams exploring something that could not be an actual problem or investigating the wrong questions. Consequently, the measurable impact of this mode of exploration risks to be only residual. It is a potential shortcoming of the sponsorship model at MIT or the practice of hiring external consultants.

An organization is only as strong as its teams' ability to collaborate with one another. Consequently, other organizations bet on the co-location of departments for identifying the right problem - and to prioritize problems. These teams generally focus on the application of Data Science to current optimization needs. When well-executed, the delivered value generates sufficient impact to accelerate the assimilation of Data Science as part of the core business of an organization. However, Data Science teams with their nose to the grindstone do not fully benefit the organization. Busy with small improvements, short-term requests, meetings, and reporting, the brains and creativity of people are not used to question the status quo and think about alternative opportunities.

In Data Science and other domains, there is a trade-off in balancing a strategy between innovation and optimization. One of the main success factors for D&A is that it grew as an ambidextrous organization. On one hand, we had a strategy that creates efficiencies and incremental changes for the present of BBVA as a global financial institution. Besides showcasing the short-term value of Data Science, the projects provided the best context for teams to grasp the real issues in the financial industry. On the other hand, we were also convinced that the companies that want to thrive in the 21st century need to use Data Science to explore their future competitive advantages. That line of work was about value creation and pushing the boundaries of an innovation agenda. Similar to MIT or Bestiario, the mandate was to help an organization to think and execute 'out of the box', away from the day-to-day interruptions and distractions. These explorations are rare opportunities to reflect on the knowledge gained, cultivate doubts, and articulate what's next.

4. Stories over sprints
In their early stages, projects need to secure buy-in from leadership, to mollify anti-change agents and counter unrealistic expectations. At MIT we employed what SENSEable City Lab's director Carlo Ratti calls "urban demos" to translate ideas into experiments and visions. These demos do not showcase any results but rather set the stage for an upcoming exploration. They tell a tangible story about "what could be" for sponsors and people to take a critical approach.

I remember WikiCity Rome as an urban demo that particularly impacted the imaginary of an emerging 'smart city' industry. The team produced the public screening of an animated map overlaying in real-time the different types of mobility (pedestrians, taxis, public transports) in the city. This project, done in 2007, was one of the first urban experiments in which a live audience could experience feedback loop mechanisms with digital information.

At that stage, the demo did not need to build a business case; its story sought to get enough sponsor buy-in to get resources to explore more concepts and report our findings back. Similarly, at D&A we developed data stories and we designed fictions to explain in a playful way the applications and implications of our analytical capacities. Shared in forms of visualizations, videos or hacked advertisements, the content helped disseminate insights that are simultaneously rigorous and speculative. These stories give sponsors or clients a clearer point of view about potential futures and how their organization will use innovation to respond. Unsurprisingly, Bestiario also has in their DNA an ability to build narratives about complex systems through exhibits mixed with the capacity to deliver enterprise data-driven solutions.

With the widespread of Agile or similar methodologies, many companies are becoming great sprinters. But innovation takes time and patience. Stories are a way to build a sustained focus over time and develop a 'patience capital' for teams to execute beyond quarterly reporting. Fictions, visualizations, demos or other types of stories empower Data Science teams to pause, question and focus more on learnings and visions than on sprint planning. Narratives offer an opportunity to thoughtfully consider a business through interactions and systems dynamics, and how the evolution of analytical capabilities could impact an organization. The material they generate also act as connectors to share knowledge across disciplines and outside of the organization. They disseminate an open, positive and creative spirit.

5. Ethics over profits
When I was completing my Ph.D. at the crossroad of Ubiquitous Computing and Human-Computer Interaction, one research team would recurrently stand out for the quality and boldness of their investigations. The Persuasive Tech Lab at Stanford was designing - and still does today - digital technologies to change what people believe and their behaviors. Its research was typically directed to individuals who wanted to quit smoking or people who needed to stick to a specific diet. After their graduation, some of the lab students naturally moved to next door campuses at Google and Facebook. There, they applied their habit-forming algorithms in the attention economy, a business that exploits human psychology to keep social media users active as long and frequently as possible. The problem in that story is not the development of persuasion techniques but rather the lack of guidelines on how the application of Data Science can change people behavior in an ethical way.

As the practice of Data Science is still nascent with wide potentials, teams must grow a moral compass and techniques to foresee the limits of their discoveries. It is also their leadership obligation to help figure out the social mission of the developed analytics and algorithms. It takes a common effort to guarantee the application of Data Science with the (unintended) implications in mind. Neglecting that aspect is similar to mandating a team to behave like Wernher von Braun as described in Tom Lehrer's song:

'Once the rockets are up, who cares where they come down? That's not my department."
At MIT, the urban demos offered a way to explore whether our concepts were culturally or socially acceptable before whether they were potentially profitable. Similarly, D&A has lines of work in the social implications of Data Science as we established internal guidelines on Responsible Data Usage and developed an active public voice on the relation of Machine Learning with trust, interpretability, and fairness. Leading initiatives for social good that put people and the planet before profit, we discovered we could enrich our analytical capacities with partners that may not have engaged in commercial ventures (e.g., governments, UN, non-profits), and also attract top talent.

The Mindset of Scientific Leadership
Today's reality is that most organizations still do not have a method for measuring innovation success. I believe a well-nurtured mindset can help a scientific leadership team to compensate that shortcoming and move Data Science into their core strategy and culture. For instance, the principles I have exposed can act as ingredients to:

  • Chart out visions to anticipate what's next, to secure buy-in from leadership, to mollify anti-change agents and to counter unrealistic expectations.
  • Formally communicate success with metrics and stories on the economic impact, change, and innovation value that Data Science activities generate.
  • Maintain close ties and coordination with sponsors, clients or the parent organization guaranteeing a balanced equation between innovation and optimization strategies.
  • Hire well with an open and diverse culture that makes sure people feel fulfilled, challenged and supported in their jobs.

People in the Data Science function also need scientific leadership that talks their language. Over the years, I have learned to set my priorities to promote my version of teams set for innovation:

  • Provide the context to work with craft, curiosity, empathy, and efficacy.
  • Inspire with 'why' to cultivate doubt, motivate brilliant people to surprise themselves and request teams to show as many implications of their developed capacities as applications.
  • Empower with 'why not' to share responsibilities and take unconventional paths in a flexible working environment.
  • Don't be (too) busy to take time for you and your teams to think away from the action.

This philosophy emerged from my professional experiences. Some parts of it could apply to other high-performance teams in other disciplines. I am happy to talk and share more of it, and I am equally interested in hearing your thoughts and stories. Feel free to comment or contact me if you think about other principles or different styles to nurture an innovative mindset for Data Science.

Source: HOB