A featured contribution from Leadership Perspectives: a curated forum reserved for leaders nominated by our subscribers and vetted by the Construction Tech Review Advisory Board.

Direct Line Group

The Right Conditions for Data Science to Thrive

Stan Sthanunathan’s excellent article in the April edition of CIO Applications provides a good overview of the types of data science solutions and their practical applications. To build on this, while technology is a huge enabler, in my experience, there are a number of factors that really allow data science to thrive as a business asset. This is especially true in more established, larger organizations that have legacy systems and culture, whereas newer, start-up and “FinTech” type organizations may have the advantage of the right culture from the outset. Nevertheless, they can still fall into the same traps as they grow and reach scale, so all organizations should bear these factors in mind.

Strategy

Yes, yes, culture eats strategy for breakfast (more on that later), and I am not suggesting call in the Management Consultants (as an ex-Management Consultant myself), but, to quote Simon Sinek, it does all start with why. Why does the business exist? For what purpose for which customers? And how does your business strategy and business model link to your: 1) data strategy (what is the state of my data and data systems, and data skills?); 2) analytics strategy (what insights can I gather, visualise, automate and predict?); and 3) data science (AI) strategy (what business process can I automate, how can I personalise customer experience etc?).

Culture

So, now that we see data as central to our business strategy, what allows us to find the pearls in the data oysters? Data science teams need to have a clear understanding of the commercial imperatives of the business, for example, ways to increase revenue, or how to improve customer experience. Once this is understood and potential initiatives to achieve these goals (like optimizing digital customer journeys), data science teams are best left with the autonomy to explore the data and suggest solutions. This does not mean it’s R&D and results are slow (see section below). This does mean a “fail fast” mindset, where things are tried, value is found, or if not, move on quickly to the next solution. Data science teams who combine autonomy with commercial awareness do well.

Data science is harder, and AI harder still when data and technology is not enabling


Similarly, politics is an impediment to progressing data science. Data science initiatives work well where functional silos are removed, and cross-functional working encouraged, sometimes changing traditional organization structures and behaviors. Also, politics that relate to who “owns” technology and data can hinder progression. Data science teams should be equipped with the right tools and technology (especially where open source can drive innovation) and access to data. So long as data science teams are aware of organizational data governance and privacy, they don’t need the data police on the patrol.

Methods

An observation (criticism?) of data science teams is lack visibility on progress. Quoting a colleague, ”data is hard.” Therefore, data science is harder, and AI harder still when data and technology is not enabling. To mitigate this, data science projects should be iterative and agile to ensure a regular feedback loop and business input to get to the best solution. This provides rigour and visibility. Patience also helps. In presentations to business teams, I often hear “we already knew that!”. Good! It shows the data is right, now help your data scientists turn your data into gold.

Sponsorship

This is largely an extension of strategy. If this is in place and data science is valued to deliver on big commercial challenges, then executive sponsorship should be in place. You are not serious about data science and the potential of AI if executive-level personnel are not engaged in data science. Does this not encourage vanity projects? So long as it is tied to your business strategy and the data science projects have cross-functional representation, then it will not be a vanity project.

Investment

In mature data and digital-led businesses, data is an asset, to invest in and curate data products for commercial value, not a cost to the business. That also means, that there should be an executive owner for the data strategy and end-to-end delivery, with the freedom to invest based on a business case. As such, data science should be equal, if not more important, to other data functions like reporting, business intelligence, and analytics. This is where awareness and education are important across all parts of the organization and data literacy is now a burgeoning ethos.

Talent

Finally, people. True there is a war on talent in data science right now, with talent shortages and eye-watering salaries being offered. I certainly think that an experienced and business savvy data scientist (or two) is required to kick-start your machine learning projects. However, look around the organization for existing talent: analysts with a mathematics and statistics background. Through coaching from experienced data scientists, these people can quickly learn and develop into fully-fledged data scientists, saving on recruitment costs and silly salaries.

I reiterate, that they will only succeed with the right tools, in an environment that encourages freedom to explore the data universe. None of this is easy, but patience, positivity, and passion from all levels will reap the rewards your business seeks from data science.

The articles from these contributors are based on their personal expertise and viewpoints, and do not necessarily reflect the opinions of their employers or affiliated organizations.

Weekly Brief