How to Build a Scalable Data Science Strategy

September 12, 2017 Mash Zahid

It’s been several years since the term “big data” crossed over into the mainstream business world—and executives realized they could make smarter decisions by harnessing the reams of information generated across the enterprise. Early efforts at data analysis have borne fruit, yet the prospect of a comprehensive and truly scalable data science strategy remains elusive at most companies. Growing interest in techniques like machine learning, deep learning, and artificial intelligence have only added to the confusion.

Fortunately, building scalable data science strategies is far from impossible. In fact, it just takes the right people… and a willingness to experiment.

The case for evidence-based decisions

Historically, companies used Six Sigma methodologies to analyze and improve business processes. Data science—which combines a practical business or operational understanding with large amounts of relevant historical and real-time datasets, then applies algorithms to learn and predict outcomes—powers faster and more accurate improvements.

Data scientists take advantage of cutting-edge computational capabilities to automate their business and operations workflows. People don’t have the capacity to sit and sift through large amounts of data at the same speed that a computer does. The outcome of this process is valuable information that guides a smarter, more informed decision-making process.

Enter machine learning

Machine Learning introduces a potent new tool by applying some statistical learning precepts as well as information technology tools (e.g., Microsoft’s Cortana Intelligence and AzureML offerings) to derive predictions. These predictions are then operationalized in order to improve business outcomes. Deep learning, a richer machine learning technique that uses more sophisticated algorithms, holds even greater promise.

Operationalizing any prediction can be difficult, because it involves asking people to abandon methods and processes they may feel are already working. Imagine, for example, trying to convince a high-potential retail outlet manager to alter racetrack and end-caps ahead of Super Bowl weekend, because your data shows that customers who stock up on fresh fruits and vegetables during those three days also check out with a much bigger shopping cart.

Iteration is the key to refining your data science strategies and driving results across the enterprise.

The other challenge? Accepting that your team has to iterate on its machine learning experiments—that your data scientists and business operators have to collaborate to refine predictive models and fine-tune business processes, thereby closing the gap between predicted and actual outcomes. Without this back-and-forth, you won’t get the best results (or any results, in some cases). It may come as a hard lesson after you’ve invested in a pilot program that didn’t deliver an instantaneous impact. But machine learning technologies are probabilistic in nature, and iterating is the key to driving value.

Building scalable data science strategies

So what does it take to do this at scale across a business? Chances are, you’ve already got talent who are capable of thinking in a data-driven way—who are adept at fields like engineering and statistics and can communicate and coordinate across multiple disciplines and departmental areas. Here’s how you can put them to work.

  1. Identify and gather your Data Science Ninjas: a diverse group of self-selected people with analytic skills and the motivation to spend a portion of their time working on business transformation projects.
  2. Give them full access to all data in a secure sandbox environment and allow them to experiment. They might explore relevant business drivers and conditions, testing hypotheses to find solutions that can be influenced, operationalized, and/or automated to drive ROI. To move even more quickly in a start-up environment, you can hire an experienced advisor to help build the program and formalize structures, goals, and processes, then leave it to your internal team to manage on an ongoing basis.
  3. Encourage them not just to experiment but to iterate as they discover and refine their solutions. Iteration is how the program becomes scalable: by enabling new data ninjas to pick up with new experiments as others drop off and return to their regular full-time roles.
  4. Make sure the organization can align quickly and in an agile manner with the discoveries of these iterative, cross-functional experiments. This starts with active, patient support from highest levels of the organization—and a commitment to using evidence-based findings to transform the enterprise. Program sponsors should also get their hands dirty, so to speak, by getting more deeply engaged with projects that cross multiple knowledge domains. Even better if they also help propose and evaluate business hypotheses for the ninjas to investigate.
  5. “Rinse and repeat.” In fact, as with successful Six Sigma operations improvement programs at major corporations, you should encourage participation as a data ninja as a way to get on the fast track for career growth.

Driving large-scale transformations in a large company is rarely straightforward. But engaged executive sponsorship and patient capital will help you stay the course during the rapid experiments that are required to build scalable data science strategies.

And when you see your work start to pay off through demonstrable business improvements—and your momentum starts to accelerate—that’s when the fun really starts.

Need a data science strategy? We can help!

Get the Skills You Need

Thousands of independent consultants, subject matter experts, project managers, and interim executives are ready to help address your biggest business opportunities.

Get Started

About the Author

Mash Zahid

Mash Zahid is a data science and machine learning expert who works across industries to improve business outcomes in inventory optimization, product development, trading and valuation, and more. After beginning his career at Arthur Andersen, he worked at J.P. Morgan Chase and The Home Depot, where his data-driven approach for improving retail operations and people strategies was detailed in Harvard Business Review. He's helped BTG clients improve patient adherence and support clinical decisions.

More Content by Mash Zahid
Previous Article
The End of Loyalty: A Q&A With Author Rick Wartzman
The End of Loyalty: A Q&A With Author Rick Wartzman

What's next for American workers now that companies no longer offer stable jobs? We spoke with author Rick ...

Next Article
On-Demand I.Q. – Winning the War for Talent in the Gig Economy [eBook]
On-Demand I.Q. – Winning the War for Talent in the Gig Economy [eBook]

The gig economy isn’t just for ridesharing platforms. Now, many of the world’s best business minds are work...

Get our free guide to working with on-demand talent.

Read Now