Blog

Avoiding Common Data Science Business Mistakes

Jeff Deal

July 5, 2016

bridging the gap on data science business mistakesIn the first decade of our firm, over 90% of our analytics projects met their technical goals, but only about 65% of solutions were ultimately deployed in an operational environment1. In other words, the biggest risks of failure in data science are organizational, not technological. If your company is using analytics you need to be aware of some common businesses mistakes that frequently cause analytics projects to fall short of expectations. Awareness of these mistakes will better equip your organizational leaders to plan and guide data science engagements to successful conclusions.

Data science is one of the hottest topics in business today and the results can have big payoffs. To cite one example, a fraud-detection program we developed for a high-tech Fortune 100 company saved the client $11 million in the first year and $67 million over the first five years. With payoffs like these, it’s easy to understand why organizations that learn about and use advanced analytics techniques run ahead of their competition.

But the implementation of analytic techniques also presents substantial challenges. Many companies fail to reap the benefits because they make crucial mistakes in planning and deployment. 

Mistake #1: Failure to Clearly Define Objectives

It is surprising how often organizations jump into analytics without knowing exactly what they will do with the capability once they have it. Aware that data science is an exciting frontier and eager to apply the newest technologies, they get started without clearly defining their objectives and carefully planning their execution. That’s a big mistake!

Launching a data science initiative can be a time-consuming and expensive challenge.  Experts must be located and hired, software must be purchased, the appropriate data must be acquired or gathered, and key stakeholders must be brought on board. There are many opportunities to make expensive mistakes. Without firm objectives and a well-formulated analytics strategy, the effort is likely to fail.

Data science is a tool to solve clearly identified “points of pain.” Successful engagements address specific needs, and have the owners of those problems on board from the beginning. When the problems are solved and others within the organization get wind of these early successes, interest in analytics naturally increases. Determining in advance what is possible and what is required will help ensure the success of the analytics project.

Mistake #2: Tackling Too Much Too Fast

With an overabundance of ambition organizations strive to build an internal analytics service to create a transformational center of excellence that will produce a large and immediate return on investment. After assigning a smart, quantitatively minded manager to head up the endeavor, they make a substantial investment in a popular analytics software package, establish some broad goals that cross functional boundaries, and begin compiling analytics.

This accelerated approach almost invariable fails. Building a transformational analytics service is a major undertaking that requires extensive resources and a vast amount of organizational energy. Without a large-scale investment of resources and a corresponding shift in company culture, such an initiative can overwhelm an organization, resulting in frustration and failure.

A more prudent course is to begin with modest, well-defined projects that have a high probability of success. Quick wins generate goodwill and excitement that lead to greater institutional support.

When the United States Postal Service Office of the Inspector General (USPS OIG) approached our firm their vision was to build an organization-wide analytics service to identify fraud, improve operations, and save taxpayer dollars. But rather than trying to tackle the entire vision immediately, we initially focused on one modest challenge that promised to generate a large return on investment. Our collaborative work achieved early successes that quickly built interest and enthusiasm within the organization. In subsequent years the USPS OIG has been building toward a complete analytics service in line with their original vision and has become a high-profile success story within the federal government.

Mistake #3: Failure to Get the Support of the Data Owners

Far too often, the owners of key data within an organization are reluctant to make that data available for data mining. Database administrators, analysts, or program executives may be afraid that analysis of the data will cast them in an unfavorable light or they may think they can do the analytics job themselves and resent the intrusion of others into their area. There are many possible reasons data owners withhold data, and it only takes one to stop an analytics project dead in its tracks.

Data scientists need both timely access to data and good information about the data. They need to know how it is collected and maintained, why it is messy and/or incomplete, what each data field means, and how the data is used by the organization.

Involving all key stakeholders in a data science project from the beginning fosters a sense of shared ownership that results in greater cooperation. When the data owners participate in the formative stages, they are in a better position to provide valuable input, and they will have a stronger desire to see the project succeed.

On the other hand, bringing data owners into the effort only after key decisions have been made and the project is underway may cause them to feel diminished and be uncooperative. This is what happened with one of our clients, a moderate-size financial services firm. From the very first day, the analyst who would be providing the data was openly hostile and challenged ideas presented by our data scientists and other members of the company’s team. From a technical perspective, the project was straightforward and could have been an easy success. But after a couple of weeks of effort, it became apparent that the engagement was headed for failure and the project was cancelled. They gave up on the potential gains, having no appetite for an internal battle.

Had the data owner been on board from the beginning, he and his colleagues could have shined from the quick success that was possible. Instead, thousands of dollars were spent on outside experts, and the analytics initiative went nowhere – a costly lesson.

 Summary

Data science is a powerful tool for increasing organizational efficiency, productivity, and profitability. When a analytics engagement is properly planned and executed, it is exciting to watch the components come together and opportunities for improvement revealed. On the other hand, a poorly executed analytics project can waste considerable time and money, with attendant frustration and loss of credibility.

 ______________

[1] In our second decade Elder Research improved even further both rates -- especially implementation success.  How it was accomplished will be explained in a forthcoming blog.

Let Elder Research help you get started with analytics. Request a consultation


About the Author

Jeff Deal Jeff Deal, Vice President of Operations, leads tasks involving contracting, finances, logistics, planning, and regulatory/legal issues. Jeff has worked with dozens of clients to understand their business needs and organizational goals and, in the process, has gained insight into organizational obstacles to successful data mining engagements. His talk on the Ten Most Common Data Science Business Mistakes has been well received at analytic conferences. Mr. Deal is the Program Chair for the Predictive Analytics World – Healthcare conference, and co-author of Mining Your Own Business.