It�s Not About the Algorithm Anymore
Five years ago we knew that data would fundamentally transform everything, from the ways businesses developed products, offered them in (often new) marketplaces, worked with their customers, and made decisions. We also knew that discovering key insights buried in vast oceans of disparate data was a necessary first step. This would take considerable experimentation and lead to sophisticated algorithms that teased out key insights. Developing those algorithms was the first order of business.
Five years ago, we also knew the companies’ organizations were “unfit for data.” For example, we knew that poor-quality data would continue to be a problem, that too much of the analytic work was a tech capability searching for a problem, and that silos would complicate the needed data sharing. But we weren’t sure how managers should adjust their thinking nor how companies should reorganize to take advantage of the new capabilities. Nor did we anticipate how quickly big data and analytics would assert themselves beyond niches such as high-frequency trading. These include:
- More effectively matching clients and the best investment opportunities for them
- Improving compliance monitoring, using artificial intelligence
- Better managing internal trade inventory, by leveraging massive computations of billions of real-time risk-based scenarios
Finally, and perhaps most importantly, we did not anticipate the speed at which sophisticated data and analytics technologies would become accessible to virtually everyone. Five years ago, small groups of data scientists were the only individuals in an organization who had any real analytic capabilities. Today, even junior people (and some managers!) conduct rich analyses as part of their daily functions as they come up, and in near real-time.
We have no doubt that new and better algorithms are still needed -- and probably always will be. The underlying processes that increase computing power and create new kinds of data for that power to work on are only now gaining steam. At the same time, today the business priority is less on the algorithm and more on building the organization to create leverage from the algorithm.
Over the last five years, we’ve learned a great deal about what is required here. First, both individuals and managers must think long-term, but act with urgency. Long-term, it is clear that those who inculcate data and analytics into their DNA will have an enormous advantage. This means that everyone must get involved. And while we don’t expect everyone to be a data scientist, we do expect everyone to think like one and, from time to time, act like one (e.g., collect simple data on their own, properly interpret basic plots, understand the difference between correlation and cause and effect).
Second, build end-to-end “data processes.” Think in terms of a D4 (Data, Discovery, Delivery, and Dollars) process. As we’ve noted, most of the work so far has focused on the algorithm (i.e, Discovery). The Delivery and Dollars steps are much weaker. Drive the effort by business problems and opportunities. Define business-oriented metrics such as “new dollars from data” and set aggressive growth targets. Work with customers to sort out what new insights are worth hard dollars versus those that are “merely nice to have.” Finally, while fully recognizing that data scientists and business analysts play separate roles, pair them up.
Third, much more work is needed on the data, specifically data quality. We can’t make this point strongly enough. The cold, brutal reality is that few decision makers will base important decisions on data they do not trust, and today, too much data is not worthy of trust. Importantly, the organizational structures and methods needed to make order-of-magnitude improvements in data quality are now freely available to all.
Fourth, take heart and show courage. Facts can disrupt the power centers of organizations as they upset “sacred cows” and challenge conventional wisdom. If you’re on the receiving end, do your best to stay calm, understand the full context of the facts, and look for opportunity.
If you are delivering the facts, expect push-back. Be sensitive and patient, fully explain limitations in your analysis, and look to build partnerships. But do not back off -- the facts are the facts, and those who deny them almost always pay stiff penalties in the end.
Lastly, get on with it. The data scientists and technologists have done their part by finding some good algorithms and making powerful technologies available to all. Companies must act -- and pretty soon, before competitors get too far ahead. Most of all, individual managers must act even faster! Get in front and show your company the way.
— Jeff McMillan is co-author of this article. Jeff is Managing Director, Investment Products and Services, at Morgan Stanley Wealth Management. He is currently leading the firm’s strategic analytics efforts, which are focused on scaling product delivery capabilities through the use of a sophisticated idea targeting and measurement framework. Before joining MSWM in 2009, Jeff was the Head of Product Strategy for Bank of America Merrill Lynch Research, where he led strategic efforts for the Research Division to include the development of research-based products and the monetization of research content.
Dr. Thomas C. Redman, "the Data Doc," is President of Navesink Consulting Group. He helps leaders and organizations craft programs to get in front on data quality, learn to compete with data, and build the capabilities to do so. His most ... View Full Bio
Tweet This