Algorithms Can Make Your Organization Self-Tuning

By Martin Reeves

Never let an MBA near a marketplace that can run itself.”

So said Ming Zeng, CSO of Alibaba, when I was interviewing him for my upcoming book. His words struck me as deeply provocative. As we dug into what he meant, we realized that we had both been thinking independently about what we came to call the “self-tuning enterprise.”

What Ming meant was, in essence, don’t try to manage what is better left to market mechanisms. As a young enterprise in a fast-changing environment, Alibaba had been trying not only to institutionalize change but also to bring the marketplace into the organization. Adapting to change through managed experimentation had already become a popular idea and a reality for many ’net-native companies. But Alibaba was taking it a step further by trying to continuously update not only its product offering but also the elements of the business that we might ordinarily assume are fixed: the vision, business model, organization, and information systems.

At BCG, Georg Wittenburg and I had, in parallel, been modeling the effectiveness of different approaches to strategy and execution in different environments. We carried out this modeling using a so-called multiarmed-bandit algorithm—coincidentally, the same sort of algorithm that Alibaba and other digital-market players use to recommend products to customers. These algorithms turn out to be extraordinarily good at rapidly tracking and stretching customers’ changing needs—tuning themselves into changing circumstances. They do so by operating three learning loops in a rapid, automated fashion: adapting to changing customer needs, modulating the exploration of new needs, and shaping the emergence of new needs.

So we decided to marry the two streams of thought and codify what Alibaba had been implicitly trying to do for some time: build an enterprise that could tune itself at all levels to changing circumstances. And we tried to derive a set of principles and actions that any organization could use to create a self-tuning enterprise. The basic idea was to replace elements of a company’s business system—elements that are fixed or specified in a top-down fashion—with self-directed mechanisms that continuously evolved, guided by and shaping the marketplace.

Algorithmic competition is an increasingly important aspect of business. In this case, we are not focusing on machines replacing or supplementing man. Rather, we are highlighting the importance of algorithmic thinking for updating some key business concepts like vision, organization, business model, and information systems that were framed in simpler, more stable times.

Who needs self-tuning? Any enterprise that, like Alibaba, faces a complex and dynamic environment requiring readjustment at all levels. And that’s an increasing number of not only ’net-native companies but also ones that may be facing disruption from new technologies or competitors.

This blog was originally published at  It is reposted with permission of Harvard Business Publishing.