Big Data and Big Change Management: A Path Forward

During a recent executive education course, MIT professors impart change management strategies.

MIT Sloan School of Management recently held a two-day executive education course, Big Data: Making Complex Things Simpler, designed to provide executives with both an overview of big data and a few techniques to harness some of the elusive power of that data.

During the event, MIT researchers Alex “Sandy” Pentland and Erik Brynjolfsson talked about the big picture of big data and the finer aspects of utilizing data, from building trend models based on free Google data, to combining disparate data sources to create new business models.

But in all the classroom discussions that surfaced around the big issues from big data — privacy, security, costs, infrastructure, data volume, data quality, and data governance — the reality that many organizations grapple with is change management: whether or not they can manage the human and process changes necessary to make the most of their analytics initiatives.

Bottom line: the real question is whether or not they can change the culture of their organization.

Consider the experience of one course participant, who talked about how her organization, a software company with 100 million customers, is struggling to implement a new analytics program that will enable both multi-channel customer communications and upselling. To achieve these goals, the company needs to change the way it handles sales data (replace batch processing of sales information with real time processing) and communicates with customers (replace blanket email communications with more personalized, one-on-one communications).

In other words, a significant amount of human and process change management would be required to make this analytics initiative work. This, of course, is a big bottleneck.

Figure 1: Matrix of Change

Matrix of Change modelView Exhibit

Matrix of Change model by Erik Brynjolfsson

Matrix of Change model

“Internally our discussion was with IT, around what the new program is going to cost,” the student explained to the class. “We went from $3 million to $10 million. Then we hired a consultant. While there was a lack of scope planning, really what it turns out to be is very emotional. Now we’re evaluating whether this project is even worth it.”

For this problem, the researchers offered two approaches: Pentland talked about how to get people to do what you want them to do, while Brynjolfsson provided a model to evaluate the process end of change management.

To get people to do what you want them to do — in essense, to get buy-in on analytics — Pentland talked about creating a model that combines the intuitive thinking employed by HiPPOs (highest-paid person’s opinions) with quantitative reasoning from “the quants.” Bringing the two together enables organizations to explore data, visualize relationships and understand in a human, intuitive way what data is telling them.

While this is no easy task, noted Pentland, the way to accomplish a combined model is through interaction with the real world. “I will advocate this strongly,” he said. “Set up some sort of test bed — a living lab — something that stores what you have and that does things differently. Maybe a different type of service for a region.”

One example of a living lab he offered was Nike, which introduces 60 models of shoes each week. They watch to see which models take off — colors, texture, price points — and then they kill 50 of the models that aren’t working. “They are able to do a small number of custom things that are highly instrumented,” said Pentland.

Brynjolfsson’s approach to managing process change is a pen and paper exercise that takes the form of his Matrix of Change model (Figure 1: Matrix of Change). The model helps companies determine three key points: Which processes need to change, which can stay the same and how processes interact. It does this by pitting old practices against new ones to determine which are opposing and which are reinforcing.

The goal is to implement processes that reinforce one another. With enough reinforcing processes, change is feasible.

If, on the other hand, there are a lot of processes that are not reinforcing, the best approach might just be to do a radical implementation and schedule changes to happen all at once. Brynjolfsson gave the example of Sweden, which in the 1960s changed the side of the road that drivers drive on, from the left to the right. “They didn’t do this incrementally… motorcycles first, then trucks,” he said. “They bit the bullet and did a total transition.”

There is another angle to consider: whether initiating changes is going to produce results that make it worth the trouble. “Not all change management is worth doing,” Brynjolfsson told the class. “In Great Britain today, maybe changing driving direction isn’t worth it. That becomes an easier judgment to call."

1 Comment On: Big Data and Big Change Management: A Path Forward

Add a comment