What’s IT’s Role in Analytics Adoption?

Beth Holmes, IT analytics lead, Monsanto Co.

With a Ph.D. in plant breeding and genetics and years of information technology project management, Beth Holmes brings the discipline of a scientist to her role as IT analytics lead at Monsanto Co., the seed and crop protection chemical company.

She’s going to need it. In its essence, her job is to answer questions for the company through exploratory analytics — and to advance Monsanto’s organizational strategy of embedding analytics more deeply into all corners of the company’s operations. Using analytics, her group has scoped out high-value sales targets, done cost modeling, improved the accuracy of sales forecasting and used multiple methodologies to aid long-range planning. The “exploratory analytics” team routinely tests assumptions about such things as commodity prices and agricultural trends. “Understanding the possibilities of things that may happen [is] really critical to our ability to operate profitably,” Holmes says.

“Five years ago, people would have viewed analysis as almost synonymous with reporting,” she says. But seeing what analytics can do is starting to change what executives ask for. And because Monsanto’s analytics-adoption strategy leans heavily on newly created resources housed within the IT department, “the work changes the conversation that IT has with the business units,” says Holmes. “It changes the perspective of the value that IT brings.”

Holmes spoke with MIT Sloan Management Review editor-in-chief Michael S. Hopkins about myth busting, why the simplest solution is often the smartest and what it means to push for analytics adoption by using the IT function for leverage.

So how has Monsanto’s use of analytics and data changed?

Five years ago, people would have viewed analysis as almost synonymous with reporting. That work is critical, but being able to take the broader view and doing the analyses to support hypotheses that impact the broader view is what’s different.

In the end, you need the best information possible for a variety of decisions that range all the way from daily ops to strategic planning. And the level of sophistication that you need to generate that data differs based on the problem that you’re trying to solve. That said, the best choice is always the simplest choice. The simplest choice that works is always the best one.

That’s very interesting. How did you come to that conclusion?

In individual solutions, the tendency is to beat the bushes to wring out everything you can. But the efficiency of that solution may not be enough to warrant the effort. So you have to exercise judgment. We do that by always starting with a baseline that’s either the existing methodology or the simplest solution possible. We generate that first, and then everything else we do is an effort to beat it.

What has surprised you as you’ve done this work over the last couple of years?

About two years ago there was a recognition that we’re a data-rich company, but to some degree there was still some “sitting on the edge of the pool.” Our new hypothesis was that maybe you don’t need to have everything in place in order to start.

We did some work in sales forecasting that didn’t really take that much effort, but it really helped our sales people improve their ability to forecast accurately early in the season. It doesn’t take long to do an analysis, and the return on that is huge. So our hypothesis was proven out. That’s the big learning.

Operationally speaking, the amount of time that we actually spend prepping data versus doing the analysis seems surprising. For the exploratory work, maybe we spend three-fourths of our time getting the data ready and another quarter of our time selecting the analysis, running it and interpreting it. But we can complete these projects on a relatively short cycle, and the benefit that is gained in a really short time period is just phenomenal. Using analytics itself to figure out what data elements we need to go after and first making sure that those are correct can narrow that field enough that it’s something that people can start to get their heads around and plan for.

Do you think this kind of burgeoning, centralized analytics capability should be housed in IT?

Yes, I do. The reality is that IT has always been a great place to be for perspective into the different business domains. When you underscore that with the implementation capability that being in an IT organization can bring, it is a natural fit.

There are really some great analytical minds at this company. They network together informally, and we have both formal and informal relationships with a variety of people doing analysis all over the company. I would say that building analytics capability is fixed in IT. That’s the one place where it’s formalized. But having a centralized analytics capability doesn’t mean it’s diminished at the point of need. They can kind of augment each other. In fact, we actually accelerate those efforts.

How has the IT department’s relationship to the company changed?

We’ve always had a really strong research and development effort, but we’re seeing an increasing appetite to use analytical methods in every level of our business. We’ve essentially found our way out of R&D and into the business units. It’s exciting to me to see that, because as business units recognize how they impact one another, there’s a feeling that they need to get more out of their data not just for their own use but for the interaction between groups. That’s how a group like my team really helps: We can step across and not only help bridge those silos, but also bring new techniques forward and help bring those folks who are already used to doing analysis up on those techniques. Because obviously it’s important that you understand what an analysis is doing for you and to you, in order to use the data.

This work changes the conversation that IT has with the business units that it works with. In some cases it changes the perspective of the value that IT brings. We’re used to being sought after for technical solutions of a different type — for the software, hardware and service that we provide to support the business units. It’s different to be sought after for the answers that we can provide.

You’ve said that analytics is about the quest for the best set of leading indicators that you can get to inform your business. Have those indicators changed over the last few years?

Yes, and actually, that’s part of the evolution of using analytical capability. If you’re running models, then of course you have to not only look at the results but also at the variables that go into that: what the relationship is between them, is that relationship behaving as I expect. Those relationships change over time, and it’s really the ability to detect that change that provides companies with a competitive edge.

We make sure that people are educated on the relationships, and assumptions about relationships, that underpin the models — and that we have the capability to monitor those relationships as well as generate the model results.

Do you find that conventional wisdom in the industry is wrong once you bring models to bear on it?

It’s definitely very common to do a little myth busting, but it’s also common to confirm conventional wisdom. Both things occur on a regular basis. But the bigger point is that even when you either dispel a myth or you confirm conventional wisdom, you have to keep in mind that tomorrow that relationship may change.

In your view, how is the changing nature and availability of information and the changing nature of analyzing it going to alter how companies work?

The expectation of having analytical results available to make a decision early on is increasing. As higher quality information becomes available faster, we need to be poised to make decisions faster and take action as well. And I think that means that you have to have some different skill sets available to either design or execute analyses.

You can acquire capability in terms of the computational power that’s required to generate some analyses. But you also have to invest in the talent within your organization that really understands what the analyses mean and can think about things objectively. The value of having a group in IT is that they can step back from the individual business units and think about the problems and solutions objectively.

We’ve seen that sometimes people in positions of power don’t want to cede their instincts to analytically driven insights. Have you encountered that?

It’s interesting. You see a variety of responses to the group, everything ranging from, “Wow, this is awesome, we needed a group that we could turn to for help with these things,” to some self-consciousness and concern about the analytical effort being an evaluation methodology.

We typically get past that by trying to collaborate deeply and closely with the business unit that we’re working on a solution for. They learn as they work through these solutions with us, and that improvement in their knowledge base translates out into their organization in a kind of ripple effect. In the end, we don’t want to own the model, we want that group to own the model. What starts off as fear can become pride of ownership and the ability to work with the answers.

Do you think leaders are going to need to be different, or differently capable, in the age of the new intelligent enterprise?

Leveraging data and analytical capability is an expression of resourcefulness, and it’s something that you can’t afford to fall behind on. If don’t like math, then you may try to avoid that to some degree, but you’re going to be impacted. The best leaders are inquisitive and resourceful in their use of data and the talent to analyze the data, but they’re also decisive and willing to take action based on data. In the transition from intuition-based decisions to data-driven decisions, that’s really the huge difference that this capability brings: the confidence to make those decisions. I think the willingness to look at it from that perspective will help leaders be effective.

What impediments do you expect to encounter as these changes become more widespread?

Sometimes we see some hesitance, based on a preconceived notion that people won’t understand how you got the answer. We really try to help familiarize people so that they don’t walk away from a solution because they’re afraid they won’t understand how it got there. It’s important for folks to realize that while sometimes it requires highly skilled talent to combine analytical techniques to arrive at an answer, everyone — absolutely everyone — is qualified to ask questions. Once people start doing that, they start to lose the discomfort.