Building a Better Car Company With Analytics

When Gahl Berkooz defended his applied math doctoral thesis at Cornell, the possibility that he’d wind up in corporate leadership for a Big Three automaker simply wasn’t on his radar. But in the 20-plus years since then, a lot has changed in the business world — and using data analytics to understand the complexities of modern business has become not only common, but essential. Berkooz joined Ford Motor Co. in 2004, and was head of data and governance and a member of the company’s global data insights and analytics skill team at the time this interview was conducted. (He is currently chief of analytics, Global Connected Consumer Experience, with General Motors.) Berkooz became acutely aware of how important analytics is to the company’s ability to thrive in the global marketplace. He spoke about his experiences at Ford with MIT SMR’s Michael Fitzgerald.

You’ve built the information management and analytics function at Ford for over 10 years now. So I’m curious, what has Ford learned in a decade of data and analytics?

I think that one of the key lessons that we learned is that you really can’t separate data from analytics — that your analytics, at the end of the day, are going to be only as good as the data that goes into it. That determines the quality of the model results, but also required for your analytics productivity.

Data scientists or analytics experts spend most of their time collecting, normalizing, and assembling the data, and part of what we’ve learned is that you really want to — “curate” is the term we use — the data ahead of time so that the analytics process is efficient.

Another key learning we had is to have appropriate scoping of the business problem ahead of time and a structured process that involves illuminating both the data and analytics requirements. You may have an idea about some clever algorithm or clever method, but once you try to apply it in reality, you find that to make an impact, it invariably requires massive amounts of data.

Are there problems that Ford finds data and analytics work well for and problems for which it doesn’t work so well?

There’s so many opportunities today for data and analytics to have an impact that it’s pretty rare that the teams get called into situations where data and analytics methods can’t really help. It’s really more about trying to get those extra percents of efficiency and leveraging the opportunity to turn insight into an advantage.

That’s really what we’re talking about — a few percentage points on the margin that an analytical method will do better than people. If you add that up across the enterprise, it’s huge. So, what it boils down to is that we know how to make decisions. It’s more about finding the opportunities to bring data and analytics to make better decisions.

Can you give me an example of an early project where data and analytics was applied that opened some eyes?

We can go back to 2005 when the company started to look at globalizing operations in earnest. Ford historically operated in three different regions: America, Europe, and Asia-Pacific. They developed models for those regions, and if you looked at the competition, like Toyota, they had global vehicles. Because of their global vehicles, they had better efficiencies, better economies of scale.

We realized we needed to operate as a single, global enterprise, so we started generating metrics around how close we are to that. How common are our parts, for example? That’s the most basic question. Once we generated those analytical metrics, people realized that there was an opportunity to deliver huge savings and efficiency by increasing our commonality and globalizing our operations.

What we lacked was a common taxonomy or common structure to talk about the vehicle as it’s composed into vehicle systems. So, we came up with a process based on being able to analytically understand how the product structure drives complexity into product development, and from there optimize the product structure to make it both global and minimize complexity in product development.

That project was immensely successful, and the result has been the product structure that has been put into the company and into how it does business. It kind of opened the floodgates.

Just to clarify for people who maybe don’t know the car business so well, which could include me, product structure means that the parts —

The way you develop a car is, you break it down into vehicle systems, like your fuel system, your brake system, your seating system, and then different teams get responsibility for different systems. And the problem that we had, when we operated as regional companies, was that there wasn’t alignment on that product structure. So, if you had a list of parts and went to one engineer in Europe and one in North America and said, “Mark off all the fuel-system parts,” you’d get lists that were 80% or 90% common, but not identical.

When you’re operating separate regional companies, that’s okay. But if you are trying to do global engineering, and you’re looking to uniquely divide the process so there’s no overlap in parts and no missing parts, you can’t have that situation. We all need to agree explicitly what the parts of the fuel system are. We can’t have gaps or overlaps.

So, that’s what the vehicle standard taxonomy or structure means. And then the question is, how do you optimize that structure? It turns out that you can optimize that structure to reduce the complexity in your engineering process.

What kind of pushback did you have to overcome during that period of time?

A change in the structure involved changing people’s responsibilities, and people don’t like that. We needed to have strong executive support to do this, and because of the imperative of globalization, we had that. We also had to have executive buy-in to the principles of how to do this optimization. Once we got the executives to buy into the principles, we could actually apply them at the granular level to get that agreement.

The company knew it had to become global and people understand that you can’t do global engineering if you don’t agree on how you’re going to break down the vehicle into systems. But not everybody accepted a single way to do that. What it boiled down to was management support, and clear analytical principles, not just for the need to do it, but the specific method that we proposed to do it to optimize it.

Was there a signature moment where you were able to overcome a skeptic or win a battle using the analytics?

From my experience, there’s typically two phases: There is a phase where the data and analytics team gets a chance to prove the concept. Executives get convinced; “Yeah, there’s merit here, let’s go try this out.” And you go and develop a prototype, a proof of concept, or a detailed proposal, whatever it is. The next phase, you go back to the executive and the affected stakeholders, and then the decision typically is made.

So, I would say that it’s up to the data and analytics teams to get to that second, true decision point with enough support — which is driven by the analytics, but I wouldn’t put it as analytics “wins the day.” Bringing stakeholders along is essential.

Has it become a strategic tool?

I wouldn’t talk about it as a strategic tool. I would pose the question differently. I would say that the company today is much more aware of making decisions more data-driven, and that’s the opportunity for analytics to come in. So, I would say that there’s a lot more awareness that when we’re going to make a critical decision, we need to ask, “What’s the data-driven way to make it?” And analytics ends up being the tool to translate the available data into the appropriate decision variables.

Has the value of analytics changed in terms of, say, strategic efficiencies at Ford?

Definitely, yes. I started the team in 2005 — myself and another person. And the team grew to around 100 people in the beginning of 2015. Such growth speaks to the value that people saw in data and analytics during a period.

Let me give you an example of where we applied analytics to optimize the introduction of new capabilities into product development. We created what we call an “information factory” model or view for the product development process, where we borrowed concepts from Lean. We looked at the key information, all the facts that flow through that process, and at the fundamental operations that engineers perform on them — which of those are value add, which of those are non-value add. We did a significant data collection exercise, analyzing how engineers spend their time, and we got some very good analytics from that — for example, what percent of the engineer’s time is spent on so-called value-add activities versus what we call information alignment activities. That was a very interesting number.

Once we had that model, we had a list of projects that were candidates for investment to improve the productivity of the engineers, and we were able to optimize that list of projects to maximize efficiency. We showed that the optimal selection process brought 50% more lifetime efficiencies for engineering than a manual selection process.

Ford has done some exciting stuff over the last 10 years. Are there techniques or emerging trends within analytics that you’re keeping a close eye on or that you’re hoping will become really valuable for the company?

Yes. I personally see a big transformation coming similar to what happened in the automotive industry. Originally, we had a few master craftsmen making hand-built cars. One or two people would build a complete car, and then they’d move on to the next car. And then the assembly line came in and similarly changed how cars are built and allowed lower-skilled people to build a high volume of cars at good quality.

I think we’re going to get the same kind of transformation happening in analytics, where we’re going to move from the individual data scientists solving a problem to what I call “pervasive advanced analytics” — analytics on an industrial scale. This is one area that I think is going to make a big impact, and we’re right on the cutting edge of figuring that out.