Nationwide CIO: Here's how we centralized IT
According to the Society for Information Management, the average CIO tenure is just under five years. Mike Keller, CIO of Nationwide, has held his position for more than 15 years.
What’s his secret? Keller says he’s had only two bosses during that time, noting that CEO transitions are one of several factors that can “end things early.” But clearly, he wouldn’t have lasted so long if he hadn’t successfully moved Nationwide forward.
That’s no mean feat in an IT organization the size of Nationwide's, with approximately 5,000 permanent employees and 3,600 contractors. Many of the advances have come through adoption of lean and agile processes along with what Keller terms “unsexy” modernization of legacy systems.
But Keller also bucked a widespread trend: Instead of decentralizing IT and allowing it to become absorbed in lines of business, he has centralized many major functions. InfoWorld’s hourlong interview with Keller, edited for length and clarity, began with that very question.
InfoWorld: One of the trends we’ve seen over time is that IT becomes less and less of a separate entity from business units. Does that track with the way things are organized at Nationwide?
Keller: No. When I joined the company in 2001, it was a decentralized IT organization. There were some shared corporate staffs, but the bulk of IT hard-lined into the business units. Shortly after I got here, we created a federated model, a matrix-managed model where we left the application development groups hard-lined into the business but with a strong dotted line in to me -- and then hard-lined in infrastructure, architecture, security, those kinds of functions.
In 2008, I believe, we centralized IT and went through a pretty major transformation of how we do IT. That’s when we established our IT Delivery Model and did our first generation of that.
What I have seen and experienced both at Nationwide and other places I’ve been is that a lot of people like to keep IT decentralized to keep it very close to the business. You tend to get high degrees of inconsistent performance and you don’t leverage economies of scale. Other organizations will centralize and they will get the benefit of standardization and economies of scale but lose touch with the business.
We’ve created a construct that we call our Organizational Reference Model. Basically, we try to architect the organizational model so that the right things are very close to the business and the things that would really benefit from centralization are centralized.
InfoWorld: Break that down for me. How does that work?
Keller: The business-facing component of our organization model we call a business solution area. But accountability for planning, building, and running of all of the applications and data assets that support business are managed by an IT group that is very tightly aligned with the business area.
As an example, in our world if you look at our property and casualty claims function, we have a business solution area -- Claims IT, which is a group of people who are dedicated to claims and who learned the claims business processes. They manage a relationship with those partners and they own all assets and all projects that face off with that part of the business. That’s the part that has kept the business partnership and the business intimacy in place.
If you look at the things that have the biggest economies of scale -- infrastructure, application development, governance-and-control functions such as cybersecurity -- those things are centralized. We built a model where we try to be very clear which organizational units are accountable for what and how they interact with each other. We have tried to tweak this in such a way that you get both efficiency and effectiveness as an organization, but also business intimacy.
It’s been tricky. I will tell you that when we centralized a lot of our application development, we made major progress and won awards for how we design, develop, test, and implement. Yet when you looked at the end-to-end process, it really wasn’t as efficient and as effective as it needed to be because of the handoff from plan to build, if you will -- things like the requirements pipeline and the handoff from build to run -- there were just some orchestration problems.
We’ve done a lot of work to get better processes implemented, like lean, but also worked a lot on culture and teamwork so that you can basically optimize and have it feel like one cohesive organization. We have over 20 of these business solution areas that are the business-facing, business-intimate groups streamlining how they work together with the centralized application development group so you get the best of both worlds.
InfoWorld: If I want an application development project done, say a customer-facing web or mobile application, walk me through how that process would work.
Keller: Let’s say that our direct property and casualty business wants a mobile app. The business team that would be working on that has a very tightly integrated and aligned team, our digital business solution area, and they would work together on requirements.
Development lines are typically staffed with people coming from both groups. The people that are in that business solution area who understand that part of the business best are in the same development line with scrum masters and developers and testers that are coming from the central shared services group. We can very quickly bring these lines together, and if the demand goes away and the project is done send people back to their home in a variable-capacity model. It’s highly efficient and it brings together the right skills from across the organization.
InfoWorld: Do you see IT as an internal services organization to the rest of the business? Is that how you think of it?
Keller: I’d say that’s accurate. I hope and certainly have aspired to have that internal service provider be one that provides consultation and advice, not just development services or operational services.
InfoWorld: I’m interested in your implementation of lean and agile and how you recognized that had to happen. As you know, there are many organizations that talk the talk, but they’ve just solidified the agile methodology into something that ends up resembling waterfall more than agile.
Keller: There wasn’t a clearcut point at which we said we’re going agile or we’re going lean. It was more evolutionary. When we still had decentralized IT, we found that most of the larger teams had started to do some agile, but when we brought all of those practitioners together we discovered we had a dozen different flavors of agile.
Some were working better than others. When we began experimenting with centralizing IT, we didn’t do all that in one step. We brought together the best thought leaders and told them that they needed to come up with one agile methodology for the enterprise. We had some really smart, capable people who did the work to try to sort through all the best practices internally as well as comparing notes with external consultants and partners to learn from them.
We also decided to go through external process assessment so we pretty quickly got to CMMI level three, which typically isn’t done in an agile environment. We were actually one of the early lean, agile, CMM-certified groups. What we found was the ability to create working software quickly, create software that met user requirements, the ability to have low-defect (or even in a lot of cases, defect-free releases) was just producing substantially better results than waterfall and better than a lot of the less mature agile practices that we started with.
We started bringing together our practices, trying to refine them, seeing that we liked the results, putting more and more scale in there to the point where now we’re moving into a model where it will be all of our application development in that model. One of the core principles of lean, as I’m sure you know, is creating a continuous improvement engine.
InfoWorld: That’s what it’s all about.
InfoWorld: To do that you need measurement of how your application is doing. How do you measure the effectiveness and then turn that around into direction for improvement?
Keller: A lot of what we do with lean is decide what we want to measure -- measuring story point velocity, measuring defects and using local visual management systems -- but then we aggregate the data. We bring it together and measure it and then we also benchmark it so we know where our productivity, quality, and velocity are compared to benchmarks. Just like you try to continuously improve practices, you also try to continuously improve the metrics.
We measured things like productivity and velocity in development, but we didn’t have an end-to-use view. As we started to look at devops concepts we realized you really should be measuring it all the way into production. We’re continuing to refine and move more and more to full end-to-end requirements through production.
InfoWorld: One of the perspectives I’ve heard across several organizations is that one of the transformations of IT, whether you call it “digital transformation” or not, is the idea of transforming IT from being a cost center to a revenue generator. Since you’ve been there for 15 years, have you seen that kind of change in attitude?
Keller: Yes. I guess the way I would look at it is we build and run different kinds of systems. You can pick your terminology, but historically IT was back-office automation, transactions across what most people today would call systems of record. Most of those systems of record were really geared towards automating tasks to be more efficient. That is still an important part of our portfolio.
InfoWorld: Roughly how much of your portfolio is systems of record versus systems of engagement?
Keller: I haven’t measured it that way, but we’re also starting to look at a third category: systems of insight. Sometimes it gets pretty fuzzy. For example, we’ve made a big investment in customer information management. That includes an operational customer information file so that we know all of our customers, we have clean addresses, we know their preferences, we know all their products with us, and we capture all of the interactions in the interaction history.
In fact, this was very heavy lifting. Our customer system of record used to be the individual product or policy systems. We changed that in most cases, so the enterprise customer file became the system of record and we had to create some very challenging two-way interfaces between that new set and all of those legacy systems that were not designed to have common customer formatting and weren’t designed to have somebody else be in control.
There was a big traditional system of record component of that, but as I said, when we built that out we also started doing things like keeping all the interaction history and taking those operational data stores and feeding data warehouses and doing predictive analytics on it. We spent well over $100 million on customer information management, but I’d be hard-pressed to say how much of that was system of record and how much of that was system of insight. It was both.
I can tell you that on our project portfolio, we have more going into replacements of core systems of record than we do in either systems of insight or systems of engagement -- but we are investing pretty heavily in all three. What we’re finding is you can’t really do the systems of insight or the systems of engagement well if you have broken core systems, broken systems of record, where you can’t feed those things in real time or you don’t have the data quality you need.
InfoWorld: Would you say that what you just described is one of the biggest initiatives on your watch?
Keller: Yes. To your earlier question if IT has to save money or be more top line, my answer is yes. A lot of times they go together.
InfoWorld: A flaw in the distinction between systems of record and engagement is that people act as if they’re not connected.
Keller: Yeah. In other industries like retail, where you might have 20x the volume of customer interaction on shopping than you do on actual buying, you can separate that shopping engagement and have a lot less dependency on systems of record. But once they buy, you have a touchpoint there.
Given the nature of our businesses, you really can’t have somebody shop or change their products or services without heavier degrees of interaction with systems of record. We can’t create the kinds of systems of engagement we want without modern backend systems. If you’re going to let somebody go on a mobile phone and … change their auto insurance from one car to the other, they can’t do that without hitting the backend, and if you’re dealing with 1980 Cobol back end system it’s not very pretty.
Our conscious decision has been invest in digital, invest in data, invest in next-gen technology so that we can provide the kind of customer experience we need. But we’re making an even bigger investment in core systems replacement because we don’t think those other investments would actually work or pay off if we didn’t do the former.
InfoWorld: Tell me about the most exciting initiatives happening right now.
Keller: I’m very, very excited about our core transformation programs. We have just about completed moving all of our claims to Guidewire and the productivity of the claims organization is up dramatically. The productivity of IT in that area is up dramatically. We’ve cut development timelines and costs by 60 percent, meaning that we can do two and a half times as much work in a given amount of time for a given amount of money than we used to do -- and we can start to build customer-facing digital capabilities radically faster and better and do things we couldn’t even have done before.
We’re beginning the rollout of our personalized policy administration, and that’s where a lot of our direct consumer interaction will be. When both personalized policy and claims are done, I really think we will save a whole bunch of money internally. We’ll be much more agile as a business, so we can change products or pricing or internal processes much better and we can create a much better customer digital experience.
Getting these core transformations done isn’t sexy. It’s very heavy, hard lifting, very expensive. But when you see what it enables, the dramatic business and technology simplification and the real benefits of being on a modern architecture, I actually find that very exciting. The cool stuff that you can build on that is also very exciting.