How systems of intelligence drive change at P&G

Javier Polit, CIO of Procter & Gamble

Javier Polit, CIO of Procter & Gamble

Until now, IT has largely existed outside the realm of analytics and metrics that we apply to every other area of the enterprise. There’s an obvious irony here: IT is ensuring that every other segment of the company has products to assess performance while not having the same accountability for itself.

That is starting to change. I’ve profiled Numerify’s technology before and how by using systems of intelligence for IT, companies can radically reshape how their IT functions across the organization.

Recently, I spoke with Javier Polit, the CIO of Procter & Gamble. Our conversation focused on these systems of intelligence or packaged, productized forms of BI for IT. Systems of intelligence have a lot of similarities with productized analytics, which I’ve examined extensively in my Forbes column (see examples here and here) and at my content marketing firm, Evolved Media.

Polit highlighted that systems of intelligence provide data that can guide automation, such as Robotic Process Automation and bots, that bolster processes overall. And for Polit, the four main benefits of using systems of intelligence are that companies can:

  1. Gain end-to-end visibility
  2. Find root causes of repeated problems
  3. Be predictive to determine which changes are likely to cause problems
  4. Improve the efficiency and effectiveness of IT overall

He also offered insight into how to get buy-in across the organization to achieve these benefits. What follows is a slightly edited version of my interview with Polit about systems of intelligence and the role they can now play in businesses.

Woods: Why does IT need systems of intelligence that apply metrics in a much broader and more detailed way than in the past?

Polit: The IT footprint is expanding dramatically. IT is in every single corner of the business now. And you have business leaders looking at different solutions and platforms and we need to improve how IT operates.

So I look at that development from two different dimensions.

The first dimension is we need to not only measure customer experience, but I think we have to start looking inside our four walls as well at our internal users. We’re starting to look at partners and tools that provide guidance and analytics about how systems are performing and the levels of service that we’re giving to our internal users. Understanding those levels of service allows you to improve the internal user experience, especially as we start looking at using bots inside our environments and the Robotic Process Automation (RPA) work that we’re doing. You need that type of information and insight to make sure you’re focusing on the right internal behaviors, metrics and capabilities.

The second dimension is about performance. We’re instrumenting IT in the same way that we’re instrumenting the business and providing the same capabilities and the same insights inside our four walls. And the optimum goal is to continue to use data and insights to get to a distinctive level of performance, both from an industry perspective and inside our four walls so the business sees a significant difference. We’ve already dramatically reduced the number of incidents, which improves availability. The business has already seen those improvements.

Woods: In what ways does the prevalence of data improve operations?

Polit: We have tools that monitor incidents and exception reports that look at time of day and at events. We’re building algorithms to make sure we understand the causes of these events. We then go back and find root-cause solutions for them. Again, at the end of the day, the goal is to improve performance and the user experience.

Woods: Can you give me the before and after of what happens after you implement systems of intelligence for IT?

Polit: Before we had this data, we were just looking at the number of incidents. We were being reactive in possibly finding out what could have been the issue. We didn’t have historical data that showed on a given week at a given hour or time of day or on a particular day we were having recurring incidents. We didn’t have the ability to look in a designated area based on what was happening in the environment at that particular point in time. We just had an overview of the number of incidents and we were trying to find root causes to drive those down. We were making progress but not in an exponential way. When we started getting more granular data, we really started improving those areas.

Afterward, what goes away is that 48-hour, week-long root cause analysis into what actually triggered the instance and what could have happened. What comes to the table is a broader perspective. Your aperture is wider in regards to what caused the specific incident, and you can go back and say, you know what, we have to rewrite this code or change these application interfaces. But at the same time you improve governance models and frameworks.

In short order, a system of intelligence gives you clear insights into what the issues are and for the short or medium term, you have some additional work to do that’s meaningful. In the long term, you have exponential and distinctive performance improvement.

Woods: So you think that this helps make you more effective at managing your IT portfolio?

Polit: Absolutely. It helps us with the simplification of our environment by identifying tools and applications that we can sunset since they are not strategic for the direction we’re going. It certainly helps you simplify or prune, which by definition also reduces your total cost of ownership because you’re no longer maintaining systems or tools that you don’t need, which are taking up storage capacity and CPU cycles.

Woods: And how does it change the way the entire leadership team approaches IT?

Polit: You start seeing correlations. The leadership teams with ownership of certain verticals or certain businesses, they start saying, yes, that’s the time of day when we do more of this type of work in that particular facility or in that particular area. And they start linking the incidents to what’s happening in their business environments. As a result, they get a lot smarter about the dependencies and start understanding the whole ecosystem of systems and applications.

Woods: I see. So the idea is that it’s not some enterprise architect over in the corner that understands this, it’s actually the VPs who now can basically go under the hood and see a lot more about how the engine is working.

Polit: Exactly. So our data gets shared on a six-week basis with myself, at my level and my vice president’s level and then with the actual line managers who are working the issues. Its gets elevated all the way to VPs and to my office’s level.

Woods: How can this data and awareness help you improve your overall architecture?

Polit: As you start getting the data in, you start realizing that at a certain point it could be your API that is causing issues and then you may find out that you don’t have a consistent strategy around your APIs or a common standard governance or structure around your APIs, as an example. And you find different areas of code or interfaces or ETLs that you basically need to transform. And then you start setting governance and a framework saying this is the way that APIs are going to be built going forward, and we’ll have a standard library of objects or API capabilities that we start leveraging systematically across the organization.

Woods: So are you saying that the problems that are causing the most trouble become obvious much faster and then you can come up with a strategy for addressing them?

Polit: It gives you a variety of insights that you need to start looking at. It could be that a particular partner isn’t writing good code, or maybe it’s something we implemented.

Woods: Does a system of intelligence then help manage partner organizations as well?

Polit: Yes. Systems of intelligence start tracking how many instances or exceptions you’re seeing and what happens in regression testing. This can reveal that the partner wasn’t doing adequate testing. For instance, you can learn that before we put a change into production, the application failed five or six times and then after it was in production for about eight months it started to get heavy loaded and it failed another fifteen times. So with systems of intelligence, you start looking at performance and capability of the architecture of the partner that’s helping you deliver those capabilities. And I think due to those insights you start determining which types of solutions the partner is good at developing. You start giving work to partners who have demonstrated a long-term end-to-end delivery capability that has caused very few incidents in your ecosystem.

Woods: Does this then also help with outsourced agreements with defined service levels? For example, I’ve heard about people resetting the clock to reclassify incidents, in essence gaming the system.

Polit: With that type of reclassification, an outsourced partner’s reports don’t always tell you the truth about what’s happening in your environment. Systems of intelligence give you much better clarity and insights in regards to what’s actually happening and the kind of incidents you have. And then it gives you an opportunity to bring those incidents to the table and talk with the partner about classification of incidents.

Woods: When you first start dramatically expanding the mechanisms to create metrics, what low hanging fruit do you find? What are the areas where this has an immediate impact?

Polit: I think the first time you start demonstrating the insights that you’re getting from having these type of capabilities, you very quickly see that in the past you thought you were doing a good job of reducing incidents and understanding incidents, but you really didn’t have a deeper understanding of what was causing them. That is eye opening for the leadership team holistically.

You start finding out root causes and which applications are having issues and what time of day they’re having issues. You can see dependencies between different applications that are making API calls to other systems so that you can make certain that those systems are available. For example, you could see that latency in one system versus another created an incident. And you don’t have those insights normally until you start having these performance tools looking under the covers.

Woods: Given these benefits, as a senior manager, how do you ensure that your line reports act on this information?

Polit: Well, they are aware now that we have a review every six weeks for twenty minutes where the leader in that space comes and provides us all the insights and the actions they’ve taken and how they’re improving the environment. And the leadership team is aware of the progress that they’re making and the work that they’re doing. So we actually have built a routine.

Woods: So every six weeks they have to explain how they’re using the metrics to improve management.

Polit: Yes, there’s a cadence.

Woods: Now, why in the past has IT been so bad at modeling itself and defending the value of technology investment? Because this fact has led to this kind of general notion that the only thing that you can do to an IT budget is cut it, you know? That’s the big victory.

Polit: I think in the past as you look at this, for someone who has been in this industry over 30 years now, I think we’re always focusing outside IT and how to build insights and deliver capabilities outside. And now that IT is in every corner of the business, you need to start looking inside your four walls and how you improve a lot of the work inside. And the same analytics and insights that you’re providing the businesses to improve top line sales and bottom line operating income need to be applied to improve how you’re running the business internally. Which, at the end of the day, serves your end user as well. So I think there’s starting to be a big paradigm shift there and we’ve started on that journey.

Woods: That makes sense. So it’s clear IT needs these metrics of their performance. But what is the case for using a product to implement systems of intelligence?

Polit: You know, there are areas where you have to build your own algorithms based on what you’re trying to do pertaining to a consumer ID graph or a customer’s behavior in regards to synergies of products, et cetera. But when you look at this area now, some leading tools are making big impacts and several Fortune 50 companies are starting to use them. That’s what we did. We chose something off the shelf that was very quick to plug in and gave us quick results.

Woods: So it’s a classic buy-versus-build analysis.

Polit: Exactly. Buying a productized system of intelligence allows us to get the benefits of systems of intelligence much faster than if we built it ourselves.