Why So Many Organizations Struggle With Data Management
Nearly all organizations recognize the value in the data they hold, but a large percentage struggle with what to do about it. Information Management spoke with Paul Blasé, principal for the Global and U.S. Analytics for Consulting at PricewaterhouseCoopers, on what’s going on here, and what CDOs and CIOs can do about it.
Information Management: Why are some companies still struggling so much with data analytics and big data?
Paul Blasé: In the paradigm where we more data and therefore more ability to use analytics techniques that can specifically be used to impact financial metrics through better forecasting or provide scenario analysis where the c-suite can understand all the paths that the ship can go down and pick the right one, one challenge is c-suite accountability. Meaning the right level in the organization has to provide the mandate and the accountability and the purposes for using data and analytics and be very clear about the type of value they’re searching for. The value could be improving products and services with data and analytics-driven services. It might be improving a long-term financial metric. It might be improving operations. All of those things have very different implications in terms of what you have to build to get the value. I think frequently that accountability and clarity of purpose is just not well articulated.
IM: Who would you say is most at fault – is it IT? Is it senior leadership? Or is it a combination?
PB: I think a paradigm we see is that it’s very easy for organizations to say I have a BI group and start saying well shouldn’t they be the organization or the function that I should depend on to advance my analytics capabilities. Very often those groups aren’t equipped with the right talent. They have a different mandate. Their mandate isn’t necessarily to solve complex business problems. It’s to provide dashboards and reports and meet efficiency-oriented service levels. From that vantage point, it’s not fault. It’s just a misunderstanding in the terminology of data and analytics and what it is and where it potentially sits in the organization. In that case its incumbent on whoever the leader is running BI to articulate the case and explain why this isn’t your father’s data and analytics, so to speak.
When we look at the business side of the equation, there are a lot of function-based analytics groups. Marketing departments, supply chain departments, sales organizations, service organizations, pricing, credit risk, you name it. They usually have some small group of business analysts and I think the challenge there is they frequently may not have some of the skill sets they need in computer science or mathematics or operations research and statistics and the like to really push the envelope given the data available and the types of modeling techniques that can be applied. In that case, what we’re seeing in organizations is that we’ve got the talent spread all over the organization. If we want to solve problems that are really about cross-functional improvement – so it might be how do we improve conversion and value of customers across the demand funnel – we have to pull marketing, sales, product development, pricing all together if we’ve going to build a model that tells us how we do that more effectively. In that case it’s about figuring how it should be organized and who should be accountable for a cross-functional problem like that because when you rely on any one function to do it it’s not how they are incented, the reporting lines aren’t clear, the objectives aren’t clear, and it doesn’t work.
IM: Stated goals around analytics usually center on improving business decisions, better engaging customers, reducing operating costs, etc. How well are organizations doing in these areas?
PB: When we think about domains of improvement, we think about the value of the domain and then how to apply analytics and data to achieve the goals, we think about where can we improve the speed of the decision – let’s say the decisions you make that impact how you engage with customers, like what’s the next logical offer or how would you tailor a price based on a situation. There’s a speed component implied. There’s also a sophistication component implied, which is what type of data are you using, what type of modeling techniques? When you think about it that way there are four dimensions.
One is just achieving basic goals like diagnostics – let’s say profile-based analysis or light diagnostics of ‘what happened’. Organizations are generally pretty good at profiling and to some degree diagnostics of why something happened. What drove an operational cost up or down, or what happened in terms of customer acquisition over the last five years. Where there is a lot of room for improvement still is diagnosing and really understanding the underlying drivers. If costs are up, why? Understanding the five variables that are most correlated and then saying, we’ve actually built models to prove a cause and effect relationship. Once you have that then you can say, alright I know there is cause and effect, I can deploy a team differently in this way to reduce cost or I can cut headcount here to reduce cost. I think there is room for a lot of improvement there.
Then, when we get into situations where you need very fast models, let’s say companies that provide a lot of online retail, they’re constantly making offers, testing offers, measuring responses to different types of marketing messages, or price or baskets that are proposed. We call that accelerated agility and most organizations struggle to really build that type of sophistication into their work flow where there is a combination of digital channels or call centers or front office sales people or agents that are interfacing with a customer. There is a long way to go before we will see prescriptive or predictive rules-based analytics being embedded in workflow in a way that really improves performance.
Another dimension is when you look at the sophisticated models where speed is important and you’re doing long-term forecasting or scenario planning, we’re starting to see an uptick in the c-suite’s interest in building more sophisticated capabilities that allow them to combine their judgment and experience and intuition with the science, which is the data and the models. In that case, we see large-scale digitalization, we see different simulation models that employ gaming into the analysis being well received because it allows the s-suite to have the collaborative conversation and they use their experience and judgment to set parameters on a model, they run the scenarios and then they see the results, and they get to debate, why did the market grow at this rate when we assumed it would grow at this rate, or why did this competitor gain share when I assumed the opposite would happen because I dropped my price. It’s about combining the intuition and the experience with the science of data and analytics together to help an executive team make better decisions. That’s where we’re seeing traction, but I think there is also a lot of opportunity, and the challenge is that it requires a culture and a mindset shift starting from the top to want to operate in a data-driven way throughout the organization, as opposed to operating primarily with experience and hindsight when it comes to data and analytics.
Q. Who should be actually leading big data initiatives, and what are the good or bad reasons for doing them?
PB: The bad reason would be hearing about the over-used term of big data and the trends, and saying, we need to have a chief data officer or chief analytics officer because of that and putting one in place before there is alignment in the c-suite about what that individual is supposed to do, how they’ll be measured, what their incentives would look like, and what type of accountability they need to have. What needs to happen to even figure out who should undertake the data and analytics opportunity and challenge for companies is figuring out where you want to generate value. If there are big problems like thinking about reserve exposure in insurance where you could do a better job connecting real-time underwriting and claims data to exposure models that are cross-functional in nature, or you’re thinking about a product development life cycle as it feeds into customer experience and doing a better job of taking data about the experience your customers are having with social media or how they respond to service calls, to voice recordings, and have a feedback loop into product development, those are all cross-functional problems and they’re tied to a lot of business value, which in the case of many companies has a direct impact on shareholder value, then you have to start looking at somebody owning it that sits at the c-suite table that is viewed as a peer of the leadership and helps the apply data and analytics in the most optimal way to the organization.