No one wants poisoned food — or the resulting misery — and CIOs, chief data officers, business analysts and other data wranglers don’t want poisoned data. But it’s out there, perhaps spreading even now throughout your company, even as you strive to make data-driven business decisions. Anthony Accardi, CTO at online shopping company Rue La La, described data poisoning as “an illness caused by a toxic relationship with data.” It happens when people are misled by large quantities of information.
“Over a long enough period of time, you get to bad decisions, which overall can erode your competitive advantage and lead to failure,” Accardi said at the Argyle 2018 CIO Leadership Forum in Boston earlier on May 2.
During his talk, Accardi detailed a number of things you can do to counteract data poisoning. But once you recover, you’ll likely find your data ecosystems inundated by questions about the data. “Those all go to your analysts, and you now have a big bottleneck there,” he said.
There are two management treatments for “meta-data poisoning,” as Accardi called it. Apply them, get questions answered quickly and effectively, and you’ll be closer to making smarter, data-driven business decisions.
Do as product management does. Your company’s product team thinks constantly about your catalog, your products, their features, Accardi said. They lay out a roadmap, determine which ones get priority for rollout and then work through the to-do list.
“You can actually do pretty much the exact same thing in analytics with questions instead of features,” Accardi said.
For example, let’s say you wanted to quantify the impact a push notification has — the question: How much impact? Determine the business value of answering that question and the effort it will take. Add it to the myriad other questions that need answering and rank it by importance; then assign a team to build the intake process for analytics.
“You start realizing that a lot of people are asking the same question, or different variants of the same question,” Accardi said. “There’s obvious value there in reducing the noise and just consolidating and answering that question once rather than dozens of times.”
Establish a data center of excellence. A data analytics program needs an organizational model, and there are lots of them, Accardi said. At one extreme is a centralized data organization, which employs everyone in the company who deals in data. That can introduce lots of efficiency when working toward data-driven business decisions. “The downside is it might be very distant from the actual business function and business needs,” he said.
Decentralization is at the other end: Each business function has its own data and analytics resources. That’s great for day-to-day business, but it’s harder to achieve critical mass on bigger, companywide data programs.
“Most people find the magic setting somewhere in the middle,” with specialized analysts in different functional areas of the business and a center of excellence, or competency center that offers best practices and support on data and analytics, Accardi said.
This organization can enable stakeholder management, helping the people who will be asking questions about the data reach their objectives, he said. It’s also key to ultimately making data-driven business decisions, building the models and prioritizing projects on the basis of available resources, “not just the intake of questions but the larger features and then actually build those things.”