In this article...

  • A comprehensive enterprise data strategy leverages a variety of data to support the company’s overall business strategy.
  • There is a natural progression from the data strategies of today to the cognitive-enabled data strategies of tomorrow.
  • An analytically mature organization will find it valuable to add cognitive technologies to its data strategy.

The world is witnessing an explosion of data brought on by social media, Internet of Things sensors and multimedia content. The data may be from private, public or third-party sources and include information not organized in a predefined manner.

An enterprise data strategy isn’t an absolute requirement for those who are beginning to experiment with cognitive technologies—applications and services can be built in an ad hoc fashion. But for companies that aim to infuse intelligence into everything they do, developing an enterprise data strategy will pay off over the long haul through increased efficiency and speed.

Appoint a Chief Data Officer (CDO)

In many organizations, data is collected and stored within individual business units. As a result, much of that knowledge is difficult to access and use for purposes that span business units and functions. That means the most strategically important enterprise initiatives may be starved for data.

The CDO is charged with developing a strategy that makes the organization’s wealth of data easily accessible for a wide variety of uses, including such purposes as transforming the enterprise or its industry. The CDO is a key member of the executive management team because data has become critical for a company to achieve competitive advantage and fully deliver on its value proposition to clients.

According to Inderpal Bhandari, IBM’s CDO, the primary job of the CDO is to develop an enterprise data strategy—and then execute on it. He explains that the core elements of the data strategy are setting up enterprise-wide governance and management systems, becoming the central data source for the organization, building deep data and analytics partnerships and developing and scaling talent.

There’s one more critical role for the CDO. That’s to catalyze the shift to cognitive computing by sponsoring a series of “proof point projects” aimed at harnessing cognitive technologies. For example, recently more than 20,000 IBMers competed in a program called Cognitive Build, an enterprise crowdfunding project that resulted in nearly 3,000 start-up projects that leverage cognitive technology. Business Insider called the program “an enormous artificial intelligence hackathon” and possibly “the largest AI hackathon every held.” By driving adoption of new capabilities in this way, the CDO can test the strategy and operations against real-world challenges and accelerate the journey to becoming a cognitive business.

Establish a data governance system

Data governance is the top-level policies, processes and technology frameworks that guide all of the keepers and users of data in the enterprise. Once this rule book is established and followed, users of the data are better able to access new data sources, integrate various types of data with one another, innovate using superior data-driven insights and, ultimately, support their company’s overall monetization strategy.

Data governance is an ongoing process rather than a onetime initiative, so organizations need to develop the rules and also enforce and modify them in response to feedback and evolving data sources. As data evolves, so should governance policies.

The governance rules should cover all computing domains, including data formatting, security, privacy, acquisition and licensing of third-party data, usage rights and access controls, and data quality assurance. In addition, if business units and functions are being charged for use of enterprise data or services, pricing, metering and charge-back mechanisms need to be established.

The goal is for data governed by the CDO be totally trusted. If the business units tap into these sources they can rest assured that they will be able to use the data in prescribed ways without running afoul of security, privacy or licensing rules and commitments.

Think of data scientists who establish and administer the governance system as something akin to the corporate law department. Just as lawyers review all sales and purchasing contracts, a data governance team reviews and weighs in on the use of enterprise data.

Consolidate enterprise data

Because most data gathered and stored by an organization is captured for a specific purpose by a particular organization, many enterprises have incomplete knowledge of the data they already possess and the state it’s in. So a key first step is to use an enterprise-wide survey to do a complete inventory of the data within the organization. Then map the data, its location and its format. Over time, the enterprise should create a system that tracks the status and movement of data—automatically updating so the enterprise data map is always up to date.

“Potentially, you can take all of these processes and make them cognitive,” Bhandari said, “automating discovery of data, tracking updates, learning, and preparing the data pipeline for your applications.”

Most business executives are familiar with the data warehouse, a repository that contains and organizes data from disparate systems, integrates data of different types, and makes it available to a diverse set of users via a variety of analytics tools. In this era of big data and cognitive systems, a new type of repository is emerging—the data lake. These structures are capable of scaling massively and handling both structured and unstructured data.

When the data is brought into the data lake, it is curated, annotated and, in some cases, reformatted so analytics tools can process it efficiently. Initially, much of the curating and annotating will be done manually by subject matter experts and data scientists but, over time, a cognitive system can do this work with little or no human assistance.

A wide variety of data can be stored in a data lake, but none will be more strategic than what we call master data. Master data is information about the people, places and things that are critical to a business. Master data is connected to transactions, but while transactions live in particular business units and functions, master data cuts across many of them.

Contextual information is also important data to place in a data lake. For instance, retailers want to know not only what a customer bought, but also where they bought it, what the weather was like when they bought it, and what they tweeted when they bought it. By connecting master data and transactional data with this contextual information, businesses can get a 360-degree view of a customer. They know who each individual customer is and can better predict what that person might want or do in the future.

Since most enterprises use significant amounts of data purchased from third parties, it’s critical to consider this material when setting up systems for consolidating and managing data. Business units and functional organizations may subscribe to the same or similar data—sometimes from the same providers. Consolidating enterprise-wide licenses will make third-party data easier to find, reduce costs and, potentially, enable the licensing of some of those data sets to clients.

Train and organize teams of data specialists

Data is an abundant resource within enterprises, but many organizations do not have the human resources to manage and capitalize on it. Specifically, there’s a shortage of trained and experienced data engineers and data scientists. Engineers design, build, operate and continuously improving data-centric systems and services. Data scientists are the analysts who use them. Both are aided by subject matter experts.

In addition to data engineers, data scientists and subject matter experts, there’s another category of professionals—business unit data officers—people with deep knowledge of data and analytics who help set strategy, design governance policies, and design data frameworks.

Business unit data officers can play key roles in the CDO’s organization, but they can also be in business units and functions across the enterprise. These business unit data officers can help evangelize the benefits of the enterprise data strategy, and they’ll understand the needs that the business units have for enterprise data. These data officers can establish a cross-enterprise council of business unit data officers that meets frequently to learn about new rules and policies, share best practices, and provide feedback to the CDO.

Although business unit data officers might report to the CDO, their activities can be funded by the business units to which they are assigned. This arrangement could help keep a balance between the interests of the enterprise and the needs of the business units. The key thing is to have a set of people who understand data governance and management, and subject matter experts who can help design and train cognitive systems for specific uses.

A number of IBM’s clients have asked us to create cognitive computing centers of excellence within their organizations—basically training their people to evaluate cognitive use cases, choose the appropriate data and technologies, and build and deploy solutions on their own. These very ambitious clients will need a deeper bench of technical talent than others.

Accelerate your data strategy with catalysts

Cognitive computing has tremendous potential for transforming businesses, industries and professions. Yet it’s so new that there are many unanswered questions. How do you identify use cases within your business that would benefit from cognitive computing technologies? What data do you need, and how do you pair it with specific cognitive technologies?

At the same time, you face a chicken-and-egg dilemma. You are building an enterprise data strategy to support cognitive computing applications at the same time that you’re learning to build the applications. Which comes first? It can be both. You can explore, experiment, map out policies and practices and then iterate with experience.

At IBM, Bhandari has launched a series of projects he calls “catalysts” aimed at accelerating both the enterprise data strategy and IBM’s advance into cognitive computing. These projects are not isolated or independent. They serve as strategic proof points of progress for the whole company.

One catalyst project applies cognitive methods to improve IBM health benefits programs for employees and retirees. The cognitive solution will capture and analyze a wealth of information about the health of its people and the performance of health programs, allowing IBM to discover which programs are working best and modify those that aren’t preforming up to expectations.

Once the CDO’s team has built and refined a platform using these catalysts as learning experiences, the platform can be opened up to people throughout IBM, helping them accelerate their own projects. At that point it will be a true enterprise-wide cognitive platform.