“No technology will give you that; you need governance.”
I’ve been hearing this expression for the past few months from vendors who offer data integration products and services to support an enterprise data warehouse (EDW) and metadata management. They utter the expression when we ask how their offering meets one of our requirements: at the “landing level” of the EDW, where the cleansed data truths will live, we want the ability to place the values of institutionally-sanctioned derived terms. They respond as if we want technology to force definitional agreement, but we don’t. We want the technology to help instantiate agreements that we’ll make socially. It is a way in which we expect governance and technology to dance together to achieve what I like to call “Analytic Public Goods.”
Like traffic rules and stop lights, public goods create order and security in shared ecosystems, where many people independently strive to maximize their individual gains. Our current ecosystem for data management, business intelligence (BI) and analytics emerged from our cultural history: This deeply data-driven organization long has demanded rapid delivery of actionable information. Collective demand emanating from multiple sources routinely overwhelmed the institutionalized infrastructure, which was at an early stage of development when the demand was unleashed. Unable to wait for a more advanced infrastructure, individual information consumers began hiring their own data and analytic staff, who found ways to access source data and manipulate them independently, even though many of the independent needs were conceptually the same.
The cumulative effect of these independent actions now resembles a city that grew very fast, teeming with inhabitants who all purchased private cars when they couldn’t get seats on the public bus, which didn’t have routes to their destinations anyway. But that wasn’t a solution either, because the infrastructure also lacked the public goods of stop lights and traffic rules. In their absence, drivers trying as hard as they could to reach destinations as quickly as possible routinely collided with each other, ensnarled themselves in traffic jams, or took wide detours to avoid collisions and traffic jams. The resulting mess made clear the need for public goods, which in turn require governance.
Economists define “public goods” as assets that are non-excludable, meaning that access to and/or use of the good cannot be controlled. Unlike “common goods”, public goods are also non-rivalrous, which means one person’s consumption of the good does not diminish the quality or quantity available for another to consume. Indeed, for some public goods, such as traffic rules, the quality or quantity may well be enhanced as more people use them in order to maximize private gain. Economists also know that governments are the only entities likely to create, implement, and maintain public goods. Because access cannot be controlled, and because benefit accrues to all, there is little incentive for private entities to invest in them.
Thus in its broadest sense, the purpose of governance in an enterprise strategy around data management and BI is to ensure the existence of analytic public goods that all stakeholders require to meet information needs securely and efficiently.
The analytic public good that launched this blog is a set of institutionally operationalized definitions for commonly used organizational terms. “Show me vacancy rate, readmission rate, average census, bed occupancy rate, ventilator-days, by nursing unit.” The questions could come from anywhere, and in a distributed staff environment, could go to anyone. Each answer requires an operational definition for “nursing unit,” “vacancy rate,” and every other term. Each staff person receiving the question will struggle to perform the same definitional work independently, not sure if the definition is “correct,” duplicating the effort performed by other staff, who produce conflicting numbers, undermining consumer confidence. Traffic snarls and collisions abound.
Our enterprise strategy includes a commitment to create, implement, and maintain this public good, and we will need both governance and technology to bring it about.
To create the definitions, we’ll require governance to convene stakeholders, and manage the social process of reaching agreement. We plan to use the technology as a rationale to convene the stakeholders, and if it is sufficiently developed by the time the convention takes place, we also plan to use it by quickly demonstrating the quantitative consequences of the choices they make.
To implement the definitions, we’ll require governance to communicate awareness that they exist, and to resolve disagreements when customers ask for deviations. Technology also plays a role, but a different one. If we place institutionally sanctioned terms in the “truth layer” of the EDW, which by then has become the institutional source of data, staff will use accepted definitions by default. Taking a page from the Cass Sunstein playbook, we’ll have used technology to nudge the staff to adopt behaviors that stakeholders have agreed are in their best individual interests, and in the best interests of the organization as a whole.
To maintain the definitions, we’ll require governance to develop criteria for when it’s permissible for a definition to change, or for new institutionalized terms to be added to the collective dictionary. Once again, we’ll be relying on a technological assist: We want to be able to search the dictionary easily, to identify potential duplicates; we want the technology to track the status of derived terms as they progress through approval processes.
This is our vision of the choreographed dance between governance and technology. So when we ask the vendors how we can place derived terms in the platform, we’re not asking that technology perform a governance task. We’re asking how the technology can support governance, as we create, implement and maintain these analytic public goods which are so necessary for providing order and security in the common information ecosystem.