“The powers not delegated to the United States by the Constitution, nor prohibited by it to the States, are reserved to the States respectively, or to the people.” – Tenth Amendment, United States Constitution
A large insurance conglomerate has been doing data governance for many years with a lot to show for it. They instituted data quality policies for financial data coming from various divisions into the corporate finance department. Because of this, they were able to improve the efficiency of the quarterly close process. As a result of their data governance efforts, this company now can close in under 10 days rather than an average of 27.
However, while they’ve been successful in some areas, they have struggled to make progress in others. They struggled with common definitions of KPIs, for example. (Data definitions are a class of data policies, along with policies for data quality, security and lifecycle.) During a workshop with Kalido, in a Eureka moment, the data governance lead suddenly understood why they had struggled: they wanted all their rules and policies to be enterprise-wide. But for some data domains and some business areas, enterprise level data policies are neither achievable, nor beneficial.
In political governance, federalism resolves this problem. Certain laws, such as those for the armed forces, currency, and immigration, need to be set by the central government. Other things, such as building and zoning laws and traffic regulations, are better handled locally. Local governments have autonomy to make their own laws as long as they don’t violate or contradict federal laws. The tenth amendment in the US Constitution explicitly declares this principle.
Another way to think about it is, every law needs to have a clearly defined scope: international, national, province, state, town, etc. When every single law is dictated by a central government, we have the Soviet Union, which was inefficient and unstable. It collapsed spectacularly.
Taking the same approach to data governance, every rule and policy for data needs a clearly defined scope. Is it enterprise wide? Or only covering the finance function in EMEA? Is it applicable to a single line of business? In our insurance company example, many KPIs are rightfully defined at the line of business level: earned premiums are calculated differently for life insurance and property and casualty insurance.
Data governance organizations should avoid the temptation to set scopes bigger than they need to be. Not creating policies for some areas should be a deliberate act.
Data is different from technology. For server farms, email, telecommunication, economy of scale matters. Centralization is the way to go. But data management is more political than technological. Here, the tyranny of centralization actually leads to inefficiency and revolt in the form of shadow IT.
We found out the hard way that for a large enterprise, managing every piece of data in a single ERP system and a single enterprise data warehouse is neither achievable, nor beneficial. We also know that anarchy without some central control is a recipe for disaster. Companies need to strike the right balance between centralization and decentralization, between central control and local autonomy. And this is where federalism comes in. It is the idea that the central authority should claim control over a limited set of things with clearly defined boundaries. The rest is up to the local authorities. This principle — which is in effect a governance model — is applicable to many disciplines in data management: MDM, data warehousing and BI, and data policy management.
Beware of promises of “a single of version of the truth,” which is a slogan for selling big and expensive projects. Instead, look for technology that gives you agility, that gives you a movable boundary between centralization and decentralization. Look for technology that gives you the ability to break a problem down to manageable chunks and solve them at the right places in the organization rather than solving everything centrally, but still give you control.
The blogs in this series are:
- Introduction to federalism as a data governance model: US Constitution and Data Governance
- Federalism for master data management: Managing Master Data Using Federalist Principles
- Federalism for BI and data warehousing: Can You Handle a Single Version of the Truth?
- Federalism for data rules and policies: How to Set Scope for Data Rules and Policies?