John Parkinson’s post on “The Importance of Meaning” lists a number of pertinent examples of how loosely defined meanings can impact business operations or decision-making.

• Critical decisions will be made on data fields. What is “claim value?” Does that include a deduction for excess? Or not?
• Does the current mortgage balance include accrued interest? What is the difference between capital balance and current balance? How is it calculated?
• What is the mortgage start date? Is it when the monies are provided to the solicitor or the completion date?

Metadata provides context to data

In the real world questions like those above are heavily dependent on context.

So we may add to these questions:

• Does claim value differ for different types of claims? What different types of claims are there?
• If we add a new mortgage product how will product management and IT engage to ensure that the relevant rules and definitions are correctly defined and reported upon?
• Are there existing business or external (legal policies that limit the interest rate, or the term, or any other variable related to a new loan? Who are the internal experts that must be consulted before deploying any new product to market?
• Can we provide internal and external auditors with all the information they need to be confident that we are following due process when defining a new product, or to prove that they can trust our report’s accuracy and consistency?

The business impact of bad metadata

“One organisation received a query from the external auditors as to why the results from two systems for the same measure were different. 2,000 man hours later (or one FTE for a year) the reason for the discrepancy was identified.”

John Parkinson

A business glossary, data dictionary, enterprise data model and other forms of metadata are important tools for establishing meaning but they are not enough.

To create meaning these repositories must be shared and maintained or risk the fate of any dictionary – they will not be read.

The data governance processes need to be established to ensure that ALL the relevant parties are consulted and their inputs are captured. Meaning needs to be defined across the business and mapped to what systems actually do.

Systems must cater for differences too.

The same term may (legitimately) mean different things in the retail and wholesale spaces.

Different terms may mean the same thing – is a customer also a client? Is a lead also a customer?

One function of data governance is to create business alignment between information and meaning.

• What systems rely on interest rates?
• What is the process for updating interest rates (for each system)?
• What reports use the term? Are there different calculations that use it?
• How do we create an aggregated view of profit on loans when we have varying profitability per product, per branch or per region?
• What regulations or laws regulate the use of interest?

Data governance must ensure not only that critical terms are defined, but that they are defined by the right people. They must be captured in the correct context and they must be kept up to date. The processes to do all this must be clearly defined and tracked to ensure that agreed service levels are met and to identify and manage issues.

This needs to be a community effort – not restricted to a small number of metadata tool specialists.

IT maintenance of technical metadata is important, but by itself, it is not governance.

Governance means business accountability and the use of these assets.