When you search on Google for “Master Data Management” you’ll have to look hard for a search result that is not vendor related. Most search results will be from software vendors about tooling related to MDM. Master Data Management seems to become a synonym for the implementation of a master data hub. On a discussion forum, I recently read a question on whether anybody had implemented MDM without an extensive/expensive toolset.
I’ll be the first to admit that tools can help a great deal. There is, however, a lot that can be accomplished without tooling. Indeed, if businesses actually sorted out the non-tooling element of master data management first, they may get a very different result in their tooling business case. However, assuming that you are gong to get tooling, your preparation work will be the foundation for your MDM efforts. This foundation work will help you to:
- get to know your data;
- understand the usage of your data;
- see the shortcomings of your data;
- spot the opportunities for improvement.
When you perform these steps without tooling you will focus on the data and not on the requirements of the tool chosen. You’ll also know much more precisely what the tooling should be capable of, once you finish this exercise. Indeed, you may find out that ou didn’t need that shiny new tool after all …
A data dictionary is definitely the basis for your master data efforts. Your data dictionary is the one stop source for all your data related questions. What is the meaning of this field? Which departments are using this information? Which values are allowed for this field? The data dictionary collects the answers to these and many more questions in one place. The content of a data dictionary most often has a technical and business section. The business section contains information like ownership, description, usage, etc. The technical part contains field name, field type, field length, default value and much more.
Creating your (enterprise) data dictionary is not something that is done in a week. It often combines pieces of (contradicting) content coming from different sources. You will be surprised by how much debate the definition for a simple field like “street” will cause. Multiply this by the number of systems and or data objects, and you have a serious piece of work. You shouldn’t underestimate it’s effort, but nor should you underestimate it’s importance.
Your business rules and corresponding data quality rules are important means for getting the data into shape.
A business rule is a condition in your data that is either true or false. An easy business rule could be: “The address of a customer should always include the country of residence”. A more complex example of a business rule might be that a contract can only contain a general discount of 10 percent if the customer purchases for more than 1.000 euros monthly. Document your business rules and make sure they can be related to the data dictionary. Relating the business rules back to the data dictionary allows you at a later stage to calculate the coverage each data object has.
Once you have your business rules down, you can start to measure the quality of your data via your data quality rules. A data quality rule is the physical implementation of a business rules against an application. A single business rule can thus result in multiple data quality rules if that business rules needs to be measured in multiple applications. Maintaining your data quality is a never ending process composing of:
- Measurement of your data against the business rules;
- Visualising the results of the measurements in reports and dashboards;
- Cleanse the data that is not correct according to the data quality rules;
- Repeat the steps above.
Be sure that the business rules can be measured and implemented in the form of data quality rules. Measure these data quality rules on a regular basis and visualise the improvement over time. Visualisation is the most important instrument you have to get your data into shape by providing the business with actionable reports that clearly indicate the issue with the data and make it clear what the solution should be.
Another easily actionable step possible without tooling is the documentation of your data management processes. Write down and share your processes for adding, changing, updating and deleting master data. Whilst, business rules and data quality reports will help you to get the data back into shape. Clear processes and controls will prevent your data from getting out of shape in the first place.
Processes are important for quick and correct data entry. Who in the company can manipulate the master data? What input is needed from which department when adding or changing data? Who needs to approve the input given? What documents need to be added to a request for either record keeping or regulations?
Create a (visual) process flow with the individual steps, the input needed for a step and the expected outcome of that step. Make sure that everybody involved in the process will know what is expected and when. (A common mistake is to document processes and then file the documentation away, never to be seen again). As with many MDM activities, it is a lot of set-up work, but which pays handsome dividends later …
Tooling or Data?
Once you have the above aspects covered, by all means, do look at Master Data Management, Governance and Data Quality tooling to support your initiative. However, the order is important here. Your first goal is to understand your data in order to be able to improve it. Getting your data dictionary, business rules and processes in place is the foundation for future activities. Once the foundation is ready, there is nothing stopping you from building your dream house…