A recent post on Information Management makes the point that master data management should not be run as an IT project.
From an IT perspective, master data management is most typically treated as a data integration problem. The focus is often on how master data can be consolidated and synchronised across various operatioanl systems.
In many cases, this approach fails to deliver business value.
The article identifies 10 best practices for MDM implementation including:
- Establish a business case. The business case should be tool agnostic and look at the use cases and potential benefit that will be delivered. The business case can then also be used to determine critical capabilities to be delivered by any proposed tool
- Get executive sponsorship. As with any enterprise project, the correct level of sponsorship is key. Sponsorship must go beyond providing budget – key business stakeholders must be involved in the MDM SteerCo that will drive decision making and resolve interdepartmental conflicts
- Get business involved. Many of the day to day decisions require business knowledge and sign off. For example, a client may have different telephone numbers in the Sales and the Orders systems. Business must be involved in determining which number is most likely to be correct. Without this active involvement IT run the risk of losing critical information.
- Invest sufficient time in planning and evaluation. A common mistake is to rush purchase of a tool. A clear understanding of the business case ( and supporting use cases), an understanding of the current state of master data, and fit to the existing technology stack should all be considered before selecting a tool.
- Institute MDM governance and stewardship. MDM governance must consider much more than the process for merging or approving duplicates. MDM governance must consider data mappings, decisions around which systems are most trusted (at an attribute level), data quality standards, match rules and much more. The MDM Governance team also brings business and IT stakeholders together to ensure sponsorship and business engagement
- Adopt the right technology and architecture. MDM can incorporate elements of batch integration, real time integration and even integration to “new” data sources such as big data elements.Decisions on how data will be synchronised back to source must be well governed and tested, and may accommodate one or many of these integration strategies.
- Define the data quality strategy. A common mistake is to assume that master data management will deliver quality master data. Instead, poorly planned and delivered MDM can actually make data quality worse. Data standardisation, enrichment, scrubbing and matching strategies must be in place if MDM is to deliver value. Data quality requirements should also be considered when selecting an MDM tool or platform.
- Get the right team. MDM and data quality management are niche competencies and you will probably require specialist help. However, we suggest that you cosuorce your MDM team – including stakeholders from your business, your IT team and any vendors or SI. This helps to drive business buy in and skills transfer, assuming that your team put in the time.
- Adopt a phased approach. MDM is a massive program and can run into years of effort. Break the program down into small steps that deliver incremental value and ensure that your Steering / Governance Committee are actively involved in resolving issues, and in prioritizing each new month’s effort to deliver maximum value based on changing business priorities
- Deliver and communicate incremental value. Each phase, identified in 9 above, should be linked ot the value generate. For example, if telemarketing campaigns are a high priority, but contactability is poor, then a focus may be to consolidate alternative telephone numbers for each client and present these as a telesales list. As the next step, one may add email addresses, and so on. The value delivered should be communicated to other stakeholders, who may have their own, similar priorities that will feature in a later phase.
What do you think? What best practices would you add?