What Is Master Data Management?

Master data management (MDM) is a set of tools and processes used to define and manage an organization’s critical data and provide data integration to a single point of reference, also known as the Golden Record.

In many businesses today, especially global enterprises with hundreds of separate applications and systems, there is a pressing need to bring accurate analytics insight. To do that, the organization needs to consolidate, cleanse, and categorized datasets from internal data sources, such as CRM and ERP systems, and external reference data aggregators and third-party datasets.

This is the promise of MDM systems: they promise to create a single, trusted view of entities such as suppliers, customers, and products to support analytics and business decision-making.

The purpose of MDM: a single version of the truth

Organizations have been using MDM solutions for decades to ensure that data for the organization’s entities operate around a single, authoritative version of the truth, regardless of where the data comes from. Curated, de-deduplicated, and enriched “mastered data” is imperative for organizations to leverage their data assets with accuracy, confidence, and impact.

Mastered data can answer many important questions such as:

• Who are our customers, and how can we better target them?
• Which products are our most popular and profitable?
• How can we get a holistic view of service consumption for more informed decisions?

Enterprises might also want to know:

• How can we reduce procurement costs?
• How can we enable smarter reporting, analytics, and monitoring/tracking, even if the information comes from many different sources?
• How can we eliminate manual, burdensome data entry and improve operational efficiency and strategic planning?

Answers to these questions are critical across a range of industries, from retail to manufacturing, financial services, and government. The downsides of not having mastered data are substantial. Most organizations see all the symptoms of poor master data in their critical business processes, whether in the form of delayed product launches, high supply chain costs, frequent customer complaints, or hefty regulatory penalties.

Who needs to be involved in MDM work?

All of this makes MDM a worthy goal, but achieving it is complex. MDM requires tools, technology, and processes to ensure that accurate, consistent, and complete master data is provided across the enterprise and to business partners. MDM also involves putting together the right people and talents. The three core roles involved in successful MDM include:

• Data governance individuals who help data stewards understand how data should be managed and then hold them accountable for following the requirements.
• Administrators who set up and configure the solution.
• Data stewards who fix, clean, and manage the data directly within the MDM tool.

Other MDM roles might include program managers, project managers, system administrators, developers, business analysts, and others, depending on the organization and project type. In addition to these hands-on roles, it is essential to engage stakeholders, such as subject-matter experts and business/IT executives.

What are the limitations of MDM solutions?

However, one of the biggest challenges with MDM is its inability to scale in today’s complex, increasingly large data environments. Data mastering involves producing clusters of records thought to represent the same entity. Once clusters are created, the next step is to construct a single “golden record,” representing each cluster. This involves finding matches. The traditional MDM solution to finding matches is to write a collection of rules in a proprietary rule language.

The rules-based approach used by traditional MDM solutions such as Informatica, IBM, Oracle, and some of the home-grown rule-based MDM approaches quickly becomes a limitation. The primary driver of this challenge is the dependency on rule-based engines, which require data experts, developers, and business teams to collaborate on coding complex formulas that capture constantly changing business logic. Whenever new data or data updates are introduced to the solution, long cycles of iterations are needed among IT, data, and business teams to refine the MDM rules logic.

The key benefits of cloud-native machine-learning data mastering

1. Accelerating analytic insights 

Our studies have shown 90%+ accuracy mastering data with Tamr technology, compared to only 50-80% accuracy with rules-based models. Tamr’s output is clean, consolidated data that can then be used to power visualization tools such as PowerBI, Qlik, Tableau, and Thoughtspot.

2. Boosting operational efficiency

Tamr Machine Learning technology is coupled with expert feedback to reduce the manual workflows required to integrate new datasets by up to 90%, requiring minimal hours of work from data stewards compared to months for IT-driven projects to adjust rules in MDM systems. The result is that data teams are better able to focus on higher-value business initiatives instead of manual data prep workflows.

3. Lowering the total cost of ownership for data projects and enhancing data operations with cloud-native data mastering

Tamr runs natively on leading cloud providers (AWS, GCP, and Azure), enabling organizations to effectively connect large volumes of highly variable internal and external data. Combined with a robust set of APIs, organizations can integrate Tamr into new and legacy data pipelines. Our growing partnerships with leading cloud providers and technology partners have positioned Tamr as a key part of the modern, best-of-breed data management stack essential for scaling data operations and lowering the total cost of ownership for data mastering projects. And with seamless data publishing capabilities, teams can trust the completeness and readiness of analytics and data initiatives powered by Tamr, further accelerating business outcomes.

The original article by Sohaiyla Khalili, the previous product marketing manager at Tamr, is here.

The views and opinions expressed in this article are those of the author and do not necessarily reflect those of CDOTrends. Image credit: iStockphoto/gorodenkoff