Formulate Essential Data Governance Practices

The creation of a  Data Governance function at your organization is a critical success factor in implementing Master Data Management.  Just like any machine on a factory floor, Master Data is an Asset.  An Asset implies ownership, maintenance and value creation: so too with Master Data.  To borrow an analogy from the Manufacturing world, Transactional data is the Widget, and Master data is one of the machines that makes the Widget.  It is part of your organization’s Value Chain.

Unfortunately, firms starting on the road to MDM fall peril to one of two pitfalls on either extreme of the Data Governance mandate.  The first pitfall is to treat MDM as a one-time project, not a program.  Projects have an end-date, Programs are continual.  Have you ever heard of an Asset Maintenance Project?  That’s a recipe for crisis.  Firms which maintain their assets as a Program do far better.

The second pitfall is ask the Data Governance team to do too much, too fast.  Governance cannot do much without some asset to govern.  Have you ever heard of a Machinery Maintenance Program instituted before you figured out what type of machinery you needed, what the output requirements were, or before you made the capital purchase?  I haven’t either.  First, you acquire the capital.  You do so with the expectation that you will maintain it.  Then you put it into production. Then you formulate the maintenance schedule and execute that plan.

In order to successfully stand up a Data Governance function for your Master Data Program, you’ll need to understand these essential roles in Data Governance: Executive Steering, Data Owners, Data Stewards. 

Follow these Do’s and Don’ts:

Do establish an Executive Steering Committee for all Master Data practices in your enterprise, focused upon strategic requirements, metrics and accountability.

Do establish Data Quality Metrics.   Tie them to strategic drivers of the business. Review them regularly.  Your MDM toolset should provide analytics or dashboards to provide this view.

Don’t ask the Steering Committee to own the data model or processes – that is the Data Ownership role.

Do establish a Data Ownership group for each domain of Master Data.  Ownership teams are typically cross-functional, not simply the AR manager for Customer Master Data, or the HR manager for Employee Master Data.  As you evolve down the maturity path, you will find that master data has a broad set of stakeholders – Do be ready to be inclusive.

Do establish regular status meetings where Data Ownership meets with the Executive Steering Committee to review priorities and issues.

Don’t require that Data Owners “handle the data”.  That is the Data Stewardship role.

Do formalize a Data Stewardship team for each domain of Master Data.  Data Stewards are “data people” – business people who live in the data, but with no technical skills required, per se (though technology people can contribute to a Data Stewardship team).

Don’t restrict Data Stewards to just  the people who report to the Data Owner – think cross-functional!

Do anticipate conflicts – Data Owners should have some political skills.  The reality is that Master Data is valuable to a broad set of constituencies within an enterprise.  Be practical as it relates to one faction’s “Wish List” and keep moving the ball forward.

Without a Data Governance function, MDM tends to be a one-time project (“Clean it Once”) and fails to deliver real value.  Without a clear vision of how Data Governance support MDM, it can hold things up. A rational Data Governance function does not need to hold up the execution of a Master Data project – it supports it. Keep Data Governance strategic, cross functional, and flexible. Then, let the MDM technology team deliver the tools.

Face API and Power BI

At last week’s Build2015 developer conference, Microsoft demonstrated many great new tools. One demo which got quite a bit of attention was the How Old Am I? app (http://how-old.net) The demo allows users to upload pictures and let the service “guess” the age and gender of the individuals in the photo. Within a few hours, the demo went viral, with over 210,000 images uploaded to the site from all over the world. The result was a dashboard of requests from all over the globe.

Power BI

This solution shows off the use of a number of powerful technologies.

Face APIProject Oxford is a set of Artificial Intelligence API’s and REST services which developers can use today to build Intelligent Systems. In addition to Facial Recognition, the Project Oxford AI services include Speech Recognition, Vision (or Image Recognition and OCR), and Language Understanding Intelligent Services – leveraging the technology capabilities of Bing and Cortana.

Azure Event Hubs –  a highly scalable publish-subscribe ingestor that can intake millions of events per second, the Event Hubs API is used to stream the JSON document from the web page when the user uploads a picture.

Azure Stream Analytics – a fully managed low latency high throughput stream processing solution. Azure Stream Analytics lets you write your stream processing logic in a very simple SQL -like language.   This allows the solution to take measurements every 10 seconds of how many requests, from which countries, of which gender and age.  These measurements become Facts for your analysis.

Power BI – choose PowerBI as the output of our stream analytics job (click here to learn how). Then the team went to http://www.powerbi.com, and selected the dataset and table created by Azure Stream Analytics. There is no additional coding needed to create real time dashboards.

The only down side to this is that my worst fears have been confirmed – I look older than I actually am by over 10 years! :(
How old do I look?!?!

The Business Value of Microsoft Azure – Part 5 – Notification Hubs

This article is part 5 of a series of articles that focus on the Business Value of Microsoft Azure. Microsoft Azure provides a variety of cloud based technologies that can enable organizations in a number of ways. Rather than focusing on the technical aspects of Microsoft Azure (there’s plenty of that content out there) this series will focus on business situations and how Microsoft Azure services can benefit.

In our last article we focused on virtualization and the use of virtual machines as part of an Infrastructure as a Service (IaaS) solution. While this is a great approach for traditional server workloads, there has been a significant shift in the way individuals interact with and consume information suggesting the need for something different. Specifically, a mobile device has overtaken the PC in terms of unit sales/year and this presents a scenario that many municipalities can tap into.

Let’s think back to our fictional town of Gamehendge. A hurricane is approaching and Mayor Wilson needs to warn its citizens. To handle the scale required to communicate in this fashion would require a significant notification infrastructure. Why pay for this type of scale when it’s only needed on occasion? Microsoft Azure Notification Hubs is a massively scalable mobile push notification engine for quickly sending millions of messages to iOS, Android, Windows, or Kindle devices. It’s possible to tailor notifications to specific citizens or entire groups with just a few lines of code, and do it across any platform.

Further, in Gamehendge there is a population that doesn’t speak English as their native language. Traditional communications can often go without understanding. The templates feature of Notification Hubs provide a handy way to send localized push notifications so you’re speaking to citizens in their language. Templates also eliminates the hassle of storing the localization settings for each group.

Combining the scalability and configurability of the Notification Hubs solution, along with its ability to work with either on-premise or cloud based systems, your municipality gains the ability to notify your citizens of any information that can prepare and inform them of upcoming events in the event of an emergency or as part of a more generalized community awareness system. While the Notification Hubs feature is just one small component of the Azure platform, it can have a significant impact in your community.

As a partner with BlumShapiro Consulting, Michael Pelletier leads our Technology Consulting Practice. He consults with a range of businesses and industries on issues related to technology strategy and direction, enterprise and solution architecture, service oriented architecture and solution delivery.

 

First Accommodate Master Data, Then Clean It

In this blog post, I want to challenge a deeply held notion of Data Quality and Master Data Management.

I have had many, many conversations with technology professionals seeking to implement MDM in their organization. In those first conversations, among the first questions asked is a complex one, disguised as a simple one – How can I start with clean data?

Listen: if you try to start your Master Data implementation with clean master data – you will never get started!

Instead you need to embrace two fundamental realities of Master Data Management. First, there is no clear authoritative source for your master data (if there were, you wouldn’t have a problem). Second: Data Quality is “Front of House” work. The IT department may have data integration, data profiling, third party reference data and matching algorithms in their toolbox, but they can only do so much. IT tools are Back Office Tools and the IT data cleanups happen in the shadows. When they get it wrong, they get it very wrong and comprehensively wrong (and the explanation is hard to understand).

This sequence of events is straightforward, enables the business to take ownership and provides a clear path to getting started.

  • Accommodate your Data – in order for business people to steward and govern their own data – they need to see it with their own eyes, and they need to see all of it, even the data they don’t like. In order to do this, you must:
    • Maintain a clear relationship between data in the MDM hub and its source – don’t attempt to reduce the volume of records. The Federated approach to MDM does this best.
    • Keep rationalization/mapping to a minimum – avoid cleaning the data as you load it. Its wasteful to do it in ETL code when your MDM toolset is ready to do it for you much more easily.
    • Take a “Come as You Are” approach – avoid placing restrictions on the data at this stage of the project, because this only serves to keep data out of your system. We want the data in.
  • Establish Governance of your Data – once you have all of the data loaded into a Federated data model, you have the opportunity to start addressing the gaps
    • First, take some baseline measurements. How incomplete is your data?
    • Next, begin developing rules which can be enforced in the MDM Hub. These rules should be comprehensible to a business user. Ideally, your toolset integrates the rules into the stewardship experience, so that rules declared in the hub are readily available to them. Once you have a healthy body of rules, validate the data and take another baseline measurement
    • Now your data stewardship team can get to work, and you’ll have real metrics to share with the business with regards to the progress you are making towards data compliance.
  • Improve your Data – MDM toolsets automate the process of improving master data sourced from different systems. They do this in three ways:
    • Standardize your Data – MDM tools help data stewards establish standards and check data against those standards
    • Match your Data – MDM tools help data stewards find similar records from multiple systems and establish a grouping of clusters of records. The Group becomes the “Golden Record” – none of the sources get to be the boss!
    • Harmonize your Data – MDM tools help data stewards make decisions about which sources are most authoritative and can automate the harmonization of data within a grouping

Organizations whose starting approach with MDM is “Get the data clean and Keep the data clean” often fail to even get started. Or worse, they spend a lot of time and money requiring IT to clean the data, and then abandon the project after 6 months with nothing to show for it. Clean, then Load is the wrong order: Flip the Script and stick to these principles.

  1. Design a Federated MDM Data model which simplifies identity management for the master data.
  2. Identify where your master data lives and understand the attributes you want to govern initially.
  3. Bring the master data in as it exists in the source systems.
  4. Remove restrictions to loading your data.
  5. Establish some baseline measurements.
  6. Devise your initial rules set.
  7. Use MDM Stewardship tools to automate standardizing, matching and harmonizing.