Modelling for Master Data Management: Expert interview with John Owens
Master Data Management (MDM) has rapidly become one of the most in-demand skills within the data management industry.
One of the core skills that practitioners require to deliver MDM is the ability to construct and manage a variety of data and functional models.
To help members understand the techniques involved I interviewed international modelling expert John Owens of Integrated Modeling Method.
Dylan Jones: Master Data Management (MDM) as a discipline and technology marketplace has really exploded to prominence in recent years but you’ve got some strong views on where many practitioners and organisations are going wrong. Would you care to share your views on where companies make mistakes?
John Owens: Wow. A big question! There are quite a few major errors which I’ll list and then cover in detail if you like.
The very first error is calling it Master Data Management, as it is not this. I’ll explain why I say this later.
Enterprises look in the worst possible place for Master Entities – in their existing data. They can’t really be found there.
Because they start in the wrong place, they then use totally inappropriate techniques to try map what they think are Master Entities. This compounds their errors.
Too many enterprises suffer from the ‘Curse of the Systems Silo’ due to inappropriate implementation of the off the shelf software, that causes fatal Master Entity fragmentation.
Data Quality and MDM Practitioners try to manage data in isolation from Function – a fatal error.
Too few MDM practitioners are familiar with the most powerful modelling tool for Master Entities, the Logical Data Model or LDM. This makes it almost impossible to implement any quality MEM solutions.
Because they do not make use of the LDM, practitioners are unable to define and model the Unique Identifiers of the Master Entities – an essential step in effective Master Entity Management.
Again, because they do not make use of the LDM, most enterprises replicate the key Master Entity of Party on numerous systems across the business, calling it by different names and holding different data for it in different structures in each system.
Dylan Jones: Ok, let’s tackle that first point. Why is it wrong to call it Master Data Management?
John Owens: The problem is that this essential set of activities is all about identifying and managing the Master Entities of the enterprise. For that reason it should be called Master Entity Management or MEM.
Dylan Jones: Isn’t this just splitting hairs John?
John Owens: No, by calling it by the wrong name enterprises are trying to manage the wrong thing. You can manage all of the Master Data in the world and still not be managing the Master Entities.The only thing that you achieve by managing data is managing data, which on its own is of no benefit to any enterprise.
Dylan Jones: What are the impacts of starting in the wrong place with MDM?
John Owens: Most enterprises look in the worst possible place when trying to identify their Master Entities, which is in their existing data. I have likened this to looking in a scrapyard for the parts to build a Formula 1 racing car.
Looking in existing data simply suggests what the Master Entities might be based on the data that the enterprise has created to date. However, unless Master Entity Management is well established in an enterprise, then the data relating to Master Entities in existing data is about as far from where it ought to be as it is possible to get.
The only way to establish what the Master Entities of an enterprise ought to be is to ask the senior executives what these ought to be.
This shocks most practitioners. They say, “Why bother the busy Senior Executives? Surely there are lots of other people in the enterprise who can tell you this?”
Well the fact is that there are not. Other people in the enterprise might have opinions, in fact you can be sure that they have, however, only the Senior Executive team can say WHAT it is that the enterprise OUGHT to be doing. Only they can answer the six Multi Dimensional MDM questions.
The answers to these six questions lay the foundation for Master Entity Management in the enterprise.
Because they start in the wrong place, they then model the wrong thing using the wrong techniques. Because they are erroneously starting with existing data, many enterprises and practitioners are actually forced to use normalisation techniques that were outmoded twenty years ago. This simply results in normalised nonsense.
Dylan Jones: This curse of the silo, we obviously see this all the time with poor data quality but how does it impact MDM?
John Owens: Because they have implemented so many off-the-shelf software packages in standalone mode, enterprises have created a set of silos that has resulted in total fragmentation of core enterprise information, including that related to Master Entities.
This fragmentation is worst when it comes to master entities as, not only are they held in different forms in different systems, they are even called by different names in each system.
Dylan Jones: Ok, I understand this but the problem remains for many companies that they’re saddled with these systems, they often cost millions and can’t be replaced overnight. For companies that can’t scrap their systems and start again but do need to resolve the challenges of MEM, how can your approach to modelling help them?
John Owens: By using the Multi-Dimensional MEM approach the enterprise will know, perhaps for the first time, what the true Master Entities of the enterprise ought to be.
Because they will have built Logical Data Models of these Master Entities, they will know exactly what their data content and structures ought to be.
If they have existing standalone systems this enables them to a) do gap analysis on all existing systems to see how well or badly they support these structures and b) design and build central hubs that can effectively support all of the satellite systems from a single point that holds the one true definition of each Master Entity.
If they have no existing systems, the logical data models are a very powerful tool that will enable them to pick appropriate third party systems and prevent them selecting those with critical structural errors.
Because Data Quality and MDM practitioners look at data in isolation from Function they get completely lost. They imagine that data has some intrinsic value and that by attempting to manage what they define as its quality that it will benefit the enterprise. This view is fatally flawed as it totally misses out the purpose of data, which is to support the execution of the Business Functions.
Knowing the Functions defines the Purpose of the data and enables ‘Fitness for Purpose’ to be very clearly understood and and defined.
Dylan Jones: You talked about the “Six Multi Dimensional MDM questions” earlier – what are they so that people can get started on their MEM journey?
John Owens: Sure, they are:
What Types of things do we make, buy, sell, improve or trade?
Who do we buy from and sell to?
Who benefits from our products and services?
Where are our customers, suppliers, beneficiaries, etc. located?
Where do we deliver our products and services?
What resources do we use to make, buy, sell, improve or trade our products and services?
Having asked the Senior Executive team all of the Multi Dimensional MDM questions, you will get answers that can be grouped into the following categories:
Products
Parties
Locations
Assets
These are the major categories of Master Entities of every enterprises.
Dylan Jones: Ok, but obviously these structures will look very different depending your own business model?
John Owens: Absolutely. It really depends on what industry the enterprise is operating in. For example, the Product structures of the Master Entities for a supermarket will be very different to those of an oil and gas exploration company. ‘Product’ refers to anything that can be bought, sold, improved or traded and can include any or all of the following:
Raw Materials
Component Parts
Wholesale Products
Retails Products
Marketing Products
Services
Dylan Jones: So many types of product! Would you really need all these Product Categories in your enterprise?
John Owens: The only way to know is to look at the Business Functions. Is there a Business Function (and by Business Function I mean a core business activity and not a business department) that requires these categories to be differentiated? One of the most effective questions to ask in order to be able to establish this is: “What products do we need to measure the comparative performance of?”
Answers such as:
“We need to know what our most profitable retail or wholesale products are”
OR “We need to know our most widely demanded services.”
OR “We need to know what it costs to market a product.”
Will tell you exactly how you need to structure and segment your Products.
The question, “What do we need to measure the comparative performance of?” is really the seventh Multi Dimensional MDM question, as it is a major means of determining how each of our Master Data Entities need to be categorised or segmented.
Remember, Function Defines All. If data is not used in the execution of at least one Business Function in the enterprise, then it is not needed and should not be held.
So, if a Product Category is not used in at least one Business Function, then Products should not be categorised in that way.
Dylan Jones: Thanks John
About John Owens
John Owens of Integrated Modeling Method. is known as a Thought Leader, Consultant, Mentor, Practitioner and Writer in the worlds of Strategic Requirements, Business Process and Data Modelling, Data Quality and Master Data Management.
He has built an international reputation as a highly innovative specialist in these areas and has worked in and led multi-million dollar projects in a wide range of industries across the UK, Ireland, Europe and New Zealand.
Both as a consultant and hands-on practitioner, he has a gift for identifying the underlying simplicity in any business, even when shrouded in complexity, and bringing it to the surface.
John is renowned for his ability to train, coach and mentor both technical and non-technical people at all levels in the enterprise in all areas of Project Management, Business and Data Modelling from strategy definition, through hands-on analysis and modelling, to quality design and implementation.
His methods and insights remove the mystique from these subjects and get people up to a high level of proficiency in a short space of time.