Interview with Len Silverston Focusing on New Volume 3 "Universal Patterns"

By Robert Seiner
April 1, 2009

RSS: Mr. Silverston. It is nice to connect with you as always. Several weeks ago I received the latest volume of your best selling Data Model Resource Book series and it is a welcome addition to my resource collection. I would be remiss if I did not mention your co-author, Paul Agnew, before we get started. Thank you for taking a few minutes to discuss your latest book with me.

LS: Mr. Seiner. It is always my pleasure and honor to connect with you and to be able to share with the TDAN subscribers. Thank you for your many years of service to the data management community with this excellent publication.

RSS: For the benefit of the readers, can you please quickly summarize the three volumes of The Data Model Resource Books and tell us specifically what drove you to write the third volume?

LS: The Data Model Resource Book series, Volumes 1, 2 and 3 provide re-usable data model constructs for standard models, industry models and underlying data model patterns. The first volume provides standard Universal Data ModelsTM that are very common for all types of enterprises such as models for party, product, orders, invoicing, accounting and many other re-usable constructs. The second volume provides industry Universal Data Models that extend the standard models to cover data modeling requirements for a variety of industries such as models for manufacturing, telecommunications, health care, insurance, financial services, professional services, travel and e-commerce. The third volume provides a library of templates and alternatives for common “Universal Patterns” or themes in data modeling that appear over and over again in most data modeling such as ways to model roles, statuses, classifications, hierarchies, business rules and contact information.

Why Volume 3? For many years, when I was involved in helping clients extend the models in Volumes 1 and 2, or helping them to develop new types of models, clients would ask me “How do you extend these models?” and even more importantly, “How can we quickly extend and customize these models for our organization and our needs to quickly develop any data model with higher quality, even if it is specific to our enterprise?” Volume 3 of the model resource book shows how to develop any type of data model consistently and it looks under the cover of the previous books and examines the common underlying structures that are applicable to all data models. So this book provides standard ways of modeling very prevalent themes in data modeling.

Another driving force was the need for a book showing how the same data modeling requirements may be modeled using either specific or more generalized styles of modeling and that there are pros and cons to each. Volume 3 shows variations on the Universal Patterns ranging from specific to more generalized versions of the same pattern. We use the concept of levels, with a level 1 pattern being a very specific way to model something, a level 2 pattern being less specific, a level 3 pattern being more generalized and a level 4 pattern is even more generalized. A more specific level of the pattern is not better than a more generalized level of the pattern or vice versa; is just has different pros and cons. Therefore the modeler can decide from numerous intelligent choices and apply what makes most sense in the current situation. The more specific patterns are easier to understand and better for communicating, especially to nontechnical audiences, and the more generalized versions of the pattern offer more flexibility when implemented.

In Volume 3 we also address the need to show and illustrate, with concrete examples, how these patterns can be used across a wide variety of efforts such as for prototyping, applications, enterprise data modeling efforts, master data management or various types of data warehousing efforts. Finally, there was also a need to show how to socialize patterns so they can be adopted and in this volume we address human dynamics and principles that help with this critical area of data integration.

RSS: You mentioned that this book focuses on assisting people to save time and improve the quality of any type of data modeling effort. How does this book accomplish that?

LS: In any type of data modeling effort, there is a need to model the same types of data, for example, the various types of roles (that people and organizations play), statuses, hierarchies, recursions, classifications, contact information and business rules. There are many different ways to model these types of data constructs and instead of starting from scratch, modelers can save a huge amount of time by re-applying the “Universal Patterns” in this book, which have been through many years of iterations and scrutiny, and have been proven to work in many production environments. In application development, it is a very common practice to re-use routines, functions and services, instead of developing them yourself. Just like re-using a standard function or piece of code that is well tested, organizations can re-use models that have been well thought out and improved over many years.

Another way that it saves a great deal of time is by providing a way to model consistently. When data models are schizophrenic and the same type of data is modeled differently in different parts of the model, there are much higher maintenance costs for the model, the resulting database, as well as the routines that access the database. This happens because there is a great amount of overhead involved in recreating different models, databases and routines, when the same type of structure or function could have been re-used. Usually, modeling efforts do not even realize that the same type of thing is being modeled inconsistently in different parts of the model. For example, when modeling statuses, one part of the model may show statuses as attributes in an entity (e.g. order received date, order confirmed date, order cancelled date) and another part of the data model may model this completely differently and record them with a “status type” entity (“customer status” entity), without even realizing that the same “status” pattern could have been applied in a consistent fashion across the model. Likewise, when modeling classifications, a model may contain a “gender” attribute, a “marital status” entity, a “person demographic” entity, a “party classification” entity, subtype entities, and other ways to model classifications. I have been involved in many efforts where we have realized that we have modeled similar things differently at the end of the modeling effort and at this point, it is often too late to correct it.

The patterns can also be used as a way to quality assure a data model by checking it against these third-party “Universal Patterns” to see if something was missed or to investigate other possibilities and ways to model that type of data. Thus it can help correct possible big mistakes that could cost a lot of time and money.

The patterns can also help to not only develop data models quicker, but also help to integrate data by using very holistic, patterns that make it easier to integrate data by facilitating common types of data structures.

RSS: That makes perfect sense to me. Explain to my readers how the patterns you use assist you reaching your goals and the goals of the people that read your books and bring you in to assist them in their data modeling efforts?

LS: The patterns assist in many ways for data modeling efforts. They assist by:

  1. Providing standard, mature patterns and templates that help to data model very common types of constructs in a consistent fashion.
  2. Showing various ways of modeling the same type of common data modeling scenarios and sharing the pros and cons of modeling them different ways.
  3. Providing a toolkit of re-usable data model templates that data professionals can use to “jump-start” their efforts when developing or extending data models.
  4. Providing a third party source against which an enterprise can evaluate and check its data models so it can evaluate alternative options or see if it perhaps may have missed something.

RSS: Are these patterns used in Volumes 1 and 2 or your book?

LS: Volumes 1 and 2 use many of the patterns that are described in Volume 3. We knew that there were re-usable patterns and even before Volume 3, we applied them to many of the standard and industry models in volumes 1 and 2. After publishing hundreds of models in Volumes 1 and 2, we gained a lot of clarity regarding different options modeling these fundamental constructs and we felt a need to publish these patterns in Volume 3. In fact, some readers have recommended that a new reader could even start with Volume 3.

RSS: So it seems to me as though the three volumes would be a wise investment for anybody that undertakes or will undertake data modeling initiatives now and in the future. How does the third volume extend what was already discussed in the first two volumes?

LS: The third volume offers patterns which can be used as tools to extend the models in the first two books. They can also be used to develop models that are not in the first two volumes and we have found that these patterns apply to just about any type of data modeling effort.

Volume 3 is very complimentary but also is quite different than Volumes 1 and 2. Volumes 1 and 2 of the Data Model Resource Book focuses on Universal Data ModelsTM , which are common and industry re-usable models that can be used to jump-start “standard” types of efforts. Volume 3 focuses on Universal Patterns that are more fundamental in nature and can be used as building blocks or templates for modeling efforts.

Another way to think about it is the Volume 1 is designed to provide standard models, which will generally cover at least a third of the data modeling requirements. Volume 2 extends these standard data models to cover an additional third of the data modeling requirements that often occur for specific industries. Volume 3 further extends the ability to jump-start a good amount of the last third of the data modeling requirements that can be for any type of data model and that may be specific to a particular enterprise’s requirements.

In other industries outside of data modeling, the use of patterns is very common. For example, when carpenters build tables or chairs there are underlying patterns that they re-use over and over, regardless of the type of table they build. So a blueprint for a specific type of “table” could be akin to the “Universal Data Model” and the blueprint for a “cross lap” or “tongue and groove” pattern in carpentry, would be akin to a “Universal Pattern”. Thus many table designs may use the “cross lap” pattern. And in data modeling, many data designs may use a certain “classification” pattern.

RSS: Can you summarize the pros and cons of having alternatives when delivering a data model rather than always following very specific or very general data modeling patterns?

LS: There is an argument that says that if you only provide patterns for either a specific or more generalized style of modeling, then the models will be more consistent and easier to manage. When there are many alternative patterns offered, there is a risk that the data model using these patterns may be more varied and have inconsistent styles of modeling.

One possibility is that people using the Volume 3 book that only want to use specific styles of modeling, could choose to standardize on the level 1 or level 2 styles of pattern, which are more specific styles. Likewise, others that only want to use generalized styles of models could choose to standardize on the level 3 or level 4 styles of patterns (the more generalized styles).

However, I think it is critical to use the pattern that bests the situation. Sometimes, it is appropriate to standardize on specific patterns, for example, when developing data models that are used to communicate data requirements to business representives. Sometimes it is appropriate to standardize on generalized patterns, for example, when developing data models that form the basis for a flexible database design, such as for a master data management system. In many environments, there is a need to use specific models for one purpose (communicating data requirements and/or prototyping) and generalized data models for another purpose (when flexibility in the database design is needed). Often the modeler must decide how much flexibility is needed within different parts of a data model and then pick the appropriate level of generalization for that part of the model since there are tradeoffs (that are discussed in the book) for the different levels of the patterns provided. Thus, it may be appropriate for some models to use both specific and generalized patterns, even within the same data model.

We felt it was imperative that we offer various alternatives for modeling these patterns as well as label them by levels (e.g., levels 1, 2, 3 or 4) so that if the modeler wanted to keep the style of a data model similar, he/she could choose to use the same level of patterns.

Very often, people want me to just tell them the “best” way to model something instead of providing alternatives. They sometimes say, “you are the expert so just simplify tell me the ‘best practice’”. However, is it better to model using a very specific style or is it better to use a more generalized style? My answer is often: “It depends on the circumstances.”

Thus, I believe that while there are ways to model that are more effective than others, there is usually no “best” way of modeling something because there are pros and cons to various styles of modeling.

RSS: Len, over the past few year I have seen you evolve not only as a data modeling and data management industry leader, but also as an individual that focuses on the human elements and aspects of data management. Can you please explain the relationship between your Data Model Resource Books and this evolution of your areas of expertise?

LS: After spending more many years developing data models, I came to a realization that while these re-usable data model constructs can help organizations, there was something else that was so critical that needed to be addressed more thoroughly. When I looked at the clients that had achieved the greatest successes in data integration efforts, there was another factor that was critically important: they used what we call “Universal Principles” of integration. These universal principles have to do with the how people and organizations interact and the human dynamics involved in any data modeling or data integration effort. Examples of this include understanding the motivations that are at work, how the vision is developed, how trust is created (or not created) and how people manage conflict.

Many years ago, I became keenly interested and focused on what really leads to successful data integration and I started compiling case studies, principles, strategies, techniques and toolsets that can help people and organizations from an effective human dynamics perspective.

While I believe that effective data modeling is important, I also believe that human dynamics are an essential aspect to integrating data, systems and people, and we have continued to develop training, consulting, publications and tools to help in this important area.

RSS: I was fortunate to be able to attend a portion of your ‘Human Side of Data Integration’ session at The Data Warehousing Institute (TDWI) conference in Las Vegas this past February. Thank you for picking on me throughout the time I attended. As you know, I presented on “Non-Invasive Data Governance” at the conference which involves a lot of the elements that you mentioned, specifically “trust”. How do these human elements play a role in successful data modeling?

LS: The last chapter of The Data Model Resource Book, Volume 3 is actually dedicated to answering this question about some of the human elements of data modeling and it is called “Socializing the Patterns.”

One of the primary purposes for data modeling is to help integrate data. For example, an enterprise data model helps create common constructs that can be used consistently throughout the organization, thus helping each project to use similar semantics and models. Another purpose of data modeling is that within a particular application, the data model can help with common semantics and a common understanding of the data for that application.

So, a key need in data modeling is to get to a common understanding of the data and in order to do this, there are many human factors and human dynamics involved. It is essential to be able to effectively communicate and to understand the various perspectives and motivations at hand. It is essential to develop trust for any type of data modeling that helps to integrate data because without trust, there will be barriers that keep people’s views, semantics, models and data separate (See TDAN article Data Integration Requires Trust).

One key issue in data modeling is “What to do when data professionals disagree” (I will be giving a presentation on this on April 9th, 2009 for DAMA International’s Enterprise Data World). I have found that a common issue on many data modeling efforts is that data modelers often disagree on the way that something should be modeled. I have witnessed efforts that have failed because of modelers insisting that “their” way is the “right” way and then badmouthing the effort when they didn’t get their way.

In short, data modeling is largely about integrating data silos. These data silos, at their root, come from people silos and it is so important for us to address the root of these issues.

RSS: You are I are on the same page when it comes to leveraging the personalities that already exist within an organization. Can you tell us a little bit more about the book, what it took to put the book together, how it is structured, who the audience is, … anything that will assist people in understanding what they can expect by reading the volumes?

LS: This book is based upon decades of research that we have gathered from many experiences implementing hundreds of data models internationally. Even before we started developing the book, we had developed many of these patterns over the years and had a great amount of experience successfully implementing them at many clients and giving presentations and seminars about the use of these patterns. It took thousands of hours to develop the book and we spent a lot of time reviewing the patterns and going through many iterations of each pattern to make sure that the patterns represented very high quality, re-usable constructs.

This book is designed for anyone that has a basic knowledge of data modeling and wants to be able to be more productive and increase the quality of their data models. This includes data modelers, data architects, data analysts, database administrators, database designers, data stewards, computer science teachers and students, corporate data integrators, as well as anyone involved in any aspect of data modeling. The content of this book is suitable for use by professionals in the fields of data management, data quality, metadata management, master data management, data warehousing, data governance and any other field where data models are used.

The models in this book can help anyone involved in a data modeling effort to re-use constructs, see different alternatives, understand the pros and cons of each alternative and make effective choices when modeling.

The book is organized into 10 chapters which includes an introduction, 7 chapters that each offer a particular pattern (e.g. “patterns” chapters), a chapter focused on how to use the patterns in various types of situations (e.g., master data management, data warehousing, application development, and so on), and a chapter on how to socialize the patterns, which involves principles and techniques on human dynamics in modeling.

Each of the seven “patterns” chapters provides different variations, or levels, of the same pattern, starting with the most specific version of the pattern and moving toward the most generalized versions of that pattern. For each variation or level of the pattern, there is an example showing how to apply the pattern to a fictitious scenario. For example, in the classification chapter, there are numerous classification patterns and each of those patterns is applied to fictitious scenarios where there are needs to create specific, less specific and more generalized data models. Each data model that is rendered has data illustration tables that show examples of the instances of data that may be included in the model, in order to clarify and illustrate the models.

RSS: I will ask you the same question I have asked others through these interviews. Was this book the result of a “labor of love” or a “love of labor”?

LS: Interesting question. This book was very difficult for us to produce as the hours were very long and it took a great deal of effort and a great number of iterations to put out what we thought was a very high quality product. We thought it was so important to release this information as we believe that this can really change the way that data modeling is conducted and that this can help revolutionize the data modeling field by moving us to what only makes sense: re-use what we know works.

So I guess for me, it was a “love of labor” is that I had an intense desire and love to produce something that I thought could help immensely.

RSS: Why do you think the readers have such great interest in reading your works and how has your published material evolved over the years?

LS: I think that the publications that I have produced have been very practical and the idea of re-using data model constructs is such as an intrinsically productive prospect. Over the past several years, these works have not only evolved to incorporate new learnings on the universal models and patterns, but the more recent publications have focused on human dynamics, and are just as important (if not more important) as the publications on re-usable data models.

RSS: Another consistent question…You are a regular speaker at DAMA events and other events nationally and internationally. How much of the content of these books comes directly from your experiences at clients and at these events?

LS: Most of the work in my publications comes from my direct experience at clients over the last 25 years as well as from other consultants’ direct experiences, such as the experiences of my co-author, Paul Agnew, who has worked with me for the last seven years on Universal Data Model engagements, and his experiences working with clients over the last 18 years. We are also able to get a great deal of feedback from many presentations and seminars at DAMA and other national and international conferences.

RSS: Len, you and I have had a long friendship and I have always respected you for many things including your “passion” for the subjects that you address. A lot of that passion (as well as the passion of others) has rubbed off on me. People have described me as showing a lot of passion for the subjects I speak about. Where does YOUR “passion” come from and how does that differentiate you and your business from the others in the industry?

LS: I also appreciate your passion as well as your friendship. My passion comes from my desire to be of service and live a good life. I am so blessed to be able to write, train, consult, provide software and offer these services to help people and make connections all over the world. It is such as great feeling to know that I have made a difference in the lives of others. While I believe there are very unique aspects of our business that differentiate us, such as the level of focus and passion on these specialty areas of data modeling and human dynamics, I also think there are many things that are similar. I am honored to be in a field with other great industry leaders that have shared their passion such as you, Steve Hoberman, Dr. Graeme Simsion, Bill Smith, Karen Lopez, Bill Inmon, Dr. Peter Aiken, to name only just a few.

RSS: What types of engagements do you typically get involved in and how important is that experience in putting together the Data Model Resource Books?

LS: I typically get involved in helping organizations either quickly develop or quality assure their data models via either consulting, training or helping clients customize the UDM models that we license to them. I also am involved in helping organizations cultivate their human dynamics tools, strategies and techniques for more successful integration efforts.

At the heart of the engagements is providing service to help our clients integrate their information, systems and people via “universal” solutions. These include all different types of services and re-usable models such as “Universal Data Models”, “Universal Patterns”, and “Universal Principles of Integration”.

RSS: And on a side-note, I know that life in the data management business and also raising two beautiful daughters is a very busy life (I am having the same experiences but maybe just a few years behind you), what types of books do you read in your spare time, if in fact you have any?

LS: Yes, I am so blessed to have such as wonderful wife and daughters, as well as great friends, and that keeps me pretty busy with family activities, my daughter’s soccer games, and family and friend get-togethers. When I have some spare time, I love reading books on personal and spiritual development.

RSS: Len, thank you very much for taking the time to answer my questions. I hope that will be permitted to publish excerpts from the book and new articles from you in the near future. Any last words for the readers?

LS: To all the readers - thank you so much for your support over the years and thank you for taking the time to read this interview. I hope this was helpful to you.

RSS: Thank you again and I look forward to connecting with you again soon. Best wishes for success with your new book.

Published in TDAN (The Data Administration Newsletter)