Choosing "A Single EA Repository Of Truth For Enterprise": A Dream Turning Into A Nightmare

As Forrester’s EA tools analyst specialist, I am regularly receiving inquiries from EA teams that are encountering trouble choosing the "single repository of truth" for the entire enterprise. Generally, they are oscillating between two products after a long decision process, hesitating in many cases because no one product is able to satisfy all the architects: the EAs, the solution architects, and sometimes the business architects. One product satisfies some architects and not the others, and vice versa; in the end, choosing one single product would not satisfy anyone because for each option that will satisfy a few, some will not use it (generally, for good reason), and it will not give others the information they require to do their job. Therefore, for these EA teams, the dream of getting a "single repository of truth" is becoming a nightmare. I encounter this sort of dilemma in half of the inquiries I receive about EA tools and particularly within the largest companies.

My answers are sometimes difficult for these EA teams to hear:

  • First: Do all team members agree on EA objectives for the next two to three years? Do all architects know and share the same IT objectives and priorities? If EA and IT objectives/priorities are not clear, it is not surprising that they want different tools, because a universal EA tool does not really exist at this time. The recent document I published about the EA management suite as a third generation of EA tools explains how the most recent two generations complement each other.
  • Second: The dream of a single EA repository should probably remain a dream at the moment. Even in the longer term, it is better to think about building a series of repositories and interfaces between them to support the industrialization of IT as well as to allow IT to participate in APM, PPM, COBIT, COSO, CMMI, ITIL, and other IT best practices. The next document I am preparing with Gene Leganza about metadata architecture will bring more on this subject.

If you faced this problem in the past or if you are currently facing this dilemma at the moment, please share your experience and how you solve the debate internally by commenting on this blog post.

Comments

There's no single EA repository

This is no secret, there's no single EA repository ... even if some people continue to look for the graal. The architect as the guardian of knowledge in its ivory repository is dead. Business is agile, and EA should be agile. EA information is everywhere and what is iportant is to enable access to it from a central place and to offer some "big picture view" that could be described in detail by other documents and tools.
The knowledge should be built from different sources and aggregated dynamically.
Putting un place a single EA tool takes at least 6 months and cost a fortune.
What is needed here is:
1) interoperability between tools through a unified metadata language
2) way of integrating/aggregating/making mashup of any EA related data
3) innovative solutions using the power of semantics ... Try to get the definition of application in your company ...

Today, the old generation of EA tools is outdated, and the needed tool for today's need does not exist ... You will have to choose between a portfolio and financial oriented tool, or a pure EA description tool (the ones with boxes and arrows).

I do not understand why this sector was not more innovative ...

Next gen EA platform

Thanks for sharing your viewpoint William. I agree with the requirements you are describing, but it seems to me some vendors are coming on 2 of your 3 requirements :
- Mega, Casewise, Metastorm (CIP) and Troux, NoMagic Cameo and others are providing an ETL environment and/or a canonical format. They tend to go to the metadata architecture I am asking to help exchange models between IT siloed applications.
- Mega, Metastorm (Workspace) and Software AG (Mashzone) are offering mashup capabilities.

I fully agree with you the semantics capabilities are not there and it is missing to "value" the EA knowledge.
I would add a 4th requirements which will become increasingly important which is different type of collaborations to help EA share and get knowledge from EA stakeholders. It should include workflow (structured collaboration), IM, wiki, blogs, forums, and new collaborations types. It should go further than what does Sharepoint because the collaboration platform should provide metrics about the contributions. The collaboration techniques should help to recognise the people who share the most. We start to see such collaboration technologies with Software AG Alignspace, Metastorm M3, Lombardi Blueprint, IBM Blueworks. But we are just at the beginning of this trend because the missing piece is the value and metrics

And you what would you add in the future platforms for EA?

Next gen EA platform

Henry,

Not sure that collaboration is not too constraining. You do not need to collaborate to share in Web 2.0 ... And sometimes you do not know where to "hook" the information you have to an existing body of knowledge. That's why semantic (or tagging or folksonomy) is more suitable that collaboration. The objective is to benefit from informal crowdsourcing, feedback and information sharing ...

Then users can "like" your document or "tag" it. Documents should have an owner, but document "connections" should be dynamic and evolve over time.

Today EA tool are too proprietary ... Most of the time you need to program to extend the functionality ... EA tool today are like ERP, with proprietary module and repository (and sometimes even proprietary DBMS or language ...). And to be funny, none of them is providing an architecture description of their tool ... in their own tool. As a customer, that's a minimum you need to have, the app you use being described.

The first reason why architects and company can not deploy EA is the cost of licences. The pricing model is killing EA by itself. The more people you will let access the EA tool in the organisation, the best it will be. Costly user based licences reduces people capability to modify or enahnce the body of knowledge.

So for me, EA 2.0 will come by democratizing access to the platform and enabling everybody to have the capability to create/modify/enhance the descriptions made. Tool miodules enabling to access and manipulate the information (not the metamodel or metadata) should be free for any employee of the company.

I would like also to have my EA tool in SaaS mode.

We all need EA 2.0 ... But the market is not important enough to justify innovation and investments, until now ...