The End of PLM Communism?

central-vs-decentral.pngIt is hard to find somebody in PLM industry not familiar with the idea of "single point of truth". I discussed the latest modification of that idea called "the whole truth" here earlier this year. I’ve been thinking about PLM implementations over the weekend and some perspective on PLM concepts. In addition to that, I had some healthy debates over the weekend with my friends online about ideas of centralization and decentralization. All together made me think about potential roots and future paths in PLM projects.

PLM was started as an idea to provide a central point for data about the product and the lifecycle. It was heavily influence by two factors: 1- protectiveness of CAD business to IP and CAD file formats; 2- ideas and successes of centralized ERP implementations. In CAD and PLM space the discussion was always presented as "best of breed vs. integrated", but in fact, PLM vendors tried to apply their influence to implement centralized data management role in PLM. Attempt to copy ERP businesses was another factor pushing PLM towards centralization.

PDM was another influencing factors in future PLM implementation. The idea of PDM was centralization. It is a great model to centralized all document records, drawings, revisions, etc. PDM was never designed to be decentralized. Even more, all trials to implement decentralized PDM systems in the past failed. The challenge of organizations was to grow beyond a single point of data storage, data model and data organization. Later in time, companies started to experiment with "virtual layers" beyond multiple systems and product data domains. Virtual layers were sort of bastardization of centralized PDM idea. PDM systems worked well with a central repository of data, but failed massively to support distributed and decentralized process management. Virtual layer was complicated, costly and didn’t work well.

The challenge is that number of companies brought to centralized PLM concept is large these days. They are continuing to spend money and resources trying to implement centralized data and processes. Eventually, they might discover at the end (after spending a lot of money), centralized idea of data and processes cannot support what they want to implement and cannot work in a distributed environment.

What is my conclusion? PLM implementations (or how some people calls it "PLM Journey") reminded me "5 years plans" in former Soviet Union. They came originally from the same place – centralization and control. The old formation of PLM systems cannot work in distributed organization environment and support distributed processes. New formation of PLM systems need to focus on network, data connectivity and processes to connect data. Just my thoughts…

Best, Oleg

9 Responses to The End of PLM Communism?

  1. Lars Taxén says:


    Interesting column! The metaphors we think by are indeed powerful and deceptive. The “single point of truth” as a beacon for PLM implementations has cost companies tremendous amounts of wasted money. However, we need to be careful about jumping from one extreme to the other (full decentralization); this can be equally hazardous. We have to find a balance between “despotism” and “anarchy”, so to say. But concentrating on federated databases only will not do the job; this has been tried many times in the past.

    I believe that we need to find a completely new metaphor for guiding PLM implementations in the future. I’m quite convinced that such a metaphor must put the search light on “context” (I have touched on this in previous comments).

    In your illustration, the black dots would each represent such a context; or an area of work supported by a PLM-system. These work areas should be the point of departure, not data, processes, rules or any other element we start from in today’s implementations. Departing from context will force us to reconsider many things taken for granted. For example, it will become clear that there is no such thing as a “true” product definition. The product will be completely defined by the needs in each work area (structures, attributes, models, etc.), which in turn needs quite different modeling capabilities from PLM system than what exists today.


  2. Dijon says:

    If the diagram is considered as geographic in nature and the nodes represent physical locations then “single source of truth” can lead to “single point of failure” or “enterprise down” in other words.

    Dispersed operations, whether those are product development centers or factories or …, often need a higher degree of autonomy. If the data center at headquarters is inoperable then should factories be affected by that?

    The separation could also be regulatory induced or to keep things flexible when a company expands or contracts (e.g. acquisitions, divestures, mergers, etc.). Another influence is time zone related for companies which are distributed amongst wide spans of those.

  3. You talk about what drove PLM but I think you miss an important item, the desire to take control away from the ERP system. The real driver was to become a powerful enterprise application by taking control of the BOM away from ERP. At the time there were PDM system but I would say these were not popular and/or widely used.

    Although the plan or objective was to have a single repository the truth is we have ended up with a distributed solution, due in part by the failed attempt to take control of the BOM. Product data is absolutely distributed today. PLM has not taken any market space away from ERP. The PLM effort has certainly increased the amount of PDM that is being used but it has not changed how people deal with the BOM.

    The bigger missing part of the truth is the ad-hoc interaction. Even if the PLM system has all the files, the BOM and all configurations it is missing how the team made decisions. It is missing the discussions used to move the design forward.

  4. Very interesting discussion here and the political back drop from your perspective is a excellent analogy. I think the key point here is control. Everyone wants control of their work environment and as long as the work is localized then they can have it and not really negatively impact the company aside from more cost and less value for many localized systems because you do the setup work over and over. I see typical companies more like the right diagram with no central node only disconnected bodies that are independent because they were acquired or just grew up autonomous.

    Its like in the cold war when countries exported and imported less they grew up and could survive (not thrive) with localized monolithic systems and they paid more for less. They got along. They were internally focused and in control. Today the world is more connected and dependent it becomes a problem to have a lot of localized systems and local control and expect to work easily together across borders. You get it done but it requires a lot of effort and negotiations. So at the point where you need to work across borders to get things done you have to begin to think about the level of standard processes and tools required to get things done smoothly. Its not the left diagram at all in my thinking. You need to begin to establish the central node in the right diagram and decide for your business the level of collaboration that’s appropriate and determine how big that node needs to become and the scope of the processes inside it. You also need to have some way to develop and manage that standardization in your company and make decisions relevant to your business needs. You must have to have systems that facilitate bi-directional information transfer from that node to other nodes and provide information flow because no one system can do what it needs to today across borders. Ask your self can your PLM system do that or was it designed in the cold war times when all was local and centralized?

    So back to the political discussion…. the PLM thinking needs to be more like China and not Russia. China is opening up and accepting some of the models being used by the west so they can begin to play together and grow with the global world. They are careful but becoming more open but they are becoming more open and allowing trade and they are reaping the benefits. Russia continues to lock down and play a little but they are not really intent on openness. PLM systems have to begin to open up slowly like China and find a way to protect their legacy interests while allowing proper information transfer across borders to let their citizens (customers) grow with the world.

    Fun discussion as always.

  5. Interesting idea, and interesting distinction between “whole truth” and “single point of truth”. The key in my view is to have the “whole truth” – that applies to all corporate information assets – as a consistent interoperable whole, readily accessible at the point of need, but not necessarily from a single point!
    In our white paper “The Road to Data Integration”,, we refer to creation of “Continents of Automation” from “Islands of Automation”, then building “bridges” between the “Continents of Automation”. The concepts described in the white paper can be applied in any PLM or IT environment today.
    The rationale behind this approach is that interoperability problems arise not as a result of how many different applications need to be connected (to form the “whole truth”), but how many different data models there are in the mix. The creation of “Continents of Automation” drastically reduces the number of data models.
    The second important point is that only a small subset of information within each “Continent of Automation” is of interest to people outside it. So building the bridges on this basis is also drastically simplified. This is made even simpler if people start thinking in terms of “Corporate Data Models” that define what information is of general interest corporate wide, i.e. not specific to any “Continent of Automation”) and which “Continents of Automation” are “masters” to its constituents.
    This approach answers the issues of control raised by David Sherborne and “balance between “despotism” and “anarchy”, raised by Lars Taxén. It also helps put an end to “religious wars” as to whether ERP, PLM, or indeed, CAD should be “the centre of universe”. Each “Continent of Automation” Tsar has control on what is inside his/her local area, but has to accept the corporate rules that applies to the information of general interest.
    As mentioned earlier, this is not just theory, but an approach that may be applied today in any company environment. For those interested, we have see the “The Road to Data Integration” Workshop on the Datamation home page. What makes this approach implementable is effective scoping of the problem. No one can find a general solution to manage all information as a single unit. It is simply not possible because such a problem is unbound. In contrast, the approach refered to here is based on defining the scope of each “Continent of Automation” which is finite and bounded. It is a “divide and rule approach” to use a political metaphor.

  6. Lars Taxén says:

    Good points made; it seems that we agree that the extremes of both centralization and decentralization are equally bad. We need to find a mix between these, but how to achieve that is far from clear. A seemingly straight-forward way is to agree on the data models in each node and make sure these are compatible. However, what is often overlooked is that “agree” is an extremely unwieldy process since the models have to be interpreted and negotiated by human beings. I cannot remember all the projects launched at Ericsson, where I worked for many years, to create a “Corporate Data Model” and they all failed, because people could not agree on it.

    I’m convinced that if we ever are to come to grips with these issues, we need put humans up front, and find ways to make the consensus creation process more efficient. Technology, systems, processes and data are still with us, of course, but making any one of these the prime target of efforts will not yield the result. To begin with, I believe we must abandon the idea of “truth” whether “single” or “whole”. If we are chasing the “truth” we will never find it, simply because there is no such thing in the PLM context. What we should focus on is “usable”: does it work or not? And this can only be validated in practice in each particular situation.


  7. Lars, you highlighted a very important point. I agree, consensus is not easy, even within a company. However, there is a way to make the consensus creation process more efficient. To use a political metaphor again, all is needed is to have democracy to make everyone aware of what is needed, coupled with authoritarian leadership to enforce a solution.

    In the “Continents of Automation – CoA” scenario that I outlined in my last comment, this can be achieved at two levels. At the corporate level, the decision is made on what information is needed for use outside the CoAs that generate them, and the vehicle for delivering it, e.g. Business Intelligence or portal solutions. Then, at the local level, each CoA is tasked with delivering its contribution to the corporate environment. This is achievable today with web services or similar technologies.

    Again I stress the point I made in the last comment. It is not lack of technology or humans wanting to control their “local empires” that lead to difficult / impossible implementations, but ill definition of scope. If the problem is adequately scoped, then the job can be done.

  8. All, thanks for comments! Great discussion! Sorry I missed the last few days.

    @Lars, I agree both centralization and decentralization are bad. However, I didn’t suggest a complete decentralization. My methaphor is network, which can be adaptive and flexible.

    @Dijon, you’ve made a very interesting point about company growth, acquisitions, etc. Network model opposite to central model, can be expanded and the same time be connected.

    @David, very interesting associations. I like you comparison of China and Russia. You are absolutely right with regards to control. It is not about “central storage”, but about how to control storage and changes.

    @Kais, I think you are right with regards to island of automation and fragmented information. Yes, only part of information from each island is interesting to people and systems from another island. To facilitate the connectivity of information can be a very interesting approach that will improve system integration.

    Best, Oleg

  9. […] Oleg in his doomsday blog post: The End of PLM Communism, was thinking about PLM all the weekend. My favorite […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


Get every new post delivered to your Inbox.

Join 290 other followers

%d bloggers like this: