Saturday, July 11, 2009

Kolmogorov and the measure of competitive value

I was fortunate enough to attend USI and, although I could not participate to all the sessions, it has been quite fruitful. The (summer) "University of the Information System" is organized by Octo, BCG, le Monde Informatique and TV4IT. It is a great gathering for "bosses and geeks", with lots of opportunities for networking (and meeting old friends as far as I am concerned) and learning exciting stuff (the list of keynotes is amazing).

1. Complexity

I'll start with the key idea that attending a brainstorming session moderated by Luc de Brabandere generated. It may be stated in a pompous manner as:

  • CompetitiveValue(IS) = f(complexity)
    The competitive value from Information Systems is a function of their complexity (in the Kolmogorov sense)

It starts as follows: what is not complex is easily reproduced and become a commodity, something that anyone can use and that may not, therefore, seen as a competitive advantage. For instance, the chisel in the hand of the stone carver is such a commodity tool. Although it is crucial to the task, and is taken great care of, the chisel is not a differentiating factor. Anyone can get a great chisel. What makes a great stone statue is the talent and the craft from the hands of the stone carver. For those companies for which IT is a differentiating factor, there is a fair amount of complexity that has been mastered, from a size, a technological or a business integration perspective.

This is actually very close to the concept of information measure as defined by Kolmogorov. Let's recall that Kolmogorov measures the complexity of an information sequence as the size of the smallest program that can generate the sequence. Anything that is very rich but generated from a set of few rules has a small Kolmogorov's complexity, while chaotic and random structures have a high Kolmogorov complexity. Here the measure of the information systems is precisely what cannot be reduced to a set of rules and a few enabling technology. If you have a large information system which is using standard tools, standard techniques in a usual manner, its complexity measure will be small. If you have an information system that is uniquely tuned to the business, where the practical know-how built over years has helped to resolve technical difficulties, its complexity measure is high.

This approach is strikingly compatible with Nicholas Carr's position on "IT does not matter", that is, IT without complexity is a commodity. One must read the original article or the book to see that Carr is talking precisely about the competitive value of information systems. His vision, which is fairly optimistic in its timing but generally accepted as target architecture, is that Web Service mash-ups will transform IT into a commodity. This Web Service / Cloud IT will not be without value (as necessary as electricity) but without differentiating value. I disagree about the availability of this “commodity IT” (cf. my book “Information Technology for the Chief Executive”, whose second chapter talks about N. Carr’s position), but I definitely agree with the (obvious) statement that there is no differentiating value with a commodity service.

One could say IT without complexity does not exist, but it is not true. Software as a service, for instance, is clearly one direction to remove a fair amount of complexity. Really simple IT exists; unfortunately it cannot solve all problems. At the end of the day, there remains a lot of complexity, irrespective of the technology or the procurement options that are chosen (cf. the Web site of the SITA: Sustainable IT Architecture). This is why companies need a CIO in the first placeJ. For me, the first job of the CIO is to manage complexity. This includes:

  • Reducing complexity through an Enterprise Architecture approach,
  • Removing complexity whenever possible, that is empower users to manage their own information system;
  • Taming complexity through collaboration and training

The paradox is that the CIO’s mission is to constantly reduce the perimeter of her/his job. But since the mission of any enterprise is to be smarter than its competitors, new challenges keep been thrown at the CIO …

Cloud computing is about complexity “de-materialization”: the complexity does not vanish (cf. SOA is not scale free), only its nature changes. If the IS is managing a complex set of business processes with a lot of service interactions, rapid changes and performance constraints, moving the IT on the cloud does not make the complexity disappear.

Back to this idea of Enterprise Architecture, the challenge is to produce a flexible architecture which may sound as an oxymoron. More precisely, one must create structure without rigidity. This prompts a few suggestions:

  • Be wary of “invariants” - there are few of them. So-called invariants are traps to install rigidity (for French readers, this is a topic covered in my first book)
  • Reference designs are living objects. Architecture relies on a number of reference designs: data models, service catalog, integration framework, etc. Any good Enterprise Architecture methodology will tell you how to build extensible designs. It looks like design violation is closely associated with innovation (?)
  • Diversity is key (a theme from the second day of the USI, I’ll be back in a moment).

This line of thoughts brings us back to biology and the general theme of this blog. In the living world, “invariants” (building blocs) are small and they are versatile (better than flexible). As explained by Albert Jacquard during his magnificent talk, diversity comes from reproduction, hence from randomness.

2. Uncertainty

The brainstorming session which generated all these ideas was about uncertainty. How to live, how to create, how to be relevant in a uncertain world?. Luc de Brabandere used a different set of scenarios, which are somehow similar to the four scenarios of Dan Rasmus in his book “Listening to the Future”.

I am a big believer in this approach to define a proper strategy for information system (cf. previous reference to the second chapter of my book). A scenario is not forecast, it is a virtual situation designed to foster creativity. This is a key point: the scenario’s value is not to be as close to what will happen in the future, it is to help build skills that will prove useful in the future (for the information systems or for the employees). In a world, the goal of the “scenario exercise” is to develop one's situation potential.

I have developed a “theory” over the years (cf. my other french blog), that the only tool to master uncertainty is gaming (as in "serious gaming"). Games are based on virtual scenarios but may develop true skills or help better understand the possibility of the future. I will return to this idea in a future post. What came to me as a conclusion of this afternoon session is that “Participants must participate”: passive viewing is of (almost) no value. This is deeper than it looks: since the scenario is not interesting per se, the value is the thought experiment and the collaboration that occurs between the participants while they play with the scenario. If a summary is proposed to a set of external listeners, most of the time it sounds dull or strange. I came to express this as a communication rule: if you need to report the result of a scenario-brainstorming session to your managers (or some other managers), it must keep the form of a role-playing exercise where the audience is actively engaged.

3. Value

This first section of this post dealt with “differentiation value”. What about “regular” value? The classical issue of the value that is produced by information systems was, as one would expect, central to this year’s USI, with one dedicated session on this topic. A good reference on this topic, by the way, is Amhed Bounfour’s book “Organizational Capital”.

That session started with an outlook on the issue, stating that there is neither consensus nor any method that would be applicable to the whole spectrum of issues (I agree, I wrote as such in the previously mentioned bookJ). Octo’s proposed approach is to define a “usage value”, very similar to Adam Smith’s definition: the value of a component of the information system is the additional amount of time it would take to perform the task without this component. It is expressed as a monetary value and actualized (over a given amount of time, such as the life expectancy of the software component.

It is a convenient measure, because it is easy to understand and relatively easy to evaluate, at least when an order of magnitude is concerned. Obviously the value must be capped by the total amount of money generated by the associated business process, in order to cover activities that would not exist without an IT platform (e.g., something that would require a million hands and generates little value). It has a few nice properties: it takes the quality of service into account, as well as the true deployment of the component. A beautiful application that is almost never used has a null value with this approach, a desirable property which is not true of all methods!

It is also a shortsighted measure, proving once again that it is hard to conciliate all objectives (cf. the introductory point). This measure does not take the future into account and how the information system is ready to embrace change. One could argue that “usage value” could be made future-oriented with a scenario approach, following the tracks of the previous section … and that’s true but that’s hard.

Adam Smith’s definition was quoted one day earlier by Daniel Cohen during a great evening speech. He talked about Philippe Askenazy's work on the Solow’s paradox (the absence of evidence of productivity gains due to IT). Philippe Askenazy, through a careful study though a large sample of data points, was able to show that IT can bring value only in conjunction with re-organization:

  • Value = Information System + Re-engineering

Those company who decided to reorganize themselves as they introduce IT in their processes showed significant returns after a few years, while those whom embraced IT but did not change, had nothing to show but costs. This is a great piece of evidence since it supports a claim easy to understand for any CIO: IT revolution only works if used as a lever to re-organize and re-optimize work (hence the importance of business processes).

I will conclude with a great idea from Pierre Pezziardi and Laurent Avignon: foster innovation through opening new territories, places where anyone can contribute to the information system. Obviously one cannot allow anyone to touch anything anywhere, hence the concept of “new territories”. This would be a “zone” where almost anyone can make a contribution (a piece of software) that may be used by others. There is a lot of value there:

  • Foster creativity and innovation
  • Improve the image of the information system, from a dark mystery to a friendly tool J
  • Support collaboration and engage a dialog with users
  • Find all the talents that hide in the company (i.e., not in the IT department) and who could contribute
  • Introduce diversity: the peaceful coexistence of safe, structured islands with active (even chaotic) agile platforms

For instance, one could open a few web services that give access to the heart of the information system (in a read mode for a first experiment J) and pick a sand box (such as Excel, Microsoft,, Google App Engine) that is exposed with a SaaS (Software as a Service) philosophy. The ease of deployment (and adoption) is crucial here to make this a meaningful event. This should turn into a big innovation contest!

The general theme of Pierre and Laurent’s show “L’informatique conviviale” (it was nicely executed on stage in a very lively manner):

Diversity is necessary, Pleasure is key.

The first point is well developed in OCTO’s book “Une politique du Système d’Information”. Their dialectic analysis is too rich to be reported here, but may be summarized as follows: there are too many conficting constraints on the information systems to adopt a unique set of policies. Different parts of the information system must be governed with different approaches, to reconcile the needs for innovation, agility, safety, performance, reliability, etc. This “zoning” of the information system requires clear “passage rules” to support the exchange flows between the different zones.

The second point was the heart of their talk: build an information system that brings pleasure to the users and pleasure to the developers. This is a great insight since biologists tell us that there is no learning (continuous improvement) without pleasure. I will conclude with something that I learned a year ago at a conference about complex systems, on the process of learning for all living organisms, from the smallest to the most complex (us).

I had learned ten years ago about the PDCA cycle of continuous improvement from Deming: Plan, Do, Check, Act. There is a similar cycle that nature has invented for learning: Desire, Plan, Execute and Please. Desire creates the will to plan, to formulate a goal (for conscious livings), to get ready. The execution yields pleasure, which strengthen the desire and reinforces the cycle. I believe that pleasure is an integral component of corporate/collective learning and continuous improvement. Pleasure can take many forms, from pride (in Japan) to simple fun (in a Silicon Valley's start-up).

No comments:

Technorati Profile