To centralize or to decentralize, that is the question

Rafe Needleman, writing at ZD’s AnchorDesk, thinks that small companies are smarter than big ones when it comes to managing technology.

I think Rafe missed the boat with this article. But it provides me a great springboard to talk about managing in a centralized way (which is what he’s describing when he talks about the big-business approach) versus a decentralized way (which is what he’s lobbying for).

My pet theory is that companies are constantly going back and forth between centralized and decentralized management of internal resources (marketing, IS, distribution, whatever). At any given moment in time the organization is either centralizing to make operations more robust and cheaper, or they are decentralizing to make them more nimble and responsive to customer needs (which is what Rafe is proposing in his article).

Just like kids squabbling in the back seat of the car, we’ve got to stop! that! eternal swinging back and forth between the two monolithic extremes. Neither is really the answer — the right place to be is as follows:

  • Centralized things are those for which the customers all share the same priorities.
  • Decentralized things are those where customer priorities are diverse.
  • An example or two? read on…

    Let’s take the distribution function in a consumer-products company. Who are the customers? The product managers. Do they all have the same priorities for the way their products are distributed? I dunno, depends. But if they do, then the distribution function can be centralized, made very robust and very low cost. If the needs of product managers are different (for example; some products are perishable, some aren’t) then the distribution function isn’t a good candidate for complete centralization because some product managers are going to be ill-served, and after a while they’ll bootleg their way around it because it isn’t meeting their needs.

    Another example — the IT function. Should it be centralized? I dunno, depends on what the customers need. Maybe parts of IT qualify and parts don’t. Maybe all the customers of IT have exactly the same priorities when it comes to the server farm (“I want it up all the time, really fast, and really cheap”) but different priorities when it comes to the applications that reside on the servers. In that case it makes sense to centralize the server farm and decentralize application support.

    The same can be said about almost anything a big company does — sometimes it pays to spend more money on diverse modes of function-delivery (especially if that thing is strategic), sometimes it pays to put all your eggs in one basket and blow the doors off with super capability and ultra low cost. *That’s* the way to stop the pendulum…

    This idea was shamelessly lifed from Bob Alloway, who long ago taught this to a bunch of us geeks, later went on to develop the “grades” system that was used to track the progress of US government agencies towards Y2k readiness, and has since disappeared, at least from my address book. If anybody knows what Bob is up to these days, I’m all ears.