Wednesday, June 15, 2011

the manual process

How often do highly complicated systems rely on some level of manual processes?

Or how about this: how often do highly evolved models (used for valuation or forecasting or decision support) incorporate highly subjective data?

In my experience, both of the above happen far more often than one might assume.  I find this situation unsettles me.  Our society derives much comfort from "systems", and yet very often the efficacy of a given system is highly sensitive to, potentially, one person's subjective input and/or one person's manual manipulation of relevant data.

On a somewhat related topic, how often does "success" in business or life come from non-intentional actions?  Stories like the invention of Post-It Note glue being an accident that someone realized could be re-positioned as a new product in it's own right speak to this thought I'm having...


  1. Nice post! I think about this a lot. I'm often struck at just how reliant mechanized fabrication is on the human hand. My perspective is from architectural scale construction, which to me is a highly complicated system, but might seem tangential to your original question...
    In design, we constantly have to juggle the ability to be incredibly precise with the necessity of scaling it back to keep the whole system in check. We could design and engineer components of a curtainwall system to within 0.001" precision. But we have to allow for up to 0.125" of tolerance to accommodate the human assembling it in the field. (And to allow for deflection. And to make up for intersecting construction that's less precise. And so on...) If we are overly precise with any one part, the system in aggregate quite literally won't fit together and might actually cause bigger failures.
    This tension between high and low precision gets attenuated as certain parts of the construction process become more mechanized, more prefabricated, more automated, and as our digital building models get more detailed. We still will have moments where less precision is critically important, due to limitations of manual labor or physical realities of material and environment.
    I find the necessity of tolerances reassuring. It keeps me from thinking I have more control than I actually do.
    Is there a data systems analogy to the construction tolerance?

  2. Looking at the last question, I think about the saying "chance favors the prepared." 3M may not have set about trying to make Post-It Notes, but they sure as shootin' assembled an impressive team of scientists, engineers, and marketers who were ready to take that new adhesive and run with it. That's what they set out to do.

  3. Jennifer - sorry for the delay on publishing your comment. I thought I had done so on my phone days ago...and to your question, I do believe there are analogs in other fields.

    What I have seen in my own work are highly detailed models for forecasting X or assigning a probability to Y, but there is inevitably one (or two) variables that are informed by the entirely subjective WAG (wild ass guess).

    Adan - Your point is a good one and well taken, but it also raises other considerations we could discuss (I would suggest that MMM's culture and composition of talent are rare), but I'm still curious how often the random, or volatile, or unexpected XYZ fuels success.

  4. Also, it's early Monday after a long weekend and I have realized that my response to J's comment was essentially comprised of the same words I used in the original post...D'Oh!

    Maybe I could come up with an example to illustrate my point...Say you are an architect, and you are trying to plan for this phenomenon you describe, the disconnect between your precise plans and the ability of human labor to meet those precise specs. So you and your math nerds come up with a formula that solves for something like Aggregate Tolerance Loss and your variables include things like the known tolerances for the materials in use (steel can be very tight to specs, brick not so much?), but then you get into bits that are a little more fuzzy...if you are mixing concrete on site and climate conditions outside of a certain range produce outlier results in the maybe that variable gets a volatility coefficient to try to capture the information supplied by those outliers... I'm getting frustrated just writing this out - my point is that at some point in the development of the model someone is going to have the chance to plug in an assumption and that number is likely to be subjective.

    Here's a far less complicated example: there's a picture hanging on your wall, and it looks a little out of alignment (not hanging level) and you have a couple of choices; if there is an engineer in the house, a laser level and multiple inspections could be involved, but if a history professor is doing the work, there may just be a nudge by the thumb and nary another thought. To tie this back in to the original post, is it possible that Steve Jobs (or insert other person of influence in the business world) is a thumb nudger? Is it possible that Warren Buffett (or Jimmy Buffett) said at some point in his early career "yeah, that's about right"?

  5. My sense is the answer is "Yes."
    The subjective is there, always. Maybe all of the objective systems we have for coming up with precise predictions are all ways of letting Jobs, or Buffet, or any of us, pull the trigger on our WAG's. We design models that validate our hunches. From there it's survival of the fittest hunch.
    Another reference to think of: The ancient Greeks had immensely precise geometric models for the design of important buildings (temples), prescribing specific compositional moves. The spacing between columns, relationship between plan and elevation, slope of the pediment, and other features all conformed to a general set of geometric relationships agreed to be correct, appropriate, and beautiful. Part of that system included 'optical refinements' which might also be thought of as compensatory imperfections. These are not about geometric perfection so much as they are about how humans perceive geometry from a vantage of about 5ft above the ground. Columns physically bulge, or have entasis, because they would look too spindly if they were straight. The main floor surface and front stairs of the Parthenon has a subtle curvature to it so that they will be perceived as flat. To my understanding, there's no precision to this type of refinement. Just an awareness that absolute forms look 'wrong,' and a long process of trial and error to arrive at what looks 'right.'