Wednesday, May 30, 2012

What taught me about engineering....

I have made some posts before about an idea of needing a more Comprehensive Methodology for Many engineering problems. Spefically, the windmill I choose to tilt at is Sheet Metal Stamping Process Engineering (formability). While I wont user this space to get too far into that topic (you can follow up at the post listed above or in the book Sheet Metal Forming Processes and Constitutive modeling (shameless plug).

But I digress, I have been trying for a while to find an analogy that people would understand for the need and drive of people to be able to digest information based on multiple-dimensional-Decision-Drivers. For most people you could weigh most decisions based on Function-Quality-Cost-Time. We want something done (function), and done well (quality) for a reasonable cost (duh) on time. How well any solution achieves that desire should be measured upon those criteria.

Whenever, we discussed this idea with others; they frequently tell me that either they are not responsible for cost or project time, or that they already have rules and standards in place that are used to govern whether or not they achieve their cost and time goals. This has become an unassailable truth in their minds one which I sought to alter (as we are selling software that aims to address Quality-Cost-LeadTime for sheet metal stamping processes).

Finally, while searching for flights to our international sales meeting I stumbled upon an analogy I think should hit home. showed me the way...

Assuming that arrival time was the most important quality function the flights as shown above are ranked as the best. If departure time was the most important quality metric then the views would be completely different.

Now these flight options must clearly come with different costs. That economic reality is very clear to everybody. But what if the internet did not allow for explicit display of costs? What if our company followed a process flow that stipulated that one department sets financial rules and the other departments had to fit those rules while not knowing the numbers used to generate the rules.

This is not such a far fetched let's imagine how these flight plan engineering rules might look:

  1. Airlines charge more for convenience, therefore direct flights should be avoided for they will cost more
  2. Airlines route to hub airports to optimize on flight capacity, so flights that make use of major hubs will be cheaper (more legs to more hubs are cheaper)
  3. Flights that provide shorter layovers will be marketable at higher prices, longer layovers may be cheaper
With those relatively clear rules in mind, we could narrow the flight options to several viable options. 
  • The American Flight via LGA/JFK
  • The Delta Flight via JFK
  • The United Flight via FRA (and ORD)
  • The Cathay Flight via ORD and HKG
The problem is that with each of these flight options we are having to prioritize without any measurable which rules to follow. Also, we cannot put a value on arriving early or late to the meeting or having to leave home earlier. This is a more common problem than we might realize with our engineers. Our engineers have to make decisions that they know affect costs, but many times without the tools they need to recognize the cost of their decisions.

If we now introduce the cost measurement to the travel example:

We now can see that some of the "rules" did not apply to this particular scenario. The cheapest flight overall did not have the longest layovers and did not have more legs. In fact the flight that by "rules" wisdom the "best" the CATHAY flight that maximized inconvenience and therefore seemed to be the one that would be cheapest was a scant $8000 at the time of this posting. Not to mention that it effectively force the traveler to make enough miles to circumnavigate the globe without actually having the benefit of doing so.

But where HIPMUNK really shines is the further filtering function of AGONY.
With Agony as a filter we can see that the relatively low cost of $1630 a mere $48 dollars more buys us a flight with a shorter layover less agony. The fact that it also has no flight changes in the US may or may-not have any affect on the agony algorithm used, but for me and any seasoned traveler knows that EU member nations on the continent are less of a headache than the average US airport for layovers (Heathrow lately being a major exception).

What did this teach me about engineering?

  • That without a clear methodology for weighing out the multiple dimensions of decision drivers we simply cannot make sound decisions; We will most likely end up shorting one criteria for another
  • If we accept that there will be some compromise we can use background analytics to weigh out many seemingly intangible cost/benefits of the various options that we consider
  • If we did not consider that there were more than one option and thought that the first answer that successfully met the "quality criteria" we may have never recognized these better solutions; imagine if we just took the first flight that got us from A-B on time

Thank to HIPMUNK for not only being my travel search of choice, but for helping me illustrate to my colleagues that our vision of a multi-dimensional decision engine is, in fact, the way to go.


No comments: