Decision Optimization Defined

Monday we defined a basic strategic sourcing process, indicated there were five critical process driven phases that can be greatly enhanced by software solutions, and indicated that we would spend one day discussing each of these technologies this week.  Monday we discussed spend management and spend analysis, Tuesday we discussed RFX, and yesterday we discussed auctions.  Today we are going to discuss my personal favorite: Decision Optimization.

There are a lot of definitions out there for decision optimization (often called bid optimization, award optimization, etc.) as it relates to strategic sourcing, but there are very few fully correct ones.  As far as I am concerned, decision optimization is the application of one or more rigorous analytical techniques to a well-defined model to generate the absolute best decision from a multitude of possible alternatives in a rigorous, repeatable, and provable fashion.

There are four key components to this definition.

(1) Rigorous Analytical Technique
Mathematically speaking, the analytical techniques used must be sound and complete.  In everyday English, the algorithms must always produce correct results and be capable of producing the optimal result.  In my book, heuristic, simulation, or evolutionary approaches, favored by some providers, that cannot always guarantee an optimal answer do not count as decision optimization, falling into the category of decision support.  However, hybrid approaches that use (mixed integer) linear programming would count since the heuristics merely guide the search in the most likely direction of the optimal solution, but do not prohibit the identification of the optimal solution even if it turns out to be an unexpected solution.

(2) Well Defined Model
The decision optimization component must not only insist on a well defined model but also allow you to completely and accurately represent your problem in the scenario definition.  Many optimization products on the market force you to over-simplify your problem to the point where the result is truly not the optimal result because you are missing key costs, constraints, or relationships.  For example, many early products (still) assume(d) that you are always shipping to one location or always buying from one location and do not allow true lane support.

(3) Best Decision
The optimizer must be capable of producing the absolute best decision given a sufficient amount of time.  As we mentioned yesterday, decision optimization is very hard and it is conceivable that an optimizer could take a considerable amount of run-time to find the optimal answer.  However, the implementation must support a configuration where the optimizer will not return until it proves the answer is optimal to whatever level of tolerance you specify, not just when it believes it has the right answer with high probability.

(4) Repeatable
The optimizer must produce the same solution or an equivalent solution each time it is run (for approximately the same amount of time).

Of course, this means that only CombineNet and SCA Technologies appear to be offering true decision optimization solutions now that MindFlow is out of the picture, but even then I find their modeling capabilities lacking in certain areas with respect to strategic sourcing needs. (On the other hand, I do not believe that anyone comes close to CombineNet’s logistical modeling capabilities.)  However, a few other companies are starting to make strong showings, and I fully expect that within a year Iasta in particular might have one of the best offerings based on what I saw in their initial foray into what they call Bid Optimization (released last December) and what I’ve been reading in e-Sourcing Forum (WayBackMachine) over the last few months.

When you consider the recent rampant inflation in energy and raw materials, the constrained capacity of many suppliers, the pressing need for improved top line and bottom results on the balance sheet, and the diminishing returns from traditional auctions at early adopters, decision optimization technology is only going to get more important. As I hinted at yesterday, I think the future leaders in the e-Sourcing space are going to be those that master decision optimization technology and its various applications.

Since this is one of those topics I plan on discussing a lot on this blog, I’ll keep this first entry short and conclude by saying that I firmly believe true decision optimization is the heart of a good strategic sourcing process and one of the best sources of value innovation that money can buy.