When choosing an optimization solution, you generally have two choices: a Platform Optimization Engine that is integrated into a strategic sourcing suite (such as Iasta’s Bid Optimization 1.0 that runs on top of industry standard solvers such as Ilog’s CPlex) or a standalone Best of Breed product (such as CombineNet’s customized optimization platforms based on their proprietary solution algorithms customized for certain problem domains and models). The question is – which is best for you?
As you probably have guessed, it turns out there is no easy answer to this question. It really depends on the problem you need to solve, the complexity of the model or modeling capabilities you need to solve it, the size of the problem, and the effectiveness of generic optimization approaches such as linear programming, mixed integer (linear) programming, and constraint-based programming on your problem. It also depends on what you are looking for in a solution – do you want ease of use or flexibility, low cost or (potentially) high value, off-the-shelf or customized, etc?
Generally speaking, POE is an easy-going, relatively inexpensive date while BoB is a serious, expensive commitment. However, whereas POE is only comfortable in familiar situations, BoB can adapt rather well to new situations with very little effort. POE is often backed up by an extensive support group whereas BoB is a relative loner, with only a small group of close friends there to support him.
In other words, since POE is part of a package deal, the relative cost for POE’s services is usually much (much) less than the cost of BoB’s services. However, POE can generally only solve the problems that fit into the pre-defined set of models that POE comes with. In contrast, BoB can often adapt to handle new models and new variations thereof as time goes on. Since POE makes use of the services provided by third party optimization engines, POE gets better not only when the platform improves, but also when the third party services improve. On the other hand, BoB only gets better when the provider supporting BoB gets better. Since POE is generally designed to solve a small range of problems, POE’s user interface is usually customized to the problems in such a way as to make their definition easy and obvious to even the most inexperienced of analysts. On the other side of the fence, BoB, designed to handle a wide range of complex problems, is often quite difficult to use and often requires a lot of education and experience if you want optimal results.
So where do you start? I’d recommend listening to the advice of POP, Practiced Optimization Practicioner who knows that the majority of your optimization problems in a particular area (generally between 60% and 80%) can usually be approximated very well and solved near-optimally by POE and that POE is an easier, quicker, cheaper entry point into optimization than often anti-social BoB. Therefore, you should start with POE on the problems that POE is most suited for, get comfortable with optimization, and get some quick hits. Once you understand optimization, you can then dive in on your more complex problems and determine whether or not you can use POE, the kind of results you can expect, and whether or not you should bring in heavy hitting BoB to tackle those problems POE can not handle, or may not do well on.
In the future, I think you’ll see the market leaders using both solutions – POE for most of their day-to-day sourcing problems, but BoB for complicated make-or-buy, logistics heavy, or non-standard sourcing-allocation problems where large amounts of savings are there for the taking. (Large being the key decision factor on whether or not you engage BoB, since BoB can easily be 10 times as expensive as POE and you need to approach everything from a value-based ROI perspective.) This means that companies like Iasta (POE) and CombineNet (BoB) should both have a very bright future, since I predict that Iasta’s forthcoming Decision Optimization 2.0 offering will be best-of-breed in its category (on-demand integrated strategic sourcing optimization) and CombineNet is already best-of-breed in many ways in its category (stand-alone logistics, multi-variate make-buy decisions, and non-standard complex-sourcing).
For a more in-depth discussion of decision optimization, what it is, what it is not, how it enables decision support, the benefits it provides, and strategies for success, see the Strategic Sourcing Decision Optimization: The Inefficiency Eliminator wiki-paper over on the e-Sourcing Wiki.