Disclaimer: This blog, including this post, is not sponsored by CombineNet. The author is not employed by, contractually engaged with, or affiliated with CombineNet. Any and all opinions expressed herein are solely those of the author. Furthermore, the opinions expressed herein should be contrasted with the opinions of other educated professionals before the reader forms his or her own opinion. Finally, the author is neither endorsing nor dissenting the use of CombineNet’s products or services – merely trying to spread awareness on the importance of optimization and the relative uniqueness of an offering like that of CombineNet. This disclaimer holds true for each post in this multi-part series and will be repeated.
For those of you following the sourcing blogsphere, you’ll know that I’ve been giving CombineNet a bit of a (good spirited) hard time lately on Spend Matters, eSourcing Forum (ESF), and here on SI, but I’m just trying to poke and prod them into educating the sourcing community as a whole since I believe that decision optimization is still not well understood overall, and optimization is a much more involved topic than most people realize. After all, they have what should be the only real optimization blog out there, CombineNotes.
If you haven’t, I would highly recommend you read the spirited debates over on Spend Matters that resulted from the following posts:
- Old News Keeps Flowing
- What Do Rubik’s Cube and Expressive Bidding Have in Common
- An Optimization Knock-Down!
as well as the following CombineNet posts:
- Project Zander and Comprehensive Network Design
- CombineNet – The Allocation Company
- Analytics Support Negotiations
- Expressive Commerce and the Long Tail
- Perspectives in Puzzling
- The Make vs. Buy Dilemma
- ‘Right Tool for the Job’ Sourcing
If you’re new to SI and missed my Optimization series over on ESF, you can still review part I, part II, part III, and part IV, and if you missed it, my first post on Decision Optimization is still available in the archives.
For those of you who’ve read the posts, and just want a quick recap, here it is.
I fear that optimization is not well understood and that much needs to be done to educate different users on the strengths and weaknesses of differing methodologies and solutions. There are upsides and downsides – the Rubik’s cube is simultaneously the best and worst analogy I’ve seen yet – the apparent complexity is that of a Rubik’s cube, but the actual complexity is much more so.
With respect to optimization, there is a sharp distinction between the problem, model, and solution (algorithm) and confusing them can be dangerous. If the model, and the modeling capabilities of the tool, are not appropriate to the problem, or not useable by the end user, it does not matter how good the solver (or solution algorithm) is. (MindFlow proved this point.)
There are embedded POE (Platform Optimization Engine) solutions, based on COTS (Commercial Off-The-Shelf) optimizers, BoB (Best of Breed) solutions with custom (proprietary) solution algorithms, and solutions that merge the two. The best solution all comes down to the problem at hand and its modeling capabilities – a better algorithm does not necessarily imply a better solution, although it will probably reach one faster. The model is key. Furthermore, there is a cost associated with pure speed – when it comes to optimization, you can only have any two of deep, fast, and accurate.
Depending on the problem, a better model may not save you more money – it depends on how fine grained the data you have available is and whether or not the model’s constraint representation abilities can support more advanced costing models. If your data is coarse grained, chances are the additional savings provided by a BoB solution will be negligible (less than 1%), if any. If your embedded solution limits the cost factors (i.e. forces you to combine unit, usage and/or transportation costs, for example), then a fine grained model may save you a couple of percentage points if you have detailed transportation and / or usage costs at multiple freight brackets. If your embedded solution does not support sophisticated discounts or bundle based costing, then a Best of Breed solution may save you significantly more, and the quoted range of 5% to 10% is very realistic. (However, if your embedded solution does support discounts and bundle-based costing, then a best of breed solution may not save you more than a few percentage points, if that much.)
CombineNet has some of the deepest models out there, they are one of the few companies with proprietary solution algorithms (which require a lot of brain power and development time), I thoroughly believe that their logistic models are best-in-class, but I’m not entirely sure that they are “without peer” when it comes to sourcing, primarily because of what I said in the last paragraph. It depends on the problem and, since this is business, the ROI.
Compared to many of its “peers”, CombineNet has a price tag that is directly correlated to its capability – quite high! (Based on quotes I have heard, one event could cost you as much as a year of unlimited events from an on-demand provider for a small team of sourcing professionals.) For some problems, I strongly believe that, in the words of the SpendFool a honda engine will do the job just as well as a lear engine, and if we are talking about a spend in the 10M range, then I would doubt the price tag is worth it. But if we are talking 100M+ spend on a very complicated category where you need to do an embedded make-vs-buy analysis (which I’ll elaborate on in a forthcoming post), I might actually advise you to spend money hand over fist on CombineNet because even an extra percentage point will generate a significant ROI for you. (For example, even if it only saved you two percent, on a 100M category, that’s two extra million to apply against your bottom line!)
My comments, as you might have guessed, did not go unanswered. According to CombineNet, their product is designed for “everything in the bucket sourcing”, their optimization is easily accessible, their speed allows for more scenarios to be run in the same time frame – increasing the probability of a successful event -, they are now a hybrid of BoB and POE, offering their clients the best of both worlds with their preconfigured templates, and even Jay Reddy, founder of MindFlow, openly acknowledged CombineNet’s optimization capabilities were without peer.
The reality, more or less, (much?) better than it was (but easy is a relative term), definitely, more or less (but that’s not necessarily a bad thing), and pseudo revisionist history. But these are topics for the forthcoming posts in this series, so I’ll leave you with a quote from the SpendFool:
This stuff (i.e. POE & BoB Optimization) isn’t mutually exclusive! Nothing wrong with using the consultants with big brains and tools to solve the really strategic problems AND also using mass deployed tools from ERP and/or SaaS vendors. Pick the right tool for the job. Anything else would be foolish.
Looking forward to your comments on my (next) posts!