Auctions
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 and yesterday we discussed RFX. Today we are going to discuss reverse auctions.
Auctions are not new, with historical records indicating that auctions were held as far back as 500 BC in Babylon where women, sought after as brides, were commonly put up for sale. They were also very common in the Roman empire. Furthermore, English auctions, where a number of well-to-do ladies and gentlemen gathered in a parlor with their numbered paddles waiting to place bids on pre-determined lots have been occurring regularly for hundreds of years.
Furthermore, internet auctions are not new either - with Free Markets holding online auctions back in 1995. However, online auctions have come a long way since the first market place offerings which closely mirrored their real world counterparts.
In addition to supporting a host of auction formats, such as sealed-bid, reserve price, fixed-price, Japanese, Brazilian, Vickrey, Dutch, and Yankee, auto-extend features such as last-bid time, and auto-termination features, such as minimum reserve price met, today's auctions will allow you to analyze bids on more the just price.
Advanced auction platforms will allow you to define adjustments to take into account differences in a multitude of user-defined quality ratings such as waste, reliability, and on-time delivery and rank bids according to these adjusted quotes. This allows you to either award the business to the bidder with the overall best value relative to your defined adjustments and metrics or give suppliers better feedback if a decision optimization and final negotiation round is to follow.
However, even though auction technologies have been around for a decade, as with RFX, there is still room for innovation. Specifically, I predict that the future leaders in auction technology will incorporate some form of decision optimization directly into the live event. This will allow you to take best practices and total value management into account during the auction and guarantee that the identified awards not only respect all of your business rules but are optimal in that respect. You will be able to define constraints, such as no more then 50% of the volume can come from any supplier and at least two suppliers must be selected to mitigate risk, and global value modifiers, such as a non-incumbent supplier comes with a one-time fixed supplier setup charge of $10,000 (amortized over the entire buy from the supplier), and be confident that the award takes these rules and costs into effect.
But don't expect too many providers to have these capabilities in the near future. As we will discuss at a later time, optimization is hard - very hard, and even today only a handful of vendors actually have real decision optimization capability in their non-real-time analytics offerings. With the exception of CombineNet, which is attempting to claim ownership of the sourcing and logistics space (especially now that MindFlow, like many of its other former competitors, is gone), I know of only one other provider of e-Sourcing technologies working toward this capability. I know of a few that have tried and failed and a few more that decided the potential reward was not worth the perceived risk. However, if you have the skills to pull it off, when you consider how valuable this will be to a sourcing organization with a large spend and limited resources who could bring even more spend under management while automatically optimizing non-critical direct materials during an automated auction, I think the cost, effort, and overall reward is worth the risk.
I would suggest that you do a little more research on Sourcing providers who have come a long way in providing decision support tools including Optimization. There are a number of providers who have incorporated strong decision support tools including realtime optimization.
Steve, I will agree that a number of auction products (I can think of 6 easily) have come a long way in the last 3 years, and that a number of these providers also offer some form of decision support or optimization, but I know of only a couple that incorporate what I label "pure" or "strong" decision optimization" into the live event, with the other products using heuristics and customized algorithms to simulate optimization in the live event (or using a staged event).
I'm also strictly limiting the scope to providers in the (strategic) (e-)Sourcing realm, and not optimization companies or services. The class of offerings would expand a bit if you broadened the definition. With this strong definition, I only know of a handful of companies. They are far from the best known players, and I believe there is still (considerable) room for improvement with degree to which this "strong optimization" is offered and the other capabilities of the tool.
I'd be interested in your list if you care to share it. (I'm only aware of about 50 strategic sourcing and related technology providers, and I know new ones are emerging constantly in this growing space.)
Hi Michael,
I enjoyed reading you article. I agree, there is a big difference between value optimizing sealed bids and real-time value optimization in a dynamic auction. Our company has developed an advanced sourcing auction system with real-time combinatorial optimization. The optimization engine can solve for non-price decision factors such as quality, switching cost, delivery, payment terms etc. It also solves for bid package combinations, contract term options, volume discounts, and supply constraints such as max award volume per supplier or a minimum number of suppliers in the award configuration. The tool is designed to allow expressive bids, where suppliers can change the bid attribute and lot configuration in real-time.
In our experience the most difficult aspect of the problem is building a solver that can compare millions of allocation configuration options in fractions of a second and once the best value option is determined to update each bidder’s bid guidance dashboard to communicate results in real-time. It seems that some companies are focused on heavy-duty optimisation solvers with no real-time interface to bidders, and others are focused on multi-lot advanced auction platforms. But the opportunity, as you say, is the combination of a powerful value optimization engine and a real-time auction platform with simple to use bid guidance capability.
Regards
Anthony Du Preez
Director BOMweb Pty Ltd
Anthony:
Thanks.
I'm curious as to more specifics on the complexity and size of the model that you can handle in real time and the specific categories of constraints you can handle. (Keep watching this blog and ESF ... in the future I will be blogging about what I see as the minimum requirements for true decision optimization, real-time and non-real-time.)
Michael,
I am a bit surprised that you use ‘size of the model’ as a measure of true decision optimization. Instead, we consider the types of real-world sourcing problems that we can solve as the only true measure of the quality of our optimization engine.
We consider a 20% saving as a result of a relatively simple optimization problem as important as a similar saving for a very large very complex optimization problem.
Another measure of true optimization is the ease of use and flexibility of the tool.
To answer your question, our latest product release can optimize millions of total cost options in less than one second.
From a constraint perspective, our iterative combinatorial auction engine allows real-time best value optimization of: multi-variable bids, non-price bid factors, supplier set bid attributes, volume allocation constraints, lot allocation constraints, and lot packages or combinations.
Regards,
Anthony
Anthony:
Although very important, the results produced by an optimization engine are not a true measure of its underlying capability. To give an example that even a layman would understand, it's like trying to compare the towing power of a Dodge Ram 3500 with a Mazda B2300 Regular Series on a 1000 lb trailer.
For all practical purposes, the Mazda - which is much more fuel efficient, smaller, and quick to rev up would appear to be just as good, and maybe even better, than the oversized, fuel guzzling Dodge Ram since it would produce the same results in this scenario, but the reality is that the Dodge Ram is a much more powerful machine. Where as the maximum towing capacity of the Mazda B2300 Regular Series is listed as 1,580 lbs, the maximum towing capacity of the Dodge Ram 3500 is listed at 13,850 lbs. No comparison.
This does not mean that you should always use the Dodge Ram when you need to tow something. After all, in business we are concerned with results and efficiency - and if the Mazda will do the job equally well, then we should use it since its total cost of ownership is significantly lower - but we should never forget its limits!
Similarly, although the tool should be useable, especially if you want it to be adopted by the marketplace, usability is also not a true measure of the underlying capability of the tool. To use another driving comparison, it's like comparing an automatic transmission to a manual transmission. The automatic is easier to use, but not necessarily better ... especially in uneven terrain where an experienced driver can get better performance, and fuel efficiency, out of a manual transmission. (And the best way for your average driver to understand the true value of a manual transmission is to drive a fully loaded 26' UHaul truck through mountainous terrain where the automatic engine is very likely to overheat very quickly.)
Thus, although results and usability are a good measure of the value of the tool, especially in the business domain, they are not a measure of its capability and since true decision optimization depends on the capability of the tool, that is why I use model size and model complexity as these are the best, and most consistent, measures of capability that I have ever found.
As for "our latest product release can optimize millions of total cost options in less than one second", although it sounds impressive, I'd have to say I would expect this from any tool proclaimed to be market leading as I can get the same performance out of a number of generic industrial solvers, including ILog's CPlex, given a good mixed-integer linear programming representation and a dedicated high end server.
In other words, the only thing you've told me that is truly impressive about your tool in this comment is "our iterative combinatorial auction engine allows real-time best value optimization of: multi-variable bids, non-price bid factors, supplier set bid attributes, volume allocation constraints, lot allocation constraints, and lot packages or combinations" - since this tells me that you are able to support at least most, if not all, of the model complexity required to represent real world sourcing problems and your previous statement implies you can handle a significant model size. Even though I expect this level of performance, the reality is that most solutions do not even come close (since it is very hard to implement a "good" representation). In other words, you have a really impressive solution, but the reason I think it is really impressive is apparently not the same reason you think it is really impressive.
Michael
Michael,
Thanks for the very informative response.
Can you name the other two vendors working on a real-time optimized auction system?
Also, who else is working on an iterative (real-time feedback) combinatorial auction system - for sourcing or allocations? My discussions with Paul Milgrom suggest that very few (if any) are doing this today? Is that your understanding?
Anthony
Anthony:
If you have been following eSourcingForum, you will have noticed my recent entry on optimization POE or Bob that noted, as per previous ESF posts, Iasta is currently working on embedding real-time optimization into their auction system. I cannot give you any details at this present time, but you know that Iasta is currently using the ILog CPlex engine and my background (from the About post and Jason's Everything You Wanted to Know About Optimization But Were Afraid to) and can possibly infer one track they may be pursuing.
In many ways, CombineNet is the market leader and the only company that I knew of for the longest time that had a real-time feedback combinatorial auction system based on their concept of an exchange under what they are now calling the "expressive commerce paradigm".
Another company that was pursuing supply chain optimization and has real-time systems is Soligence, but a current review of their websites does not give any indication on whether or not they are still pursuing supply chain. The only other optimization company that comes to mind is Rapt, but they try to maximize profit while most companies try to minimize cost.
The reality is that very few companies do true optimization, and fewer still do real time, because, as I have pointed out in my optimization series (on eSourcingForum), optimization is very, very hard.
Michael