(Guest Post) Why Optimization Matters: Moving Beyond Price Driven Auctions

By Dr. Olga Raskina, Lead Scientist, Emptoris

This article is the first in a three-part series of guest articles on optimization by Dr. Olga Raskina, Lead Scientist, with Emptoris, the supply and contract management solutions provider.

As the Lead Scientist for Emptoris and someone who spends a great deal of time focusing on the impact optimization technologies can have on business processes and businesses, I am often asked “Why Optimization Matters?” Optimization sometimes sounds like a difficult topic, even though the idea and technology are meant to make processes and decisions easier – and more efficient.

Why does Optimization matter? Let me illustrate if I can. Recently eBay changed the policy regarding its buyer and seller rating system. One of the changes was to remove the ability to leave negative feedback for the buyers. The notorious eye for an eye rating system will soon be history.

At first I was surprised by this development. I don’t sell often on eBay, but for anyone who does, the rating system was the only way to know anything about your buyer. Then I realized that in reality this just didn’t matter. The rules of any eBay auction are such that the only criterion to select the winner is price. A seller who offered the best price, with a few simple restrictions like being above the buyer’s reserve price, or having a valid shipping address, wins the auction, regardless of rating or any other qualities. While the rating might mean something to the seller, the policy is just not built to accommodate it.

Granted, price-only auction is the best choice for a one-time sale of those barely used pair of skates. Such auctions are small and low in complexity.

At the other end of the spectrum there are events with quite a bit more at play; where things like participant ratings could make a difference. These include large procurement events, with thousands of items, and a large number of complex parameters far beyond simple price.

Transportation procurement events are a perfect example. Companies will procure services for tens of thousands of lanes and evaluate the providers based on dozens of factors, from carriers’ capacity and shipment time, to fuel surcharges, insurance costs and container types. The sheer amount of data to be collected in these events excludes the possibility of conducting them in a real-time auction manner.

Such events are usually sealed and multi-round, stretched over weeks or even months. Typical cycle involves suppliers analyzing the requirements and putting together a proposal; buyers collecting and analyzing these proposals, and then potentially repeating the cycle (possibly with some feedback or recommendations to the participants).

These two extremes – low complexity auctions and high complexity multi-round events are usually the only two options that are considered for the majority of actual procurement events, leaving everything in the middle to the mercy of fate.

The events that I find particularly interesting from this perspective are small size high complexity auctions, with only a few items, and a multitude of interrelated factors affecting the final event outcome. These could be auctions for a complex commodity, for example, bulk quantities of computer hardware, where one would need to account for factors like material, labor, and overhead prices, transportation and storage costs, as well as reliability, performance, and service levels. Events of this type would usually include not more than five to ten items and half a dozen participants.

The size of these events allows the advantage of fast-paced real-time auctions. Open competition and bidding frenzy can yield multiple attractive offers. Yet the complexity of the items and requirement calls for involved analysis of the offers. Decisions are (or should be) made considering the combination of all the factors.

With only two options available to them, many buyers opt for either price-driven competition or an analysis-driven multi-round event. Both approaches will not realize the full potential of the event and the buyers will still leave money on the table.

Price-only auctions ignore all other factors, and the outcome, while low on price, could be very high on the utility costs like quality, item specifics, or supplier performance. Sealed multi-round events, on the other hand, will not encourage fierce competition.

More savvy buyers try to combine the better of the two worlds, first collecting the bids through the (price-driven) auction, and then conducting the analysis of all other factors. At the first glance this strategy seems to be a perfect solution. More thorough study, however, reveals two major shortcomings of this approach.

First, price-only competition does leave out all other factors, even though these factors are still part of the bids and subsequent analysis. The absence of the competition leaves all other elements unchanged. The analysis will select the best combination of bids for the buyer’s policy, but the buyer will miss the opportunity to get better overall deals.

Second, the participants could feel deceived, as they were forced to lower their prices, yet some or all of the business was awarded on other criteria and they were not given an opportunity to provide counter-offers on other elements. Had they known about other bidding dimensions, they would have had a chance to provide different competitive offers, which might not always require reducing price, but could for example mean better quality, or faster delivery.

Another example could be a buyer’s policy to single-source a group of items. The most favorable response for the suppliers would be to focus on bidding on all items in the group, rather than further squeezing the prices on one or two items. Information on where they stand with respect to price only is not nearly enough to make these types of decisions.

Such events require an entirely new strategy; one that will allow a real-time competition on multiple dimensions, and will provide real-time analysis of all aspects of the buyer’s awarding policy. This is where optimization can have an impact on sourcing/auctions.

Such a strategy offers the participants the most up-to-date information about the current status of their bids with respect to such policy. If in addition the buyer will choose to share some of the policy details, the suppliers will be equipped with a wealth of information on what factors affect the buyer’s decision making. They will be able to see immediately where they stand in the event and respond with various complex offers along their strength points.

Implementation of an optimized auction strategy of this kind will, of course, demand a commitment from all players. Buyers will need a deeper understanding of the structure and driving factors for each event and will have to be able to develop an appropriate strategy in each case.

They will also need to think ahead and narrow down the awarding policy as a part of the event set-up. The set-up will have to be thorough and realistic. For example, if the buyer decides to dual-source a group of items, they will need to ensure an adequate number of suppliers are invited to bid on these items. If an item quality is a decision factor, the suppliers should be required to provide the quality description.

Buyers will need better, more extensive communication with their suppliers to explain the bidding rules and the awarding policy for each event.

Suppliers will also need to think ahead about what values and combinations of bidding elements are acceptable for them, as there will not be much time to collect and analyze this information during the real-time auction bidding.

They will have to implement new or different revenue management policies and modules, and possibly integrate them with the auction tools for fast evaluation and response.

Last, but not least, the sourcing software solution providers need to develop adequate real-time auction analysis tools. Most analytic modules are rarely built with an instantaneous response in mind, both in terms of the science and the implementation. These types of analytic tools require novel approaches in many areas, from front-end consistent multidimensional presentation to back-end scientific machinery.

Off-line auction analysis is hard in itself, but on-line analysis is an order of magnitude harder. While there has been a significant research and technology creation effort devoted to standard off-line auction analysis and optimization, investigations into these topics, but the scientific community is yet to embrace the necessity of business-driven real-time auction analytics.

Unlike at eBay, changes in all of this will not come quickly. Buyers, suppliers and technology will have to graduate to a new level of sophistication. Yet, we already observe the procurement leaders starting to realize real benefit from efforts in advanced optimization. And technology leaders continue to invest in developing cutting-edge optimization tools to facilitate growing and ever more complex demand.

Dr. Olga Raskina, Lead Scientist, Emptoris, Inc.

Olga has been an active member of the Institute for Operations Research and Management Science (INFORMS) for many years, and serves as vice-chair on the Boston Informs and on the Informs Subdivisions Council, focusing on promoting the value of Operations Research to businesses. Olga has Ph.D. degree in Operations Research from Columbia University.

Editor’s Note: As is the case will all guest posts I would like to stress that neither Emptoris’ appearance in the Procurement Insights Blog, nor the positions presented by Dr. Raskina in the article are to be construed as an endorsement by Procurement Insights. That said I can speak from first hand experience as the lead researcher/developer of an initiative that was funded by the Government of Canada’s Scientific Research and Experimental Development (SR&ED) program, that the points introduced by Dr. Raskina are both relevant and insightful.

Referencing an agent-based versus equation-based model, the initiative’s research of the utilization of advanced algorithms in the procurement process eventually led to my formation of the theory of strand commonality. The premise behind the strand theory is that unique attributes within seemingly diverse strands or streams of data can ultimately be identified, linked and measured to produce a collective “best result” outcome. When weighted-value parameters were introduced into both the development and production models to measure historic data with real-time conditions at the specific point in time when a purchase was made the results were telling. In more than 98 percent of all transactions, the system consistently selected the best supplier. This was reflected in marked improvements in supplier delivery and quality performance levels and significant price reductions that did not negatively impact supplier profitability nor increase the risk for supply base erosion. It should be noted that both the development and production models focused almost exclusively on Indirect MRO Materials.

What this means is that Emptoris’ solutions are structured around a model that is more reflective of a true Software as a Service (SaaS) or on-demand application. This of course is in stark contrast to the equation-based models utilized to develop the ERP-based applications that are usually offered under a traditional (and considerably more expensive and less efficient) licensing program. In short, on-demand or SaaS is more than just a price schedule, and Emptoris is more than just another vendor.