Early adopter strategic sourcing teams are turning to two types of sophisticated mathematical algorithms to give them the edge in sourcing goods and services. Regression analysis (big data) is being used for advanced spend analytics and linear or mixed-integer programming algorithms to optimise supply configurations during the sourcing process. This article will focus on optimisation algorithms for advanced sourcing projects.
Most Australian companies use a sealed bid tender approach to source billions of dollars of goods and services. This well understood approach has withstood the test of time and remains the preferred go to market strategy for most business and government organisations.
With the advent of web based esourcing technologies, most organisations have moved at least some part of the tender process online. From simply compiling the tender documents and emailing to prospective suppliers, to online tender boxes and tools that require suppliers to complete submissions directly into a format that can be stored and analysed. The payoff in moving the tender process online has been partial automation of the event management, bid collection and bid analysis process and reduced sourcing cycle time.
Tenders work because the process easy to understand, and when run correctly comply with most procurement policies from probity and audit perspectives. Tenders are also effective at finding the lowest cost solution for the requirement as specified by the buyer.
On the down side, traditional sealed bid tenders give the suppliers no feedback about their competitive position meaning they are effectively pricing in the dark. Also, suppliers are pricing a configuration (specification, packaging, terms etc) prespecified by the buyer that is not necessarily the lowest total cost configuration. eAuctions partially address the feedback shortcoming of the sealed bid tender process but are constrained once again by a buyer pre-set configuration with no room for supply flexibility and innovation and a one dimensional focus on lowest price.
In response, advanced eSourcing technologies, powered by mixed integer programming algorithms, have seen widespread adoption amongst the leading Fortune 500 US. Companies like Heinz, Proctor&Gamble, and Alcoa. These companies have moved beyond the sealed bid tender (eRFX) and eAuction mindset and are now sourcing most of their centralised spend categories using these technologies.
Mixed integer algorithms have had such a profound impact on organisational cost structures that Aberdeen now publishes a 'leaders and laggards' report specifically on the adoption of Advanced eSourcing technology powered by these algorithms. So what are a mixed integer optimisation algorithms, how do they work and why are they so effective at helping sourcing professionals cut total cost?
Stemming from the operations research field, mixed integer optimisation algorithms are designed to crunch millions of permutations to find the optimal solution to a multivariable problem. The smart maths behind the algorithms is effectively all about fast search algorithms and finding shortcuts to sorting through millions of permutations to solve a problem. These algorithms were initially used to solve real world problems like optimally scheduling everything from manufacturing processes, ads in TV shows and aircraft take-off and loading slots.
Early adopters of eSourcing technologies started including mixed integer algorithms in response to supplier frustration with the limitations of the basic eRFX/eAuction tool. Specifically, suppliers were giving feedback that they were 'straight jacketed' by the eRFX/eAuction event and they were unable to offer their lowest total cost solution across dues to the pre-set configuration.
As an illustrative example, and to highlight the issue, a supplier receives a pricing spreadsheet with 100 items to price, being for 2 components across 50 sites. The question is, does the supplier price each item assuming they will be cherry picked by the buyer, or do they price each item as if they will win all 100 items? This is called the 'exposure problem' and is the reason why suppliers need the flexibility to submit bids with conditions attached. Suppliers that exhibit economies of scale will price the latter assumption lower than the former, and suppliers with dis-economies of scale or 'economies of scope' will not.
Suppliers that exhibit 'economies of scope' tend to be specialised by a region, service, component, material or some qualitative factor such as quality. These suppliers will not be the most competitive across all the items, but should be very competitive against the items that they specialise in. Each supplier's cost economics will be different and probably unknown to the buyer, and it is therefore important to allow the suppliers the flexibility to submit bids with conditions attached.
In other words, advanced eSourcing technologies that use mixed-integer programming algorithms, allow the suppliers to bid 'expressively' i.e. to bid not only a price against a buyer set supply configuration, but also to bid a price for each supplier set configuration. In effect, using an online expressive bidding interface, suppliers can indicate how their price varies based on:
- Qualitative factors (i.e. how price varies based on factors such as: quality, service, delivery, payment terms etc..),
- Quantitative factors (i.e. how their price varies based on volume - overall and by location),
- Bid packages (i.e. how their price varies for each package of items that they have specified), and
- Supply constraints (i.e. how their price varies depending on specified supply constraints)
Expressive bids will depend on the category being sourced, and the unique way that the suppliers cost base is structured. In a nutshell, buyers will get the lowest total cost solution if suppliers can supply goods or service in a way that is most efficient and cost effective to them as long as the supplier's preferred arrangement is within constraints set by the buyer's stakeholders.
Buyers that move to a more flexible or expressive tendering process are faced with a new challenge. How do they evaluate and compare expressive bids when even a few degrees of freedom can result in many millions of award permutations. This is where the maths makes life easy for the sourcing professional. Buyers use mixed-integer optimisation algorithms to test multiple 'what if' scenarios against internal stakeholder constraints. For example, what is the total cost if only one supplier was chosen vs. maximum of 3 suppliers or what if high quality was worth 5% more to the buyer vs medium quality etc. What used to take hours or days to compute, can now be done on the fly with the stakeholders in the room contributing to the analysis.
Once stakeholders agree on an optimal contract award configuration (preferred suppliers, value of qualitative factors, packaging or lot splitting options, reserve price etc.), the buyer can use the technology for a final best offers round from the panel of approved suppliers. The final offers stage allows the buyer to give the suppliers feedback and overcomes a major disadvantage of the sealed bid tender approach.
Feedback gives suppliers the opportunity to finalise their offers as they now have certainty around the approved configuration, reserve prices and their competitive standing on each item. Suppliers use this competitive feedback to refine their offers and in some cases shift their bidding focus to lots where they are most competitive.
In conclusion, advanced esourcing technologies, powered by mixed integer optimisation algorithms, should be used for complex or centrally sourced categories such as transport, print, energy, hotels, packaging, MRO, telecom, services etc. Companies adopting advanced eSourcing technologies need to invest in developing their internal capability for developing sophisticated TCO models, but the payoff for this investment is sustainable and structural total cost reduction outcomes, lower risk supply chains and happier suppliers.
Published in Procurement Professional Magazine, June 2013.