Abstract
Sourcing professionals buy several trillion dollars worth of goods and services yearly. We introduced a new paradigm called expressive commerce and applied it to sourcing. It combines the advantages of highly expressive human negotiation with the advantages of electronic reverse auctions. The idea is that supply and demand are expressed in drastically greater detail than in traditional electronic auctions, and are algorithmically cleared. This creates a Pareto efficiency improvement in the allocation (a win-win between the buyer and the sellers) but the market clearing problem is a highly complex combinatorial optimization problem. We developed the world's fastest tree search algorithms for solving it. We have hosted 16 billion of sourcing using the technology, and created 1.8 billion of hard-dollar savings. The suppliers also benefited by being able to express production efficiencies and creativity, and through exposure problem removal. Supply networks were redesigned, with quantitative understanding of the tradeoffs, and implemented in weeks instead of months.
Introduction: The Trillion-Dollar Sourcing Puzzle
Sourcing—the process by which corporations procure goods and services—is a multi-trillion dollar industry. Historically, this process relied on manual, one-on-one negotiations, which were expressive but slow and lacked competition. The 1990s saw the rise of "reverse auctions," where suppliers bid down prices on pre-defined "lots" of items. However, these auctions are often economically inefficient because they force items into arbitrary bundles before supplier preferences are known.
For example, a trucking company might offer a significantly lower rate if they win both a "to" and "from" route, avoiding empty miles. Traditional auctions cannot easily capture these conditional efficiencies, leading to the "exposure problem," where suppliers bid conservatively to avoid losses on partial wins. Furthermore, buyers often have complex business rules—such as limiting the total number of winners or ensuring spend with minority-owned businesses—that simple auctions cannot enforce.
The Expressive Commerce Framework
Expressive Commerce addresses these inefficiencies by embracing complexity rather than avoiding it. The system allows for a "win-win" scenario where savings come from reducing economic waste (e.g., optimizing logistics) rather than just squeezing supplier margins.
The architecture consists of a user interface application where buyers manage events and analyze results, while suppliers submit expressive bids. These two sides are bridged by the Combinatorial Exchange Description Language (CEDL), which feeds into the ClearBox clearing engine. This engine performs automated problem formulation and sophisticated tree searches to find the optimal allocation.
Technical Methodology: The ClearBox Engine
The core technical challenge is the Winner Determination Problem, which is NP-complete. To solve this at an industrial scale, the ClearBox engine utilizes specialized tree search algorithms that can be up to 10,000x faster than general-purpose Mixed Integer Programming (MIP) solvers. Key techniques include:
- Information-theoretic branching: Selecting variables to split on based on their ability to reduce uncertainty.
- Dynamic decomposition: Breaking massive problems into smaller, independent components on the fly.
- Lookahead and Primal Heuristics: Rapidly finding "good enough" solutions to prune the search tree.
- Symmetry Breaking: Identifying and ignoring redundant parts of the search space.
Real-World Impact and Results
The system's efficacy has been proven through massive real-world industrial sourcing events. It has handled events with over 2.6 million bids on 160,000 items and 300,000 side constraints. In one collaborative sourcing event, retail giants P&G, Walmart, and Target allowed carriers to bid across their combined demand, creating unprecedented efficiencies.
Key results include:
- Financial Impact: The technology has hosted $16 billion in sourcing spend, resulting in $1.8 billion in "hard-dollar" savings.
- Computational Speed: In a head-to-head trial, the system solved a complex trucking optimization problem in 9 seconds, compared to 30 minutes for a competing product.
- Operational Efficiency: The "time to contract" was reduced from months to weeks by eliminating manual lotting and post-auction negotiation.
Conclusion and Future Directions
Expressive Commerce bridges the gap between traditional negotiation and electronic auctions. While highly successful, future work aims to automate "scenario navigation," helping buyers find the most robust solutions among various "what-if" scenarios with minimal regret. Additionally, as computational extremes still exist for certain complex instances, further refinement of the underlying optimization algorithms remains a priority.
Summary updated