In our first post, Part 26, we noted that, after covering e-Procurement, Spend Analysis, Supplier Management, and Contract Management, it was finally time for Strategic Sourcing. When it comes to Sourcing, we have to deal with the ORA et labora. The work, and the prayer (that it gets the results we want). But at least when it comes to the prayer, we have three tools at our disposal:

• Optimization
• RFX
• Auction

In Part 27 we started with the most classic sourcing tool, RFX, where RFX stands for Request for X, where X could be Bid, Information, Proposal, Quote, etc. depending on the depth of response required and the terminology used in the industry and geography the RFX is being issued in.

Then, in our last post, Part 28, we continued with the primary alternative to RFX, e-Auction. In e-Auction, instead of asking for quotes which will be reviewed in a long, detailed, often weighted process, you’re asking for real-time quotes in an online auction where you can update your bids until you self-select to drop out.

The last tool at our disposal, which does require bids to be collected first (which does not need to be through RFX or e-Auction but can be done through every buyer’s favourite tool, Excel), is strategic sourcing decision optimization. It’s not used nearly enough considering that it will practically always identify a lower cost scenario, and even if you find the lowest cost scenario impractical, you understand exactly how much more a relationship is costing you and you are quantifying how much a better relationship, better quality, lower risk is worth to you and can make more informed, and better, decisions in the future.

BASIC

Pillar #1: Solid Mathematical Foundations
The algorithms used must be sound (mathematically correct in all situations) and complete (capable of analyzing all possible solutions). An optimization engine based on Mixed Integer Linear Programming (MILP) would qualify as hybrid simplex approaches will provably converge on an optimal answer given sufficient time (and one can always compute a maximum distance from optimal based upon the calculations done to date since the longer the algorithm churns for, the more the lower bound on the optimal solution increases). In contrast, the application of many heuristic, simulation, or evolutionary approaches are likely not valid since the majority of these techniques do not guarantee full exploration of the potential solution space and, therefore, aren’t guaranteed to find the true optimal solution (although they may get close).

Pillar #2: True Cost Modelling
The model must allow you to define the full cost model, not just one (or two) fixed costs. For example, if a buyer is sourcing direct material, the platform must allow the buyer to include all indirect and incurred costs, such as freight, tariff, storage, processing, and marketing differential costs in the definition of the cost model.

Pillar #3: Sophisticated Constraint Analysis
The model must allow the buyer to build a model that capture a realistic approximation of real world constraints. If the business must select at least 2 suppliers, will not accept a product mix with an average quality or reliability of less than 8 (/ 10), if a supplier has a maximum capacity, or if a minimum allocation must be given to an incumbent because of a contract still in play, all this needs to be captured.

A strategic sourcing decision optimization platform must support four core constraint types. Capacity constraints that define a supplier (‘s location) capacity limit. An allocation constraint that defines a minimum or maximum allocation to a supplier (group) based upon existing contracts or business policies. Risk Mitigation constraints that ensure that business policies on supplier or geographic splits designed to reduce risk are captured. Qualitative constraints that allow for qualitative ratings such as reliability, quality, relative sustainability, etc. on a mathematical (e.g. 1 to 10) scale to be defined.

Pillar #4: What If Capability
The platform must support the creation of multiple what-if scenarios, each with different constraints. Buyers should be able to create them from scratch, or as modified copies of existing what-if scenarios.

Out-of-the-Box Scenarios
The solution should contain multiple out of the box scenario definitions, including unconstrained, x-supplier, incumbent, etc. that automatically generate these what-if scenarios for the bids being evaluated for optimization.

Scenario Comparison
The solution must contain a built-in capability for (side-by-side) scenario comparison that allows a buyer to easily see the cost differentials and get a feeling for what each scenario is costing them.

ADVANCED

Integrated Analytics
Optimization models take exponential time to solve. While small models can solve in minutes, and even seconds, on a high powered multi-core machine, large models can take hours or days. The key to rapid model solution is minimizing model size. This can often be done by way of a preliminary analysis that determines that some supplier bids are just to high to ever be acceptable, some qualitative factors too low to ever be acceptable, and some supplier locations are in geographic regions that are just too risky. Eliminating award possibilities that will never be made can drastically decrease model size and solution time.

Constraint Relaxation
If a model is unsolvable, but could be solved by solved with lesser constraints, the platform should be able to identify which (near) minimal constraint set is preventing a solution and identify which (minimal) relaxations would allow a solution and present those to the user, who can accept them, or use that as input for defining an alternate relaxed model that may permit a solution. (Remember best practice is to prioritize constraints and add them incrementally until the model becomes unsolvable as that allows you to always choose the least important constraints to relax for solvability.)

Sensitivity Analysis
In optimization, a sensitivity analysis tells you how dependent a solution is on a certain constraint and what the impact of removing the constraint that is currently preventing a lower cost solution in terms of hard dollars. (For example, insisting an incumbent supplier get 50% of the award might be costing you $10 Million in a $100 Million category, while reducing the minimum to 25% might only cost you $2 Million [as it the supplier is more competitive on some products than others].)

Hard and Soft Constraints
The platform should allow you to define constraints as hard and soft. When a model is unsolvable and needs to be relaxed, the solution will only allow soft constraints to be relaxed. Furthermore, it should also allow for an indication of when a soft constraint can be relaxed. For example, average quality can only be reduced from 9 to 8 if the savings increases by at least 3%.

Integrated Freight Model Support
In addition to supporting true cost modelling, the platform should also have built in freight models that understands transport types and modes (truck vs rail, refrigerated vs dry, etc.) and allow for the easy definition of complex freight models when those models might allow for overall lower costs of ownership when carrier bids are also included in the model.

Of course, this is not a complete list of what a strategic sourcing decision optimization platform might have, or necessarily should have, as systems continue to improve, but a baseline of what they must have to be considered a modern solution.

Next up: the vendor list in Part 30.