Category Archives: Decision Optimization

The 12 Days of X-emplification: Day 3 – Optimization

As I highlighted in Questions to Ask your Optimization Vendor, not all optimization vendors are equal … and, more importantly, not all vendors that claim to have decision optimization even have it! Thus, not only is it important to know what to look for when searching for a true strategic sourcing decision optimization, it’s also vital to know what to ask and what answer you want to hear.

In this post I’m going to cover five key questions that you should ask of every vendor you are considering. Some of these overlap the questions I x-emplified in my previous post (which you should re-read), and a few of them are new. Of all of the topics I am covering, this is probably one of the least understood – and since this is one of the few technologies with the capability to allow you to reduce your spend year-over-year when properly applied – this situation has to change.

1. Does the application satisfy the four pillars of strategic sourcing decision optimization?

As outlined in the Strategic Sourcing Decision Optimization wiki-paper on the e-Sourcing Wiki [WayBackMachine], the four pillars of strategic sourcing decision optimization are:

  • Sound & Complete Solid Mathematical Foundations
    such as simplex algorithms and branch-and-bound;
    many simulation and heuristic algorithms do not guarantee analysis of every possible solution (sub)space given enough time, and, thus, are not complete in mathematical terms
  • True Cost Modeling
    many bidders bid tiered bids, discounts, and fixed cost components – the model must be capable of supporting each of these bid types
  • Sophisticated Constraint Analysis
    At a minimum, the model must be able to support reasonably generic and flexible constraints in each of the following four categories

    • Capacity / Limit
      allowing an award of 200K units to a supplier who can only supplier 100K units does not make for a valid model
    • Basic Allocation
      you should be able to specify that a supplier receives a certain amount of the business, and that business is split between two or more suppliers in feasible percentage ranges
    • Risk Mitigation
      let’s face it – supply chains today are all about risk management, and you should be able to force multiple suppliers, geographies, lanes, etc. to mitigate those risks without specifying specific suppliers, geographies, lanes, etc. to take advantage of the full power of decision optimization
    • Qualitative
      A good model considers quality, defect rates, waste, on-time delivery, etc.
  • What-if Capability
    The strength of decision optimization lies in what-if analysis. Keep reading.

2. Does the application support the creation and comparison of multiple what-if scenarios?

The true power of decision optimization does not lie in the ability to find a solution to one model, but the ability to create different models that represent different eventualities (as this will allow you to hone in on a robust and realistic solution), to create different models off a base model plus or minus one or more constraints (as this will help you figure out how much a business rule or network design constraint is really costing you), and to create models under different pricing scenarios (to find out what would happen if preferred suppliers decreased prices or increased supply availability).

3. Does the application automatically identify the most constraining and costly constraints?

Let’s face it, not every constraint has a significant impact on the optimal solution, if it even has an impact at all. Restricting the highest cost supplier to 30% of the total award is unnecessary if the supplier is not going to get any award. However, restricting the lowest cost supplier to 20% of the award could be the most restrictive constraint in the scenario, as the supplier would get 80% of the award otherwise.

The solution should identify, in order of decreasing impact, which constraints are having the greatest effect on the optimal solution and, at the very least, provide a range estimate of how much the constraint is costing you in the model. Determining the constraints that significantly impact a scenario can be done deterministically – they are at their bounds. Determining the constraints that impact a scenario moderately can be done through a deterministic comparison with the optimal solution to the “unconstrained model” (where only supplier capacities, demands, and cost constraints are included). The rest of the constraints then impact the model slightly or not at all. Calculations that take into account the differences between the optimal solution to the model and the optimal solution to the unconstrained model can be used to provide a reasonable estimate of the cost of any particular constraint. Furthermore, an exact cost associated with the removal with any constraint subset can be determined by optimizing the modified model. This brings us to …

4. Does the application support the automatic creation and solution of relaxed and perturbed scenarios?

After the constraints with the most significant impact, particularly from a cost (or risk) perspective have been identified, it’s only logical that you want to know not only how much they are costing you, but how much a relaxation (as opposed to a removal) of the constraint would save you. For example, if you allocated 30% of an award to a new vendor vs. 20%, what would you save? The reality is that you really want to understand not just the cost, but the “cost per unit” of the constraint. If you have allocation splits, you want to know the effect of minor and moderate changes to the splits. If you have limit constraints, you want to know how much you could save with increased capacity (and, thus, whether the company should be making an investment into new technology or more production lines or entering into a strategic partnership with a key supplier to lock up more capacity). If you have qualitative constraints, could you save more if vendors increased their quality by 10% across the board (which is equivalent to allowing a 10% decrease in quality level in the model)?

For each constraint type, it’s pretty easy to come up with a standard set of “perturbations” that you would want to analyze using what-if analysis. The application should support standard perturbation templates that you can use to set up an over lunch (or overnight) run against a well-formed what-if scenario that would generate a variance report that would tell you not only what constraint relaxations would save you the most, but how much a per unit perturbation against the constraint would save you and let you hone in on an award allocation that will have the lowest total cost of ownership over the life of the contract – and not just on the day you run the what-if scenario.

5. Does the application support make-vs-buy and arbitrary product substitution?

If you’re only sourcing indirect, you might not care about make-vs-buy, but you should care about product substitution. Let’s say you’re a major player in the food service industry who caters to the average joe’s love of pizza. Since few pizzas are made without tomato sauce, you’re going to need a lot of it. But guess what – if you ask a supplier for “sauce” they’re going to say “how much”, “what type of packaging”, and “refrigerated or frozen”? Chances are there are 10 different “products” for you to choose from. And it’s not as easy as just saying “whatever is cheapest” or “I’m standardizing on form factor 27” because “whatever is cheapest” will vary by production plant (some types of packaging will be cheaper in some countries than others, some plants have newer technology for certain kinds of packaging, and packaging weights and volumes determine shipping costs). Furthermore, the availability of products is probably going to vary across locations. The way to get the lowest cost is to allow a supplier to bid ALL products that can meet your needs (and, of course, account for any variances in processing costs through adjustments). Thus, you want your strategic sourcing decision optimization solution to support arbitrary product substitution.

However, if you’re sourcing direct, you’re really going to want make-vs-buy analysis. Let’s go back to Mr. Martyn’s automative seat example. Do you source the seat? Do you source seat components and assemble them? If so, how do you define a “component”? Or do you source all the raw parts and assemble them? The reality is that even in this simple problem, you have over 1000 different options. Thus, creating a model where you source the finished components, creating a model where you source specific sub-components, and creating a model where you source all the parts, and doing a comparison report on their optimal solutions just doesn’t cut it. You might find that you save 10% by sourcing the components and having a third party assemble them, but who’s to say that there isn’t a different configuration of components that would let you save 20%? If you gave your suppliers the flexibility to choose their own components and the third party assemblers the opportunity to bid on the components they think they could assemble into the final product most cost effectively, who knows what innovation they might identify? In a make-vs-buy scenario, you can’t assume you know what the proper subset of models to analyze is. You really need to analyze ALL the options available to you.

the doctor Goes Mental On Optimization Myths

In my last post, I went mental on three of the most dangerous myths out there with respect to e-Auctions. In this post, I’ll attack some of the more brain-dead myths that are out there with respect to optimization. For more great information on decision optimization, I recommend checking out the e-Sourcing Wiki [WayBackMachine] paper (Strategic Sourcing Decision Optimization: The Inefficiency Eliminator) that was originally authored by the doctor, the doctor‘s joint podcast with Next Level Purchasing (Parts I and II and Free Transcript with Editorial Notes brought to you by Sourcing Innovation and Next Level Purchasing [now the Certitrek NLPA]), and the posts in the Decision Optimization category here on this blog.

Myth 1: I need a PhD to use optimization!
It used to be the case that you needed an advanced degree to use the overly complicated command-line tools that represented the first generation of commercially available optimization products, but that hasn’t been true for quite some time. Today, companies like Emptoris (acquired by IBM, sunset in 2017) and Iasta (acquired by Selectica, merged with b-Pack, rebranded Determine, acquired by Corcentric) offer very simple wizard-driven user-interfaces that can be driven by any business analyst or sourcing professional which are often easier to use then your ERP or BI tools!

It’s literally as simple as selecting the relevant auction or RFx data, defining your demands for each item at each distribution center, identifying invalid freight lanes, specifying supplier capacity restrictions, identifying any business rules (such as dual-supply, 70-30 split) and defining any discounts or “preferred award” valuations (if I buy from Quality Delivered, the joint marketing campaign will be worth $10K). Then you click “optimize” and the optimal award for your scenario is spit out. If you don’t like it, you can copy the scenario, add or remove some constraints, and see what your idea of an optimal award is costing you and make the smart decision.

Myth 2: I can’t afford optimization! It’s too expensive!
Having been involved in this industry for a while, I know that early solutions were very expensive – usually starting in the seven figure range for an average company. But that’s true for every new generation of technology, software or hardware, it’s costly at first, as companies need to recoup their massive R&D investments, but gets cheaper over time. Today, a mid-size company can get a true enterprise quality strategic sourcing decision optimization solution for its sourcing department starting in the quarter-million to half-million six-figure range – and this will include a (shared) C-Plex license and (shared) dedicated hardware resources if they use an on-demand model such as that offered by Iasta.

Myth 3: My problem’s too large / complex for optimization.
Again, the technology has come a long way in the last decade. Not only can massive problems (which couldn’t have been solved in months a decade ago) now be solved in a matter of hours, but the types and quantity of constraints available have greatly increased. If your model is truly humongous, just remember that both CombineNet (acquired by Jaggaer) and Algorhythm regularly solve problems that take millions of variables and hundreds of thousands of equations to specify. As for complexity, even the solution by the relative newcomer, Iasta (which has adopted a best-to-market strategy) supports the four basic categories of constraints required for true decision optimization (capacity, allocation, risk mitigation, and qualitative), flexible discounts that will allow you to implement just about any cost structure you can devise, and freight costs for a true total landed cost model (as you can also define adjustments to capture utilization costs).

Myth 4: We’re very sophisticated when it comes to e-Auctions. We’re not going to save enough money with optimization to make it worthwhile.
Wayne Campbell said it best when he said and perhaps monkeys will fly out of my butt!“. Although it’s theoretically possible that you could be making the perfect buy – every time – without any decision optimization, in reality, the chance that this is true is about as close to 0 as you can get. I’ve NEVER encountered a situation where optimization didn’t provide a company with a cost savings opportunity in any moderately complex bid (and, these days, what bid is not moderately complex?). Furthermore, Aberdeen has found, both in their 2005 study AND their 2007 study, that advanced sourcing and negotiation methods, which includes decision optimization, saves a company an average of 12% beyond what they would save just using e-Auctions. That’s a lot of cash you’re leaving on the table.

There are more myths, but this is a good start – and hopefully enough to convince you to check these solutions out. Decision optimization for strategic sourcing is worth the investment.

Algorhythm and the Optimization Rhythm in India

Recently, I had the pleasure to have a couple of conversations with Ajit Singh, the Founder and Director of Algorhythm, a company in Pune, India that has significant expertise in Optimization and Supply Chain Modeling. The have their own optimization engine, a set of front-ends for different types of supply chain models that can be used by anyone with modeling skills, and significant experience in helping large global multi-nationals with significant supply chain network design and optimization problems. Basically, they’re India’s CombineNet, but with a slight distinction – every model they build, including custom models, can be executed and modified completely by the client through an extension of their easy-to-use windows-based front end – you are not tied to their services. In comparison, although CombineNet has done a great job over the past few years of actually building stand-alone products and interfaces, it’s still often the case that custom models are only available through their services model.

Algorhythm has the capabilities to attack both strategic and tactical supply chain problems from an optimization and simulation perspective. They have sophisticated models for strategic planning that include inventory optimization, distribution network design, manufacturing network design and for tactical execution that include production planning, logistics planning, and supply network execution.

They also have specialized solutions for oil, steel, and packaging as well as having a considerable amount of experience in creating models for manufacturers and distributors. Major clients include Unilever (Hindustan Unilever, Unilever Plc. UK, and Unilever China), Thyssen Krupp, Hindustan Petroleum, and Parle Products among dozens of others. Their manufacturing and distribution network design models often save their clients 3-5%. Remember that we’re talking production models here – not sourcing models, so this is actually quite good. In terms of efficiency, their production planning and scheduling models often halve throughput time and inventory carrying requirements – which is also very good. Furthermore, we’re not talking small models here – Parle, for example, ships 50K trucks per year per SKU from hundreds of factories to thousands of wholesalers.

It’s quite easy to build a model in their products, which they call Prorhythm (for production-planning based models), Netrhythm (for network-planning based models), and Logrhythm (for logistics planning models), and which run on top of their Xtra Sensory optimization engine. They’ve thought through what the model is, what the core elements are that make it up are, what the costs are, and what measures you might want to optimize. Building a model is simply defining all the relevant entities (which are factories, lines, outputs, inputs, etc. in production planning), the associated costs (material, labor, overhead, etc.), the measure(s) you want to optimize (cost, throughput, etc.) and their priority / weighting if multiple, and the constraints. It assumes all relationships between related entities are valid unless you specify them as invalid (and permits groupings for easy constraint definition). It also groups constraints in a “constraint file” so you can easily run the same model against different constraint sets. Basically, it’s built to build models the way the doctor would build it.

Since there is no “one” optimal solution when you’re optimizing against multiple objectives, as it’s almost always impossible to precisely normalize each measure to a uniformly distributed 0-1 interval that can then be weighted according to the weights you want, they also support simulation. You can tell the optimizer to construct a set number of models equally distributed around the desired optimization point and it will automatically create and run all of the variants which you can then compare to see how slight changes impact solutions and goals.

It’s a great offering, and the people are quite knowledgeable. If you have a tough optimization problem, be sure to check them out. They might surprise you.

Questions to Ask your Optimization Vendor

Not all optimization vendors are equal … and more importantly, not all vendors that claim to have decision optimization even have it (as their systems barely qualify as decision support). Thus, since Emptoris [acquired by IBM, sunset in 2017] just released a new version of their offering, since Iasta [acquired by Selectica, merged with b-Pack, rebranded Determine, acquired by Corcentric] is coming out with their first heavy-hitting release in the next month, and since CombineNet [acquired by Jaggaer] is always working on something new, it’s important that you be able to distinguish between the relative strengths and weaknesses of the different products, as well as how much strength you really need, if good decision optimization is one of your driving reasons for selecting a (new) e-Sourcing solution. (And, by the way, it should be!)

Now, as I indicated in a comment over on Spend Matters (in “Emptoris 7 Pushing the Sourcing Envelope”), I’m not going to devote a post analyzing the new Emptoris announcement at this time, as I don’t yet have enough data points to even make a half-assed*0 attempt (although I do feel I have a pretty good idea precisely what they did based upon their choice of wording, the amount of time they’ve been working on it, and my perception of their in-house skill level), but I really think you should analyze it, just as you should analyze any other vendor’s solution, before buying it. (Not necessarily because I don’t think it will do the job, but because the key with optimization is buying just what you need in the majority of your sourcing events. Optimization is expensive. Buying too much power could severely impact your potential ROI, and buying too little power will be equivalent of flushing that investment down the drain as it won’t solve the majority of your problems. I’m using the word “majority” because there is no general purpose decision optimization product for sourcing that will handle all of your events and solve all your problems. As with just about everything else in business, it’s the 80-20 rule. The best solution is the one that solves as close to 80% as possible at a cost of ownership that maximizes your ROI multiple. You can always do one-time projects with best-of-breed providers or specialist outsource providers for those projects in the remaining 20% where there is enough of a savings opportunity.)

Before I get to the question list, I should point out that it’s almost impossible to cover every question, as many of the questions you should be asking depend on the answers you receive to your first few questions, but I think the question list below is a good starting point. If I get some good feedback, and some more free time, I’ll consider doing a part II at a later date. So, without further ado, here’s the starting list!

  1. Does your product meet the four critera for strategic sourcing decision optimization as outlined in the Strategic Sourcing Decision Optimization wiki-paper on the e-Sourcing Wiki [WayBackMachine]  (initially authored by the doctor, the all-knowing optimization guru*1)? Specifically, does it support the following:
    • Sound & Complete Solid Mathematical Foundations
      such as simplex algorithms and branch-and-bound;
      many simulation and heuristic algorithms do not guarantee analysis of every possible solution (sub)space given enough time, and, thus, are not complete in mathematical terms
    • True Cost Modeling
      many bidders bid tiered bids, discounts, and fixed cost components – the model must be capable of supporting each of these bid types
    • Sophisticated Constraint Analysis
      At a minimum, the model must be able to support reasonably generic and flexible constraints in each of the following four categories

      • Capacity / Limit
        allowing an award of 200K units to a supplier who can only supplier 100K units does not make for a valid model
      • Basic Allocation
        you should be able to specify that a supplier receinves a certain amount of the business, and that business is split between two or more suppliers in feasible percentage ranges
      • Risk Mitigation
        let’s face it – supply chains today are all about risk management, and you should be able to force multiple suppliers, geographies, lanes, etc. to mitigate those risks without specifying specific suppliers, geographies, lanes, etc. to take advantage of the full power of decision optimization
      • Qualitative
        A good model considers quality, defect rates, waste, on-time delivery, etc.
    • What-if Capability
      The strength of decision optimization lies in what-if analysis. Keep reading.
  2. Does it support the creation of multiple what-if scenarios and does it simplify the creation of these scenarios?
    The true power of decision optimization does not lie in the model solution, but the ability to create different models that represent different eventualities (as this will allow you to hone in on a robust and realistic solution), to create different models off a base model plus or minus one or more constraints (as this will help you figure out how much a business rule or network design constraint costs you), and to create models under different pricing scenarios (to find out what would happen if preferred suppliers decreased prices or increased supply availability).
  3. How fast is it for different average model sizes and can performance be tweaked?
    Optimization takes what it takes. That being said, if one solution takes an average of 1 hour for an average scenario, and another solution takes 10 minutes, all things being equal, if you have compressed sourcing cycles, the 10 minute solution might be better. Emphasis on “might”. This is only true if the faster solution is of the same quality – some models, and some solvers, sacrifice quality and accuracy for speed. The best solution will let you trade off “tolerance” and accuracy for speed. Sometimes it’s easy to get within 1% or 2% in a few minutes, even though that last 1% or 2% could take hours. On a model with low total savings potential, getting within 1% may be enough. And when trying to hone in on the right what-if scenario, it’s nice to get within 1% quickly and then allow the right scenario to run to completion over night after you’ve quickly analyzed half-a-dozen scenarios and settled on your preferred scenario. Thus, tweaking ability is very important.
  4. If it supports “real-time” is it “true” real-time or “near” real-time.
    Thanks to significant advances in processor and hardware performance as well as off-the-shelf optimizer technology (like ILog’s CPlex), it’s now possible to rapidly re-build and re-solve moderately sized models using off-the-shelf modeling languages in seconds, allowing for e-auction tools that keep the model relatively small and simple to incorporate decision optimization in near-real-time by simply re-building and re-solving the model every 30-60 seconds (depending on model-size) on a high-powered dual or quad core server with an appropriately configured and optimized CPlex 10. However, this is NOT true real-time optimization and could rapidly break down if the model gets too big or too complex. (For example, real-time optimization requires the ability to merge model construction and model solution in such a way that a new bid can be introduced as a parameter change that does not require the optimizer to rebuild the sparse model matrix and start the solution process over from scratch.)
  5. Describe two or three scenarios you have encountered where you could not model the situation exactly for companies in our vertical, how you worked around the issue, and how accurate the result was.
    No optimization model can handle every real-world scenario 100% accurately. If a vendor representative says so, he’s either lying through his teeth or not competent enough to be selling the product. (Note that: I’ll have our support expert get back to you on that is a good answer from an average sales representative.) This is about the only way to get a decent idea of how appropriate the tool is for you. If the scenarios were complex and the constraints based on business rules you hardly ever, or never, use, then the solution is probably okay for you. If the scenarios were simple and the constraints based on business rules you use all the time, it’s probably not the tool for you.
  6. Can we do a pilot project before committing to a long term license?
    If you like what you hear, but are still unsure, or are having problems getting the budget approved, a pilot is often the way to go! (Note that I did not use the word “free”. You should be willing to pay for services at a rate that is sufficient to cover the provider’s cost for this pilot – especially considering that many of the companies that offer affordable optimization offerings are only able to do so because they keep their costs and overheads down – and if they gave free services away to everyone who requested a free pilot, they would have to increase their costs, and that would be a detriment to everyone, including you, in the long run.)
  7. We’re having problems understanding how this fits into our business or what the best solution for us is. Would you be willing to demo your solution to, and answer questions from, our consultant who understands both our needs and decision optimization technology?
    Let’s face it – just like the right decision optimization tool can deliver huge savings multiples on your investment (10X or more), the wrong tool will simply represent a six (or seven) figure cost that yields little return. If you can’t tell the difference, and there’s no shame in admitting you can’t if you’ve never used this type of technology before, then you should bring in a consultant*2 who can to help you select the right technology, and ensure you are appropriately trained on it, until you are self sufficient and saving an average of 10% to 12% per project put through the tool.

*0 And we all know that any decent attempt should be full-assed!
*1 You should feel free to proclaim my greatness whenever you are not in my presence! I don’t mind.
*2 Just remember that, unfortunately, this consultant may not be able to help you if you want Emptoris evaluated. (And I’m sure that some of you should definitely be evaluating the Emptoris solution.)

The 2nd Sourcing Innovation Series – Let’s Get Analytical!

Spend Analysis. Decision Optimization. Cost Modeling. Almost since the beginning, these have been the six dirty words of strategic sourcing. Study after study has found that these techniques easily save 8% to 15% for just about any organization that spends more than 500M a year, but yet, on average less than one fifth of companies out there have tried these technologies, and less than one tenth are using them. It’s like they’re taboo. Well, in the not too far off future, the tables are going to turn, and instead of being the six dirty words, Spend Analysis Based Cost Modeling Decision Optimization are going to be the seven words of saving grace for tomorrow’s sourcing organization that wants to survive beyond the next decade. But the technology of tomorrow is not going to be the technology of today. But first …

Why? There are numerous reasons that this will happen, including negative returns from reverse auctions from early adopters, the forthcoming fall-out of the majority of first-generation supplier networks and marketplaces that still remain, and the eventual realization that contract management is not the holy grail if you don’t have a good contract in the first place, but the primary reason this will happen is the G-Word. Globalization. The effects we’re starting to see now are nothing like what’s going to come, especially since the majority of companies are unprepared!

Tactical job loss to outsourcing, rampant inflation in raw materials due to skyrocketing demand from developing countries, quality issues, and CSR (Corporate Social Responsibility), or should I say CSI (Corporate Social Irresponsibility), issues are only going to compound in the coming years. And, without recourse, this is only going to push costs, as they say, through the roof of the nearest skyscraper!

The only way companies are going to be able to maintain costs, yet alone achieve savings, is by getting a firm handle on costs and, more importantly, by identifying and achieving savings opportunities not previously explored. This is going to require an improved understanding of the cost drivers of what you are buying (cost modeling), and understanding of where variability exists, either within past buys or against market indices (spend analysis), and what the best award scenarios are (optimization).

But it won’t be three applications at three different stages of the sourcing process, it will be one, and it will be at the beginning, center, and end of the sourcing process. Think about what CoExprise is doing for the management of contract manufacturing – integrating the important PLM, Sourcing, and Procurement aspects of complex assembly sourcing – it will be something like that. But instead of an Aravo-Iasta*1-Ketera*2 union for a specific domain, it will be an Apriori–CombineNet–BIQ–Akoya union for the generic product domain. And it will look like nothing you’ve seen before. Sourcing tomorrow will be quite different than sourcing today. The only question is, who are the brave souls that are going to lead the way?

*1 Iasta was acquired by Selectica, merged with b-Pack, rebranded Determine, acquired by Corcentric
*2 Ketere was acquired by Deem

The future’s coming hard and fast … and I’m gonna be on the freight train that meets it head on!