Category Archives: Decision Optimization

Questions to Ask Your Optimization Vendor

This is an update of a post that originally ran way back in 2007. Yes, two, double-o seven. Seventeen years ago. It is being updated because

  1. it needs a re-posting
    (as very few of you will find it that deep in the archives)
  2. most of the vendors originally mentioned are gone

However, if you read, and remember, the original, you’ll realize that, like my article where the doctor goes mental on optimization myths (which was recently shared on LinkedIn), it doesn’t need much updating and what was written seventeen years ago is still valid to this day. (When you write to inform vs. to create meaningless buzz, it really does stand the test of time.) Let’s begin.

Not all optimization vendors are equal … and, more importantly, not all vendors that claim to have strategic sourcing decision optimization (SSDO) actually have it (since the underlying algorithms and model needs to meet a stringent set of requirements to be true SSDO), with some systems, to this day, barely qualifying as decision support. Thus, since the need for optimization is as desperate as it has ever been with costs again skyrocketing, risks rising rapidly, carbon control being critical, and supply assurance necessary for sustained operations, it’s time to make sure you know how to qualify a potential provider. This means you need to not only understand the basics of what SSDO does (see the archives), but also how to distinguish between the relative strengths and weaknesses of the different offerings, as well as how much strength you really need.

You need to buy optimization at the strength, and usability level, that you need — especially if the vendor is pricing it according to its power, or computational requirement. And while there is no such thing as too much, the reality is that a 95% solution is often more than enough as the entire point is understanding the optimal solution against each dimension (cost, risk, carbon), the cost of compromise between the trade-offs, and the cost of going with a preferred, versus calculated, vendor award. And doing this for EVERY sourcing event. Once you factor in enough discounts and constraints, it’s almost impossible to calculate the best award in a spreadsheet, and the insight of what you could be spending, versus what you are, how low your risks could be, versus what they are, and how much you could alter your carbon footprint, vs what your footprint is today, is invaluable. Even if you never select a recommended solution, the key is understanding how good your (preferred) award actually is.

Before we get to the (starting) question list, it should be pointed 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 the question list below is a good starting point.

1. Does the product meet the four criteria for strategic sourcing decision optimization?

  • Sound & Complete Mathematical Foundations : such as MILP solutions based on simplex, branch and bound, and interior point algorithms as many simulation, heuristic, and “AI” algorithms DO NOT guarantee analysis of every possible solution (sub)space given enough time, and, thus, are not “complete” in mathematical terms (and if they incorporate Gen-AI, they aren’t even “sound” in that they may not even compute an award that satisfies the constraints!)
  • True Cost Modelling :
    that supports tiered bids, discounts, and fixed cost components — the model must be capable of supporting all of the bid types being collected, as well as the cost breakdowns
  • Sophisticated Constraint Analysis : at a minimum, the model must be able to reasonably support 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 supply 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: 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., and must support qualitative factors and minimum and average scores across the award
  • 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 per event?

Furthermore, 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 of 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 lunch (or if its a huge model, over night) after you’ve quickly analyzed half-a-dozen scenarios and settled on your preferred scenario. Thus, tweaking ability is very important.

4. 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 IBM ILog’s CPlex), it’s now possible to rapidly re-build and re-solve even very large models using off-the-shelf modeling languages in seconds, allowing for e-auction tools that keep the model relatively moderate in comparison, and presolve with seed bids (current prices, market prices, last quotes), to incorporate decision optimization in real-time by simply updating a few parameters and re-solving the model every (few) parameter(s) update (depending on model-size) on a high-powered multi- core server with an appropriately configured and optimized solver (which can spin off copies and have each processor work on a different subspace). However, if the approach the product takes is to rebuild and resolve the model on every update, that’s not real-time, that’s near real time, and the slowdown could be significant for large models. (To clarify further, 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. Can you describe two or three scenarios you have encountered where you could not model the situation exactly?

And, more importantly, how did you work around the issue, and how accurate was the final result. The real world is messy, compared to models that are clean, only so much data is available, and math can only model as much as the minds who created the model could conceive. As a result, 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 optimization 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. 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 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% or more per project put through the tool.

7. Can we do a pilot project at-cost (or gain-share) 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”!) If you’re not willing to sign a license, given the sophistication of this technology and the amount of effort the provider is going to have to allocate to support you through the pilot and ensure you are successful, you need to be willing to pay for services at a rate that is sufficient to cover the provider’s cost for the 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.

The Sourcing Innovation Source-to-Pay+ Mega Map!

Now slightly less useless than every other logo map that clogs your feeds!

1. Every vendor verified to still be operating as of 4 days ago!
Compare that to the maps that often have vendors / solutions that haven’t been in business / operating as a standalone entity in months on the day of release! (Or “best-of” lists that sometimes have vendors that haven’t existed in 4 years! the doctor has seen both — this year!)

2. Every vendor logo is clickable!
the doctor doesn’t know about you, but he finds it incredibly useless when all you get is a strange symbol with no explanation or a font so small that you would need an electron microscope to read it. So, to fix that, every logo is clickable so you can go to the site and at least figure out who the vendor is.

3. Every vendor is mapped to the closest standard category/categories!
Furthermore, every category has the standard definitions used by Sourcing Innovation and Spend Matters!
the doctor can’t make sense of random categories like “specialists” or “collaborative” or “innovative“, despises when maps follow this new age analyst/consultancy award trend and give you labels you just can’t use, and gets red in the face when two very distinct categories (like e-Sourcing and Marketplaces or Expenses and AP are merged into one). Now, the doctor will also readily admit that this means that not all vendors in a category are necessarily comparable on an apples-to-apples basis, but that was never the case anyway as most solutions in a category break down into subcategories and, for example, in Supplier Management (SXM) alone, you have a CORNED QUIP mash of solutions that could be focused on just a small subset of the (at least) ten different (primary) capabilities. (See the link on the sidebar that takes you to a post that indexes 90+ Supplier Management vendors across 10 key capabilities.)

Secure Download the PDF!  (or, use HTTP) [HTML]
(5.3M; Note that the Free Adobe Reader might choke on it; Preview on Mac or a Pro PDF application on Windows will work just fine)

Coupa: Comprehensive Optimization Underlies Procurement Assurance: Coupa Supply Chain Solutions

We’ve never covered Coupa Supply Chain Solutions (for Design and Planning), formerly known as Llamasoft, here on Sourcing Innovation, but the doctor did contribute to some of the coverage over on Spend Matters, including the acquisition coverage (Functional Overview, Overlap Between Direct Procurement and Supply Chain, and Procurement, Finance, and Supply Chain Use Cases [Content Hub Subscription Required]) in 2020. Llamasoft / Coupa Supply Chain Design and Planning has also been more recently covered by Spend Matters’ Pierre Mitchell as part of his analysis of Coupa for Supply Chain Management overall. For those interested with a ContentHub subscription, see his pieces on Can Coupa manage supply as well as spend?, Coupa’s journey from Business Spend Management to Supply Chain Management: Assessing progress on seven key dimensions, and From Spend to Supply — Coupa’s direct spend management progress.

Coupa Supply Chain Solutions consists of four main offerings:

  • Supply Chain Modeller: the core solution, that can be used offline on the desktop (Supply Chain Guru) as well as online in the cloud, where you build network, inventory, and transportation models for optimization and exploration through the dynamic reporting and dashboard creation module; note that the online version can process multiple “what-if” optimization models simultaneously
  • Supply Chain App Studio: the online solution which allows users to build custom interfaces to the underlying model that can be, if desired, custom designed for different user types (procurement, logistics, demand planners, etc.) and then shared with those users who can use the app for regular analysis and what-if optimization
  • Demand Modeller: for demand modelling and forecasting — not covered in this article
  • Supply Chain Prescriptions: uses machine learning and AI to identify savings opportunities from changes to transportation and inventory models, as well as to identify risk mitigation scenarios, based upon the current supply chain design

In this article we are going to primarily cover the capabilities of the Modeller / App Studio and the Prescriptions which is the core of their supply chain (design and planning) solution suite.

The Modeller has three primary components:

  • Model : where you build the models
  • Explore : where you build what-if scenarios, that are then optimized
  • Results : the outputs of the what-if scenarios

Model building is quite easy. It’s simply a matter of selecting, or uploading, a set of data tables for each relevant supply chain entity. They can be pulled in from a relational database or from a CSV file in standard row-based column format. As long as the column headers have standard field names, the SCP solution can auto detect what entity the table represents (warehouse, lane, transportation mode, etc.) and what data is provided on the entity. It understands all the common elements of supply chain modelling, common names and representations, and appropriate business rules that can do all of the auto mappings.

When you pull in a table, and it does the mappings to the standard internal models, it also automatically analyzes and validates the data. It makes sure all entries are unique, key values required for the types of analysis supported are there (such as coordinates for warehouses, costs per distance for transportation modes, stock levels and associated product requirements for inventory), etc. and flags any conflicting, missing, or likely erroneous data for user review and correction.

When you go to build a scenario, it understands what is required in the base underlying model and validates that all of the necessary data is present. If data is missing, it warns you and gives you a chance to provide the missing data. (Furthermore, as you add constraints to the scenario, the platform understands the data is required and ensures that data is present as well before it tries to run the scenario.)

The application was designed for ease of use and speed, tailored for automating most of the model building process for standard network/inventory/transportation scenarios (including setting parameters and defaults) so that standard models can be built for analysis quick and easy (and it is also quick and easy to change or override any default as needed).

Explore provides the capability where you build scenarios for what-if? exploration.

Building scenarios is simple. You simply select the scenario requirements, or constraints, from a set of existing, or newly created, scenario items that define the parameters of the scenario. For example, for a network optimization, you might want to explore limiting the number of existing distribution hubs or adding more proposed nodes to see if you can reduce cost, carbon, or distribution time. For transportation, you might want to explore adding in rail to a network that is currently all truck to see if you can decrease cost. For inventory, you might want to reduce the number of locations where safety stock for rarely used components is stored (so you can limit the number of locations with a low turn rate and minimize the warehouse size/footprint you need at those locations) and see what happens and so on. Each scenario is built from a set of specifications that specify the restrictions that you want to enforce, which could even be a reduction in the current number of restrictions. These restrictions can be on any entity, or relationship. One can also create scenarios to explore how the network will change under different circumstances, such as demand change, cost change, or disruption. Selecting is a simple point-and-click or drag and drop exercise.

Once you’ve created the scenario(s) of interest (remembering that you can optimize multiple simultaneously in the online version), you launch them by selecting the type of optimization (the “technology”), the sub-type of optimization (the “problem type”), the horizon (the timeframe you want to analyze), and, optionally, override default parameters (if you don’t want to do a cost optimization but instead want to optimize carbon, service level, fulfillment time, risk, etc.). Then you run the scenario, and once the optimization engine works its math, you can explore the results.

The Model supports:

  • Network Optimization
    • Standard Network Optimization
    • … with Network Decomposition
    • … with Infeasibility Diagnosis
    • Greenfield Analysis
    • Cost-to-Serve Analysis
  • Inventory Optimization
    • Safety Stock
    • Safety Stock & Service Level
    • Safety Stock & Rolling Horizon
    • Safety Stock Infeasibility
    • Service Level
    • Rolling Horizon
    • Rolling Horizon Validation
    • Demand
  • Transportation Optimization
    • Transportation – Standard
    • Transportation – Interleaved
    • Transportation – Hub
    • Transportation – Periodic
    • Transportation – Backhaul
    • Transportation – Backhaul Matching
    • Driver Scheduling

In short, it’s a very extensive network, inventory, and transportation optimization modelling solution out of the box that makes it really easy for supply chain and procurement analysts to build scenarios and solve them against all of the traditional models (and variants) they would want to run. (And if your particular variant isn’t out of the box, the SCP team can code and add the variant into your deployment as the underlying solution was built to allow for as many models as was needed as well as unlimited scenarios on those models.) Note that, by default, the platform will always run the baseline scenario so you have a basis for comparison.

Results, which are output in the form of results tables, can then be analyzed in table form (by selecting the output table), graph form (by accessing the graphs), map mode (by accessing the map), or as a built-in or custom report/dashboards that the analyst can create as needed. For every type of analysis in the system, SCP includes a default set of dashboards for exploring the data set, which adapt to not only the type and subtype of optimization that was run, but the goal (objective function) as well. So if you did a cost optimization scenario, they summarize the costs. If you did a carbon optimization scenario, they summarize the carbon. If you did a service level optimization, they summarize the service level. If you did a carbon optimization relative to a maximum cost increase, they summarize the carbon and cost (and the relationship). In their platform, if you optimize one element or KPI, you can see the impact on all of the other costs and KPIs as all of the associated data is also output for analysis.

There is an output table for all elements which can be analyzed in detail, but most users prefer graph or map view on the relevant data.

Views provide custom, tabular, reports on the relevant fields of one or more tables, which can be exported to csv or pushed to another application for planning purposes. For example, if the model was a network optimization model, you can create a view that outputs the new distribution centres and fulfilment lanes for the revised network and push that to the TMS (Transportation Management System). If it was a transportation optimization model, you can output a table that specifies the carrier and rate for each lane, or, if necessary, each lane product combination and push that into the TMS. If it was a safety stock optimization model, you can output the product, location, minimum stock levels, and reorder points and push that into the Inventory Management or ERP system. And so on. There are default views for cost, carbon, service level, demand, and inventory optimizations, along with drill ins for relevant types of cost (site, production, by transportation type, etc.), but it is quite easy for a user to create a view on any table, or set of tables, with derived fields, with the view editor.

Graphs summarize the data in tables or views graphically, allowing for easy visual comparison. Select the scenario, select the data, select the graph type, and there’s your graph. They are most useful as components in dashboard summaries.

Maps provide a visual representation of the supply chain network — warehouses/distribution centers, customer locations, transportation lanes — overlaid on a real-world map with the ability to filter into particular supply chain network elements. There is a default map for the full network overview, and it can be copied and edited to just display certain elements.

Dashboards group relevant elements, such as a map of the current distribution network, a map of the optimized distribution network, a graphic summary of current distribution costs, a graphic summary of new distribution costs, and tabular (view-based) cost, carbon, and service level comparisons as the result of a supply chain network optimization scenario. These are typically custom built by the analyst to what is relevant to them.

Prescriptions, only in the online version of Supply Chain Modeller, are based on the 22 years of experience the SCP team has in building and analyzing models and uses advanced ML, simulation, and AI to automatically identify potential cost savings, and risk reductions and presents rank-ordered opportunities for you in each category, which you can drill into and explore. This solution automatically generates dozens (upon dozens) of scenarios and performs hundreds (or thousands) of analyses to automatically bring you actionable insights that you can implement TODAY to improve your network.

These savings will be grouped by type for easy exploration. For example, when it comes to cost savings, these will often be obtained by node skipping, mode switching, or volume consolidation — and the prescriptions module will summarize the prescriptions in each category, as well as summarizing the relative total savings of each category. A user can accept or reject each (sub) set of prescriptions, and then export all of the accepted prescriptions into new route definition records that can be imported into the TMS.

Note that the analysis that underlies the prescription analysis is very detailed, and in addition to the prescriptions, the platform will also identify the top network factors that are impacting the transportation costs, such as fleet distance, unique modes, certain carriers, country, etc.

When a user drills in, she sees the complete details of the prescription, including the before and after. In the node skipping example, they will see the current distance, products, quantities, (total) weight and volume, and current rates and then will see these in comparison to the new distance, new rates, and new costs. The old and new routes will be mapped side by side. The old and new lanes will be detailed.

The out of the box network risk summary for revenue at risk is quite impressive. The platform is able to compute the overall network revenue AND network profit at risk based on single sourced site-products, % of flow quantity single sourced, avg. end-to-end service times, and impacted paths. It will then do analyses to identify potential risk mitigation improvements allowing for 5%, 10%, and 15% network change (based on how product flows through the network with the current design) and compute the corresponding change in revenue and profit at risk as a result of those changes as well as the change in network cost. Usually the cost will increase slightly, but not always. For example, it could be that you could reduce the revenue at risk by 5% just through a supplier reallocation and network redesign, and if you were really risk averse, it could be that a 1% increase in network cost could result in an 8% to 10% decrease in revenue, and profit, at risk. And that could be the cheapest supply chain insurance you can buy.

Of course, you can drill into each model, the prescriptions, and the risk reduction with each individual change. It’s an extremely powerful tool.

Another thing that is really powerful in Coupa Supply Chain Solutions is the specific applications they can enable in the online App Studio, including the Cost-to-Serve App (which is just one example of the custom interfaces that can be built) that is one of the most complete dynamic dashboards for network insights that the doctor has ever seen. A summary can’t do it justice, but to whet your appetite to be sure you ask to see it in a demo, it has a full set of meaningful baseline KPIs, a visual network and flow summary, deep details on product costs and profitability, deep details on lanes and transportation costs, and so on. You can also quick-select a scenario to run and compare against the baseline in the app. It’s extremely well thought out.

Furthermore, you can build scripts in the App Studio to rebuild and run models on a schedule when you have a network in flux (because of disruptions, supply base changes, network changes as a result of prescriptions, etc.). And, of course, you can share these models and apps and dashboards with other analysts and democratize supply chain planning, easily enabling planners to analyze their own scenarios and make decisions collaboratively in a user-friendly App.

In short, Coupa has fulfilled the supply chain use cases we identified back in 2020 in Procurement, Finance, and Supply Chain Use Cases. It’s a great solution that you should check out, especially if you would like to have procurement and supply chain under one umbrella.

Even Forbes is Falling for the the Gen-AI Garbage!

This recent article in Forbes on the Supply Chain Shift to Intelligent Technology is what inspired last week’s and this week’s rant because, while supply chains should be shifting to intelligent technology, the situations in which that is Gen-AI are still extremely rare (to the point that a blue moon is much more common). But what really got the doctor‘s goat is the ridiculous claims as to what Gen-AI can do. Claims with are simultaneously maddening and saddening because, if they just left out Gen-AI, then everything they claimed is not only doable, but doable with fantastic results.

Of the first three claims, Gen-AI can only be used to solve one — and only partially.

Procurement and Regulatory Compliance
This is one example where a Closed Private Gen-AI LLM is half the battle — it can process, summarize, and highlight key areas of hundred page texts faster and better than prior NLP tech. But it can’t tell you if your current contracts, processes, efforts, or plans will meet the requirements. Not even close. In fact, no AI can — the best AI can just indicate the presence or absence of data, processes, or tech that are most likely to be relevant and then an intelligent human needs to make the decision, possibly only after obtaining appropriate expert Legal advice.
Manufacturing Efficiency
streamline production workflows? optimize processes? reduce errors? No, Hell No, and even the Joker wouldn’t make that joke! You want streamlining? You first have to do a deep process cycle time analysis, compare it to whatever benchmarks you can get, identify the inefficiencies, identify potential processes and tech for improvement, and implement them. Optimize processes? Detailed step by step analysis, identification of opportunities, expert process redesign, training, implementation, and monitoring. Reduce errors? No! People and tech do the processes, not Gen-AI — implement better monitoring, rules, and safeguards.
Virtual Supply Collaboration
A super-charged chatbot on steroids is NOT a virtual assistant. Now, properly sandwiched between classical AI and rules-based intelligence it can deal with 80% of routine inquiries, but not on its own, and it’s arguable if it’s even worth it when a well designed app can get the user to the info they need 10 times faster with just a couple of clicks. Supply chain communicating? People HATE getting a “robot” on a support line as much as you do, to the point some of us start screaming profanities at it if we don’t get a real operator within 10 seconds. Based on this, do you really think your supplier wants to talk to a dumb bot that has NO authority to make a decision (or, at least, should NEVER have the authority — though the doctor is sure someone’s going to be dumb enough to give the bot the authority … let’s just hope they can live with the inevitable consequences)?

And maybe if the article had stopped there the doctor would let it pass, but
first of all, it went on to state the following for “AI”, without clarifying that Gen-AI doesn’t fit in the process, leading us to conclude that, since the first part of the article is about Gen-AI, this part is too, and thus is totally wrong when it claims that:

“AI” understands dirty data
with about 70% accuracy where it counts IF you’re lucky; that’s about how accurate it is at identifying a supplier from your ERP/AP transaction records; an admin assistant will get about 98% accuracy by comparison
it can “confirm” inventories
all it can do is regurgitate what’s in the inventory system — that’s not confirmation!
it can identify duplicate materials
first it has to identify two records that are actually duplicates;
and how likely do you think this is with a supplier mapping accuracy of 70%?
it can identify materials to be shared among facilities
well, okay, it can identify materials that are used across facilities and could be located in a central location — but how useful is that? it’s not because, first of all, YOU ALREADY KNOW THIS, and, second, IT CAN’T DO SUPPLY CHAIN OPTIMIZATION — THAT’S WHAT A SUPPLY CHAIN OPTIMIZATION SOLUTION IS FOR! OPTIMIZATION!!! We’ll break it down syllabically for you so you know what to ask for. OP – TUH – MY – ZAY – SHUN!
it can recommend ideal storage locations
again, NO! This requires solving a very sophisticated optimization model it doesn’t have the data for, doesn’t know how to build, and definitely doesn’t know how to solve.
it can revamp outdated stocking policies
well, only the solution of a proper Inventory OPTIMIZATION Model that identifies the appropriate locations and safety stock levels can identify how these should be revamped
it can recommend order patterns by consumption and lead time
that’s classical curve fitting and tend projection

And, secondly, as the doctor just explained, most of what they were saying AI could do CAN’T be done with AI, and instead can only be done with analytics, optimization, and advanced mathematical models! (You know, the advanced tech (that works) that you’ve been ignoring for over two decades!)

The Gen-AI garbage is getting out of control. It’s time to stop putting up with it and start pushing back against any provider who’s trying to sell you this miracle cure silicon snake oil and show them the door. There are real solutions that work, and have worked, for two decades that will revolutionize your supply chain. You don’t need false promises and tech that isn’t ready for prime time.

Somedays the doctor just wishes he was the Scarecrow. Only someone without a brain can deal with this constant level of Gen-AI bullsh!t and not be stressed about the deluge of misinformation being spread on a daily basis! But then again, without a brain, he might be fooled by the slick salespeople that Gen-AI could give him one, instead of remembering the wise words of the True Scarecrow.

You Don’t Need Gen-AI to Revolutionize Procurement and Supply Chain Management — Classic Analytics, Optimization, and Machine Learning that You Have Been Ignoring for Two Decades Will Do Just Fine!

Open Gen-AI technology may be about as reliable as a career politician managing your Nigerian bank account, but somehow it’s won the PR war (since there is longer any requirement to speak the truth or state actual facts in sales and marketing in most “first” world countries [where they believe Alternative Math is a real thing … and that’s why they can’t balance their budgets, FYI]) as every Big X is pushing Open Gen-AI as the greatest revolution in technology since the abacus. the doctor shouldn’t be surprised, given that most of the turkeys on their rafters can’t even do basic math* (but yet profess to deeply understand this technology) and thus believe the hype (and downplay the serious risks, which we summarized in this article, where we didn’t even mention the quality of the results when you unexpectedly get a result that doesn’t exhibit any of the six major issues).

The Power of Real Spend Analysis

If you have a real Spend Analysis tool, like Spendata (The Spend Analysis Power Tool), simple data exploration will find you a 10% or more savings opportunity in just a few days (well, maybe a few weeks, but that’s still just a matter of days). It’s one of only two technologies that has been demonstrated, when properly deployed and used, to identify returns of 10% or more, year after year after year, since the mid 2000s (when the technology wasn’t nearly as good as it is today), and it can be used by any Procurement or Finance Analyst that has a basic understanding of their data.

When you have a tool that will let you analyze data around any dimension of interest — supplier, category, product — restrict it to any subset of interest — timeframe, geographic location, off-contract spend — and roll-up, compare against, and drill down by variance — the opportunities you will find will be considerable. Even in the best sourced top spend categories, you’ll usually find 2% to 3%, in the mid-spend likely 5% or more, in the tail, likely 15% or more … and that’s before you identify unexpected opportunities by division (who aren’t adhering to the new contracts), geography (where a new local supplier can slash transportation costs), product line (where subtle shifts in pricing — and yes, real spend analysis can also handle sales and pricing data — lead to unexpected sales increases and greater savings when you bump your orders to the next discount level), and even in warranty costs (when you identify that a certain supplier location is continually delivering low quality goods compared to its peers).

And that’s just the Procurement spend … it can also handle the supply chain spend, logistics spend, warranty spend, utility and HR spend — and while you can’t control the HR spend, you can get a handle on your average cost by position by location and possibly restructure your hubs during expansion time to where resources are lower cost! Savings, savings, savings … you’ll find them ’round the clock … savings, savings, savings … analytics rocks!

The Power of Strategic Sourcing Decision Optimization

Decision optimization has been around in the Procurement space for almost 25 years, but it still has less than 10% penetration! This is utterly abysmal. It’s not only the only other technology that has been generating returns of 10% or more, in good times and bad, for any leading organization that consistently uses it, but the only technology that the doctor has seen that has consistently generated 20% to 30% savings opportunities on large multi-national complex categories that just can’t be solved with RFQ and a spreadsheet, no matter how hard you try. (But if you want to pay them, a Big X will still claim they can with the old college try if you pay their top analyst’s salary for a few months … and at 5K a day, there goes three times any savings they identify.)

Examples where the doctor has repeatedly seen stellar results include:

  • national service provider contract optimization across national, regional, and local providers where rates, expected utilization, and all-in costs for remote resources are considered; With just an RFX solution, the usual solution is to go to all the relevant Big X Bodyshops and get their rate cards by role by location by base rate (with expenses picked up by the org) and all-in rate; calc. the expected local overhead rate by location; then, for each Big X – role – location, determine if the Big X all-in rate or the Big X base rate plus their overhead is cheaper and select that as the final bid for analysis; then mark the lowest bid for each role-location and determine the three top providers; then distribute the award between the three “top” providers in the lowest cost fashion; and, in big companies using a lot of contract labour, leave millions on the table because 1) sometimes the cheapest 3 will actually be the providers with the middle of the road bids across the board and 2) for some areas/roles, regional, and definitely local, providers will often be cheaper — but since the complexity is beyond manageable, this isn’t done, even though the doctor has seen multiple real-world events generate 30% to 40% savings since optimization can handle hundreds of suppliers and tens of thousands of bids and find the perfect mix (even while limiting the number of global providers and the number of providers who can service a location)
  • global mailer / catalog production —
    paper won’t go away, and when you have to balance inks, papers, printing, distribution, and mailing — it’s not always local or one country in a region that minimizes costs, it’s a very complex sourcing AND logistics distribution that optimizes costs … and the real-world model gets dizzying fast unless you use optimization, which will find 10% or more savings beyond your current best efforts
  • build-to-order assembly — don’t just leave that to the contract manufacturer, when you can simultaneously analyze the entire BoM and supply chain, which can easily dwarf the above two models if you have 50 or more items, as savings will just appear when you do so

… but yet, because it’s “math”, it doesn’t get used, even though you don’t have to do the math — the platform does!

Curve Fitting Trend Analysis

Dozens (and dozens) of “AI” models have been developed over the past few years to provide you with “predictive” forecasts, insights, and analytics, but guess what? Not a SINGLE model has outdone classical curve-fitting trend analysis — and NOT a single model ever will. (This is because all these fancy-smancy black box solutions do is attempt to identify the record/transaction “fingerprint” that contains the most relevant data and then attempt to identify the “curve” or “line” to fit it too all at once, which means the upper bound is a classical model that uses the right data and fits to the right curve from the beginning, without wasting an entire plant’s worth of energy powering entire data centers as the algorithm repeatedly guesses random fingerprints and models until one seems to work well.)

And the reality is that these standard techniques (which have been refined since the 60s and 70s), which now run blindingly fast on large data sets thanks to today’s computing, can achieve 95% to 98% accuracy in some domains, with no misfires. A 95% accurate forecast on inventory, sales, etc. is pretty damn good and minimizes the buffer stock, and lead time, you need. Detailed, fine tuned, correlation analysis can accurately predict the impact of sales and industry events. And so on.

Going one step further, there exists a host of clustering techniques that can identify emergent trends in outlier behaviour as well as pockets of customers or demand. And so on. But chances are you aren’t using any of these techniques.

So given that most of you haven’t adopted any of this technology that has proven to be reliable, effective, and extremely valuable, why on earth would you want to adopt an unproven technology that hallucinates daily, might tell of your sensitive employees with hate speech, and even leak your data? It makes ZERO sense!

While we admit that someday semi-private LLMs will be an appropriate solution for certain areas of your business where large amount of textual analysis is required on a regular basis, even these are still iffy today and can’t always be trusted. And the doctor doesn’t care how slick that chatbot is because if you have to spend days learning how to expertly craft a prompt just to get a single result, you might as well just learn to code and use a classic open source Neural Net library — you’ll get better, more reliable, results faster.

Keep an eye on the tech if you like, but nothing stops you from using the tech that works. Let your peers be the test pilots. You really don’t want to be in the cockpit when it crashes.

* And if you don’t understand why a deep understand of university level mathematics, preferably at the graduate level, is important, then you shouldn’t be touching the turkey who touches the Gen-AI solution with a 10-foot pole!