Just a To-The-Point Reminder of Why Shipping Costs So Much

Empty pallets, empty containers, empty loads.

It’s essentially the same reason (airport) taxis usually cost more than Ubers. Empty space one way (in the form of seats).

Think about it. If a truck is coming empty from a big city 300 miles away to your plant to carry product back, every week, that’s over 15,000 empty miles a year on that one truck. A truck which takes a driver (who needs to be paid by the hour), gas (which costs by the gallon), maintenance, and replacement parts on an accelerated schedule. That means that you’re paying twice what you would be paying if the truck wasn’t empty for the majority of that 300 miles.

And even if you sub your shipping out to a logistics company, if the logistics company isn’t working hard enough on your behalf, that’s why certain lanes could be too high. And that’s also why you shouldn’t let a supplier or a single carrier manage your shipping. All carriers are going to have long empty lanes. You need to make sure that you’re cargo is not on these. You need to be sure that the truck isn’t driving more than a few hours from it’s drop off to your pickup (or your drop-off to it’s next pickup) so that the carrier is able to give you the lowest cost possible on the lane.

That’s why the best companies do global lane analysis (using decision optimization) and award contracts to multiple carriers that minimize costs across all lanes (by directly or indirectly eliminating the empty lanes).

So if you want to lower costs in your supply chain, just like you would avoid the empty calories in your diet, avoid the empty loads on your lanes.

Why A True Supply Management Professional Still Will Not Be Replaced by Technology

Algorithms still don’t sense, still can’t read the majority of non-verbal cues (as even the best mood detection algorithms can barely differentiate between “happy”, “indifferent”, and “sad” … even when the people it is analyzing have big smiles, flat lips, and big frowns), take calculated risks that go outside the programmed parameters, or form common bonds. They don’t feel, and they are not intelligent. And while their predictive capabilities are now getting scary in some respects, they are not infallible, and as we discussed in our last post, when they fail, they fail in a big, big way.

As first noted in our original post five years on Why a True Supply Management Professional will Never be Replaced by Technology, not only do algorithms not feel, but they are als incapable of accurately predicting how a person will respond to a suggestion that has any emotional impact whatsoever. Especially in today’s individualistic society where the message is what is interpreted by the recipient and only someone with a shared understanding will be able to comprehend what that is and react accordingly. As a result, an algorithm cannot negotiate (unless it is negotiating with another algorithm — but that’s not the best of ideas. When two algorithms negotiate, they develop their own undecipherable shorthand [as evidenced in multiple studies and real world occurrences, which includes two creepy Facebook bots talking to each other in a secret language], and we won’t be able to figure out what they did or why. (Was it to optimize the best win-win situation or was it to advance the plans for building SkyNet. We don’t know.)]

Secondly, as pointed out in our previous posts, successful negotiation depends on more than a first party transmitting a message to a second party that the second party can accept, but understanding all of the possible messages which might be accepted, their likelihood, and which are the most preferable to each organization and selecting the best one for the situation at hand. And while an algorithm can compute which options are likely given certain assumptions, and which of these options are the least distance from optimal according to some metric, it cannot determine what assumptions to make. Only a person who can feel, and feel what the other party is feeling, can be the judge of what good assumptions are. And, secondly, algorithms cannot sense. They don’t feel, and they don’t have instinct —- because that requires real intelligence!

Thirdly, as described above, they can’t accurately read non-verbal cues. Even if someone is stating that they may be agreeable to an offer, the reality might be that they may have no intention of ever accepting the offer, and are only indicating the contrary either because it’s the culturally polite thing to do or they want to stall for more time while they figure out their position. It’s often the case that such a person is not as good at masking their demeanor as they are at masking their words. It might be the case that their non-verbal cues give more away than they would like, but only a trained negotiator with years of experience and instinct could be an accurate judge of this.

But, even more importantly, they still typically can’t detect patterns in unrelated data, as it’s typically the case they can only process specified data in a specified set of ways. And a fixed data pool never tells the whole story. A fixed algorithm might not know that a fire today will impact resource availability in six months, that your main competitor is likely to go out of business due to a massive loss in a patent infringement lawsuit, or that a new technology is going to make the current technology obsolete in 18 months, with prices and demand starting to plummet in six months. As a result, in each of these instances, the algorithm would buy (today) (at a much) higher (price) than it needs to.

In short, the proper application of good, assistive intelligence, technology will make you two, ten, and maybe even one hundred times more productive (depending on the metric), but it cannot replace you. No matter what a vendor may claim. So don’t be scared of new technology for your supply chain —- embrace it. But don’t trust it blindly. Verify. Then you’ll have the best of both worlds — efficiency, with reliability — provided not by the system, but by your intelligent brain.

… And Keep Your Big Platforms. Big Brains Will Still Win in the End!

Five years ago, about the time when the big data hype first reached insane hype levels, SI published a post that it was sick of all the big data hype and how it is not the answer to all our problems because, not only is this a load of baloney, the reality is that there’s no such thing as big data in business. As we said then, relative to our ability to process it, data has always been big. And, in business, big has always been meaningless. Furthermore, in business, we’ve always been able to process as much data as we need to in reasonable amounts of time if* we make good algorithm and technology decisions

Plus, the fact that all of the hype around big data is often centered around the fact that we will be able to replace science with math and processes with AI programs is even more ridiculous. There is no such thing as artificial intelligence. And even though we’ve finally taken automated reasoning to the point that we have assisted intelligence, there’s a big difference between recommendations from a leading expert system (which not only can’t know when it is wrong but how much it can be wrong the few times it is wrong) and an average, experienced, professional in the domain (who can know how likely they are right, and if they are not likely to be right, how far off they are likely to be).

But even worse than the big data hype is the big platform hype … how mega platforms backed with cognitive abilities can do it all! They can do a lot, but they can’t do it all. And any delusions we might have that they can are only going to get us into trouble. Because as soon as we start trusting them blindly, we’re going to turn two blind eyes and that’s when the 2% failure rate is going to kick in, and materialize in the absolute worst way possible.

In Procurement, it’s going to miss the fact that a new organizational vendor is a very high risk and make a 2 year sole-source award for a small, but critical, custom made component in your (engine/control system) assembly when, in fact, it should be excluding the vendor which just had its credit score downgraded from a B+ to a D-. It’s not going to predict that in all probability, the vendor is not going to be able to secure enough loans to stay afloat (until it fulfills your orders and other customer orders and grows its business after losing a major contract that accounted for one third of its production) and will shut down and stop delivering product in 3 months when no one’s watching. Your production line will go down for 2 weeks while you find a backup supplier to quickly bring a production line up, make a minimal order, and air-freight it to you. If it’s a big automotive production line that goes down for 2 weeks, that’s easily 10M down the drain, and that 1M you saved is wiped out ten times over in a second.

But that’s not the worst thing that can happen from us turning two blind eyes. In this situation, the company temporarily loses some money, and as long as it has enough inventory to keep most of its customers happy, and it can keep its failure quiet, no one notices. Now, if its a medical diagnostics vendor and a (visually-based) diagnostic expert system (designed to help with the identification of all skin conditions) used by a remote doctor fails and mistakes melonoma for a relatively benign lesion, it’s a whole different story. When you consider that skin cancer is one of the five fastest spreading cancers, by the time the patient goes back in and insists something wrong, it could be too late — the spreading could be too far and the patient will be doomed (since melonoma, while only 1% of skin cancer, is not only one of the fastest spreading skins cancers but also has the highest fatality rate and causes the majority of skin cancer deaths each year).

Big Platforms give Big Confidence, but it’s false confidence. I’ll take a real human expert any day. Yes, she’ll make a mistake sometimes. But she also knows when she’s not sure and you should get a second opinion. That’s not always something a system can tell you. It’s above a threshold, below a threshold, or on the line (and no decision is made or classification is given). But that’s not always the right way to look at a situation.

* And, FYI, hiring college drop-outs whose college experience consists of cutting and pasting HTML and javascript code and fiddling with it until it works is not a good technology decision. There’s a big difference between being able to code a web-page and develop a highly scalable, reliable, and efficient enterprise computing system. BIG DIFFERENCE!

Tomorrow is March for Science Day. That IS Important For Everyone.

Why? Besides the obvious that all modern technology is the result of science, you won’t get your next-generation cognitive sourcing platform without more advancement in, guess what, data science.

And, right now, in the US, every year, in addition to having to deal with the introduction, and passing, of more anti-science policies, you also have to deal with the fact that funding for science (education) is diminishing as well and it’s the cornerstone of all progress. What do you think inspires the continual advancement of advanced mathematics and statistics? Scientific need. And where do you think the roots of most of your analytical algorithms come from? Science.

So even though you spend your days slicing data and running reports in the back-office to meet the business goals of savings, reduced inventory turn-around time, reduced, risk, etc. — you’re still using the results of scientific research and progress. Don’t forget that. Or someday Kyle may not be able to flick the internet back on again when it starts to fail. (The sad reality is that because of a lack of science education and knowledge, some people actually believe this is how you fix the internet.)

For more information, see March for Science

Are You Ready to Get Analytical But Don’t Know How? Read On!

Now that you’ve read our last three posts and understand that you need to get more analytical if you want to get cognitive, hopefully you’re ready to dive deeper but just don’t know how to do that.

The four part answer is almost as easy as it was for optimization, just a bit more nuanced. What’s the nuance? Figuring out if your provider offers a modern spend analytics platform or is still a generation (or two) behind (when you are still behind yourself) is the nuance. So how do you determine if a vendor at least passes the sniff test? We’ll get to that, but first, let’s talk about where you start.

At a high-level, the four-part answer is almost the same as optimization. Just the vendor names change.

1) If you are using a sourcing or analytics platform from a modern provider with modern (next generation) analytics capability, use it (and acquire the module if necessary).

Who are the vendors? While we can’t say this list is thoroughly exhaustive, if you look at Spend Matters Deep Solution map, you see that the following providers make the map: AnyData, (SAP) Ariba, (Opera) BIQ, GEP, iValua, Jaggaer, Sievo, Simfoni, SpendHQ, Synertrade, and Zycus. Not all are equal, and this list is likely not exhaustive, but depending on your organizational needs, a sub-set of these providers is likely your starting point. (What Sub-Set? Depending on whether you are data, function, process, technology, configurability, or services oriented, the sub-set will vary. And practitioners who want to know which vendors match which subset can contact Spend Matters.) And if you are a do-it-yourself type, you could probably start with a platform like Spendata.

2) If you are not using a modern analytics platform or a modern sourcing platform with analytics, get a modern analytics platform or a modern sourcing platform with analytics, your choice.

Again, you can start with the dozen of providers above, which you can quickly narrow down depending on whether you prefer best of breed or sourcing suite and whether you favour technical orientations or service orientations. If the list is still too large, find the subset that bests fits your organizational size, industry, category focus, geography, and culture and focus in on those.

3) If you are using another sourcing or analytics (reporting) platform that is not meeting your needs, and can replace it, do so.

As with the optimization providers, a few of these providers have a considerable portion of their customer base that consist of customers that switched from another provider with a solution that didn’t meet their needs and, thus, have a lot of experience with change management, fear squashing, migrating your data over, and getting you up and running on the right processes quickly. Simply craft the right RFI and you will quickly zero in to the handful of providers that will likely be the best fit for your situation.

4) If you are using another sourcing platform or reporting platform that is otherwise meeting your needs, or can’t be replaced at the present time, or both, augment it with a pure-play deep-dive best of breed modern analytics solution.

So if you are in the situation that you just bought a best of breed Source-to-Contract or Source-to-Pay solution and can’t replace it, or you have a first generation BI tool that produces reports the executives love but doesn’t meet your needs, augment it with a point-based best of breed solution. From the above list,
AnyData, (Opera) BIQ, Sievo, Simfoni, SpendHQ, and Spendata fit that bill.

But what about the “sniff test”?

How do you differentiate a last generation solution from a current generation solution? Three tests. Have them, in front of you, in a live demo:

  • Build a Cube with Derived Dimensions and a new Report on the Cube on the Spot
    if they can’t do so (in 15 minutes), they are a last generation platform that can only work on pre-defined and pre-built OLAP cubes
  • Run a categorization exercise on at least 3 months of your transaction history / invoice data and at least 100,000 transactions
    if they can’t either use their AI, or powerful (collaborative) filtering and priority based rule definition, and get to the 95% mark in an hour, it’s not for you … (and, trust me, you don’t need AI to get to the 95% mark if the rule definition capability is appropriately defined)
  • Map the cube to a new taxonomy, create new derived dimensions, and create a set of filters that will allow comparison reports to be run between the cubes
    let’s face it, there is no one size fits all taxonomy for analysis, and this is the kicker test to see if the platform can support any taxonomy that is needed, run any analysis you want, and allow you to run comparison reports both as checksums and as differentials to figure out where the opportunities are hidden

All this should take less than a morning or afternoon. But it means the provider deserves to be on your short list.