Category Archives: rants

AI Will not Save Procurement … It Will Only Hasten its Demise

A recent post over on Spend Matters UK from Andrew Nichols in “artificial intelligence help businesses save thousands” boldly states that Artificial Intelligence Can Help Procurement Solve Some of the Big Challenges. In fact, he predicts that AI in the not so distant future will play a major role in the international supply chain, supporting businesses to solve a number of very contemporary problems.

In particular, Andrew believes that AI could identify new markets, manage supply chain risks, track exchange rate volatility, and find the best value without compromise on quality.

If this were true, Procurement would not be needed at all, and the C-Suite would be chanting “Procurement is Dead. Long Live the Machine! Our Samaritan has Arrived!” If an AI (which does NOT exist by the way, intelligence is not artificial) could do that, you’d all be fired, because, let’s face it, when it comes to managing supply chain risks, tracking exchange rates, identifying new markets, and always finding the best value, your batting average is less than that of a major league baseball pro.

An AI can give you market statistics, and break it down by region, demographic, competitor, and product. It can NOT tell you how appropriate a market is for you. You have no idea why a market is good for a competitor. You can cross-correlate it’s products to other top selling products on the market and identify common features, common advertising channels, and common comments across brand surveys, but the best you can draw is conclusion based on correlations. Correlation is not causation. You could take the highest ranked strategy suggestion from the analytics engine, implement it, and flop miserably because a key factor was missed, flawing the entire model.

An AI can compute, for every product in the world, the cost to value ranking using market costs, exchange rates, correlation to desired feature lists, and consumer ratings, but this is not the best value. The best value is that where the cost to value formula is based on your value rankings, which could be much more heavily dependent on reliability, safety, and service than look, feel, and flash. And since your organization will not have every product rated, the best the AI can do is suggest the most likely candidates for human review.

An AI can detect the presence of risk indicators that you have defined against known risks, it cannot identify risk indicators for unknown risks. If the algorithm doesn’t understand that a tsunami is a risk because it can damage harbours and destroy coastal plants, the risk will not be identified until it discovers a news story about how the supplier plant had to shut down. And if it does not understand that legal proceedings can bankrupt a small company, it could overlook a filing with the potential to bankrupt the supplier. If the supplier was strategic, that is something the organization would want to know about immediately.

An AI can track exchange rate and give you a real-time view into which is the most preferable rate, the short-term and likely long-term trends, and give you suggestions with an expected level of confidence within plus/minus x%, but can not necessarily predict the right currency to use to lock in a long term value for any better than a human expert. No known algorithm knows all the factors that contribute to exchange rates, how to detect their presence, and how to incorporate them. There are a lot of advanced statistical algorithms that can model the trend curves well, but they assume that markets will more or less keep the status quo, which never happens. Their projections are useful, as they can identify which currencies are likely to be best and where inflection points are likely to occur, making the best use of an experts time, but they cannot replace the expert.

And if any CPO were to try and replace a team with an AI for one or more of these functions, then he would quickly bring an end to Procurement in his organization because, while it would succeed in many cases, sooner or later there would be a spectacular failure that would cancel out all of the previously identified value, putting the entire organization at risk.

the doctor doesn’t like lists either, but the 50/50 is as good as it gets

There’s been a lot of noise surrounding the 50/50, and we know the Spend Matters client services team have received a number of enquiries from companies who felt they were entitled to make the list, or make the “to know” list, but didn’t, based on their customer status, and even Jason, the founder of the list, has gone on record as to why he hates the list in “3 reasons i hate the spend matters 5050”. Peter (in “spend matters 50 to know and 50 to watch questions and answers”) and Taras (in “why good things come in threes”) have chimed in too. Now the doctor is chiming in.

Let’s begin by reiterating the title. Generally speaking, the doctor despises lists. First of all, most lists come from the analyst firms that release the tragic quadrant procurement grave reports which, as we all know, change the requirements for consideration, and grading, every few years. (This often has the benefit of increasing the rank of some companies and decreasing the rank of others in an often arbitrary fashion. This is why the doctor is working with the maverick and the prophet to define standard requirements for different solution types so that vendors can be graded equally and fairly against a common, consistent, benchmark.)

Secondly, most are subjective lists that tend to represent the views of just one or two analysts, often heavily influenced by a small number of vendors that they spend the majority of their external interaction time with. While this doesn’t mean that they will be anymore biased to these vendors as opposed to others when doing their rankings, their view of what a product should, and should not, do are heavily influenced by these vendors and, thus, the rankings of these vendors are always good.

Third, the lists are usually limited to sourcing, or procurement, or SRM and not broad enough to identify related, emerging, complementary technology that can prove just as useful to an innovative firm. There always comes a point where the same-old, same-old fails to add value.

Fourth, any lists that stops at Vendor #X does not include Vendor #X+1, which may be just as valuable to a (slightly) smaller group of potential clients and, more importantly, may actually do more to warrant watching in the months that follow than Vendor #X, that might become complacent given their recent ranking.

Fifth, as pointed out by Jason, there are always going to be accusations by vendors not included, third parties with their own agendas, and even by vendors included (but not ranked where they feel they should be).

Sixth, and not least, no list is perfect. At any given time there are vendors the analysts are unaware of that might deserve a spot on the list, there are vendors on the list that might not be keeping up the innovation, and many of the services oriented vendors have to be subjectively ranked on limited customer interviews.

But that doesn’t mean that a properly constructed list cannot be useful. A well constructed list, that is objective as possible, can open one’s mind to options one might not have known of but should consider. A well constructed list can help vendors realize how well they are known and what they are known for and where they need to spend more effort on education and marketing. And it gives vendors something to strive for, so long as that list is created equally each time it is created.

While the list is not perfect, this is the first list the doctor is aware of that, while subjective, was created in a fashion that was as objective as a subjective list can be. It was debated over by seven analysts across three continents — the revolutionary, the civil crusader, the money, the public defender, the maverick, the doctor, and the prophet — who covered a variety of areas including, but not limited to, S2C, P2P, SRM, Analytics, Services, Risk, and Finance, and who had very strong opinions on who should and should not be considered. And client status played no part whatsoever. As the prophet said, about half were Spend Matters present or past clients, half were not. With respect to the recommendations for the list from the doctor, the split was about the same.

Basically, when all was said and done, to make the to know list, at least three of the analysts had to agree (and more than that to be guaranteed a spot), and to make the to watch list, at least two (and three to make the list with certainty), and they had to persuade the profit that their choice was better than another choice that had the support of two analysts. While each of us can point to a handful of vendors and say we would have liked to seen them on a list, the fact of the matter is that if only one analyst sees a vendor as worthy, that’s a singular subjective view point. When at least three people agree, especially when they cover different sub-sectors, in different parts of the globe, that’s a much stronger statement than just analyst X likes vendor Y (especially when, as pointed out in our recent post on 30K a Day and You Haven’t Even Seen the Solution, any company that did not demo at least one analyst was not considered). Is it perfect? No. But is it better than everything else? Yes. And it’s going to get better still. As more types of applications and services pop-up, the prophet is going to add more experts to the pool. the doctor suspects next year’s list will be argued over by nine analysts, which will make the results stronger still.

30K a Day for Advice, and You Haven’t Even Seen a Solution!?!

the doctor recently learned two very disturbing pieces of information about one of the top analyst firms in the space. And when the doctor says top, he is referring to the type of analyst firm that produces one of the tragic quadrant procurement grave reports that many, many clients pay top dollar for in the hopes of identifying the leaders to invite to their technology RFX. (And the reasons that many vendors pay top dollar for analyst relations in the hope of getting enough notice to not only get included, but featured well.)

What was this disturbing news?

1. The firms is now charging up to 30K a day for dedicated analyst time.

This is disturbing for two reasons. First of all, the doctor doesn’t know any analyst in the space that is worth $3,750 an hour. The top analysts are easily worth $1,000 an hour, but almost $4,000? Top management consultants, like top lawyers, charge a lot, often in the $1,000 an hour range, but when you do the math, that’s already equivalent to a salary of 2 Million a year. How many professionals are worth 2 Million a year, yet alone the 7.5M this analyst firm is implying its analysts are worth? Not many. Secondly, as far as the doctor is concerned, this firm doesn’t have the top analysts in the space. Do you really want to pay that much for advice from a tier 2 analyst?

2. The analysts no longer take demos.

That’s right! All they want now is customer references. And while no firm should make a recommendation without a customer reference (because demos can be misleading in the hands of a slick demo expert when done in front of a non-technical analyst), no firm should be making a reference based on a customer reference alone. Why? If the vendor knows that they are judged entirely on the references, the only references they are going to give are those to new clients where the blush is fresh on the rose and those client representatives who have been bought.

But the company has a no-bribe policy, and we have confirmed from the company that the individual did not violate it!

Irrelevant. There are non-violating ways to bribe a rep at a top-named customers, and the best marketers in the space know the tricks. First of all, find someone who likes to travel but who cannot afford to do to much of it and does not get many opportunities to travel on the company dime. Invite them as a speaker to all your corporate events at desirable locations as well as your presentations at big industry events where they can look good in front of their peers — and cover not only all their expenses, but all of their partner’s expenses as well.

Next, make sure they are on the product advisory team. If their employment agreement doesn’t explicitly forbid outside consulting to your company in off hours, or honorariums for exceptional performance, pay them for it. If not, make sure the product advisory team has to go on semi-annual “retreats”, all expenses paid.

Finally, do everything you can to make them look good. In return, the praise they lather on your behalf will be thicker than the bond between thieves. Much thicker.

In other words, regardless of the product quality or suitability of the solution, all the analyst is going to hear is that it’s the best thing since sliced bread and the vendor performs eight miracles a week.

And when you put these two facts together, it is very, very, distressing that any client who engages the firm for a detailed strategy and recommendation session in the effort to pinpoint the best technology for them is going to essentially spend 30K for a worthless recommendation. That’s a lot of money that could be used for better things, including 3 days of consulting with an analyst who’s actually seen the solution, analyzed its effectiveness and appropriateness, and knows what they are talking about.

And you don’t have to hire the doctor (for a fraction of the price) to get this insight — Spend Matters also has a no play, no promotion policy. Not only is an open demo with one of their (extended) analyst team a requirement for consideration on the 50/50, but while they will mention the existence of a provider with whom they haven’t seen a demo, they will not recommend a client engage with such a provider without a detailed investigation of product capability, and deep, live, demos. (And, moreover, if they have 3 or more other providers to recommend that they believe will suit a client’s need, the existence of a provider they haven’t seen is not likely to be mentioned at all unless you inquire about that provider specifically.)

So why would anyone pay 3X what they should for advice that is worthless? Probably because they don’t know better. So before you pay what could be a ridiculous amount for analyst time, please do your research and know exactly what type of advice you can and can not get. the doctor is sure this particular firm (that will not be mentioned, since this post will apply to any analyst firm that decides to adopt the same stance of ridiculous day rates and no-demo policies and this post is going to be archived as long as SI remains alive) still does a great job at identifying process weaknesses, opportunities for new and established technologies, and strategies for advancement — but the doctor, who barely trusted their version of the tragic quadrant procurement grave reports to begin with, will never, ever trust a technology recommendation from again — and if they hold to this “no demo” rule, neither should you.

Driverless Delivery? Tantalizing Theft Target!

With the emergence of drones and, now, self driving cars, a number of delivery companies are promoting these as low cost delivery options to companies that want to reduce delivery costs, especially for small businesses shipping low volumes (that fit in a large van or small truck) or retailers doing B2C delivery. But are they really low cost?

Yes, drivers cost money because, like all workers, they expect to be paid. And if you could obtain a driverless vehicle for the same price of a driver-required vehicle, you would save. But driverless vehicles come with a higher price tag. Now, the argument is that over the lifetime, the savings from a reduced driver workforce will cancel out the increased up from cost, and this would be true if the driverless option were as reliable as the driver-required options.

Now at this point, you’re probably asking what madness has the doctor contracted because, unlike humans that get sick, get lazy, make mistakes, and need rest — as long as the equipment gets the fuel and proper service, the software can drive it 24/7 — and this is true. The equipment can run 24/7, but this doesn’t mean you’ll get your stuff.

First of all, if there’s a programming error, or GPS error, there is no one there to detect and correct it. If GPS steers an Uber off course (and it does regularly in big cities with lots of tall buildings … sending multiple Ubers in a row a block away from where I was in Chicago recently despite the fact I was very sure to provide the address and not accept the default GPS location), the driver can say “there’s no one here”, call, and figure out where to go. If GPS steers a delivery drone off course, the customer’s neighbour gets a free gift and you get to eat the replacement cost as the credit card company is not going to rule in your favour in a dispute where the customer provided correct shipping information but you delivered to the wrong address. And the cost multiplies if an entire truck is shipped to the warehouse next door and you can’t prove it. (Even if you can, it does not mean you will get your goods or money back.)

But the biggest problem is that there is no guarantee that the goods will even make it to the destination. Goods being delivered driverlessly are very tantalizing theft targets. Not only is there no security to worry about, but there is no driver to even notice a theft as it is happening, report it, and get descriptions of the perpetrator — which means 0 chance of recovery. And do not think for a second that insurance is going to cover it in a cost effective manner. As claims start rising, and investigations into reasons continue, rates are going to either become unaffordable for driverless delivery options or become nonexistent options for the average business.

The argument that the drone is not interceptable until it drops low enough to deliver the package is not going to hold because signals can be hijacked and they can be hacked. (If top of the line cars can be hacked, how hard do you think it is to hack a bottom of the line drone?). And the argument that the delivery vehicle is secure until it reaches its destination is not going to hold either because if thieves can bust open a lock and rob a moving delivery truck with a driver unnoticed, how hard is it going to be to do the same to a driverless one. (Answer, even easier — no one to see the theft. Cameras do not count. They are easily hacked if they are digital and easily blinded by LED lights.)

Driverless trucks are already becoming theft ring targets, and delivery drones will soon be the target of bored hackers everywhere who will be able to get stuff en-route and not have to wonder if the order on the stolen credit card number will go through before the theft is detected and reported.

Driverless delivery is a tech-dream, but, for the time being, is not a Procurement one. You have been warned.

One Hundred and Seventy Years Ago Today …

Marked the first publication of the Cambridge Chronicle, the oldest surviving American newspaper. This is a very long time for a publication to survive. A very long time. Especially when many publications in today’s internet age only last a few years. Even the Red Herring ceased print publication in 2007, less than fourteen years after it was founded. (There was a time when it was as popular, if not more so, than Wired, an internet age publication that actually survived the internet age, but which still is only 23 years old.)

Even the New York Times did not start until five years later (and celebrates it’s 165th birthday on September 18 of this year). This blog, while the second oldest surviving independent blog in the Supply Management space (at 10 years), is just a blip when compared to the Cambridge Chronicle. Let’s hope that digitization does not wipe these publications out because ad-sponsored journalism is not really journalism at all. (When even South Park knows the danger of ad-funded “journalism”, you know something is very, very wrong.)