00:00:00 Introduction to oil and gas supply chain
00:05:02 High stakes and offshore logistics
00:08:20 Financial impacts of supply chain downtime
00:13:54 Complexity and automation in supply chains
00:16:15 Misunderstanding impacts planning
00:18:39 Inventory affects uptime
00:20:33 Oil price impacts profitability
00:23:00 Right inventory for efficient uptime
00:25:10 Optimization beyond spreadsheets
00:28:58 ERP’s decision-making limitations
00:31:10 Excel as decision tool
00:33:15 Systems of records’ importance
00:37:00 Approximate correctness in decisions
00:38:10 Decision focus under tariffs
00:40:30 Numerical recipes enhance focus
00:43:37 Engineering challenge distractions
00:46:10 Large scale reduces agility
00:51:30 Speed and advocacy for numerical recipes
00:53:00 Pricing’s role in the supply chain
00:57:19 Oil as civilization’s foundation
00:59:00 AI and clerical automation
Summary
In a reflective dialogue, Conor Doherty and Joannes Vermorel delve into supply chain challenges in the oil and gas industry, likening them to managing a miniature city rather than straightforward enterprises. Vermorel critiques conventional methods focused on criticality, advocating for automation and numerical models to enhance efficiency. He addresses the financial impact of long-tail events, emphasizing streamlined inventory practices over stockpiling. Vermorel highlights the inadequacy of systems like ERPs in decision-making, advocating for systems of intelligence to navigate future uncertainties. The conversation underscores the industry’s conservative nature, where agility and digital optimization are crucial for advancing supply chain strategies.
Extended Summary
In a thought-provoking dialogue, Conor Doherty, Director of Communication at Lokad and host of the LokadTV YouTube channel, engages with Joannes Vermorel, CEO and Founder of Lokad, to interrogate the intricate supply chain challenges faced by the oil and gas industry. Their conversation serves as a meticulous examination of how these complexities unfold across various operational phases—upstream, midstream, and downstream.
Doherty initiates the discussion by framing the oil and gas sector as a bedrock underpinning global supply chains, noting its profound influence on a wide array of sectors—from consumer goods to industrial materials. Vermorel underscores this sentiment, emphasizing the supply chain complexity that parallels that of aerospace and retail industries. Contrary to presumptions about simplicity, managing an oil rig resembles orchestrating a miniature city, laden with thousands of SKUs, each essential to maintaining seamless operations. Yet, the true challenge lies not in managing product stability but rather in sustaining forward supply chain needs amidst economic constraints.
Vermorel astutely delivers insights on the economic stakes tied to downtime, where even minor sites incur exorbitant costs daily, exacerbated by logistical hurdles presented by remote offshore platforms. Borrowing from Doherty’s framework of complication, complexity, and criticality, Vermorel critiques traditional management practices that focus solely on criticality, often employing methods like ABC analysis. He argues for Lokad’s preference for automation and numerical models over these antiquated techniques, positing that efficiency can be achieved without resorting to large teams of planners.
The conversation veers into addressing low-probability events, or long-tail events, which, despite their rarity, inflict severe financial impacts across operations. Vermorel identifies a gap between perceived and actual probabilities within supply chain operations, warning that service levels may falter without accurately understanding dependencies. While orthodoxy advocates hefty inventories as buffers against risk, Vermorel champions streamlined inventory practices focused on uptime rather than sheer volume, denouncing the inefficiency and counterproductivity of excessive stock.
Reflecting industry financial mindsets, Doherty questions whether maximum ROI is genuinely pursued for every investment dollar, challenging the rationale that supports reliance on manual spreadsheets in an age demanding optimization through numerical recipes. Vermorel resonates with this skepticism, acknowledging the inefficiencies spawned by spreadsheets even in large firms and underscores the need for advanced systems of intelligence rather than mere systems of records.
Vermorel categorizes systems of records like ERPs and WMS as advanced ledgers, adept at tracking data yet deficient in automated decision-making, often nudging users back to spreadsheets due to their inadequacy. Despite the promises of demand forecasting and inventory optimization, these systems oftentimes falter, revealing a need for systems dedicated to decision-making—systems of intelligence focused solely on addressing uncertainties and facilitating complex decisions.
The duo delineates the dichotomy between systems designed for perfect recall of past transactions and those systematized for addressing future uncertainties under real-world conditions. Systems of intelligence are championed for their ability to navigate intricate computations beyond mere record maintenance, poised to elevate decision-making concerning inventory and pricing strategies.
Doherty shines a light on the conservative nature of the oil and gas industry, where companies hoard stock to hedge financial risks yet shy away from software tailored to mitigate these risks. Vermorel conjectures whether the industry’s engineering prowess inadvertently overshadows supply chain optimization, hinting at inherent rigidity compared to agile sectors despite digitization efforts.
Agility, epitomized by responsive software adapting to unforeseen changes, challenges traditional methodologies like FIFO, paving a path toward more dynamic supply chain strategies. Vermorel acknowledges the prevalence of trader automation, particularly in midstream and downstream segments, while documenting the physical complexity at upstream levels where systems of intelligence remain glaringly underutilized.
Transitioning to systematic change involves blending existing manual strategies with burgeoning software solutions—a notion Vermorel encapsulates through Lokad’s dual run approach. This methodology empowers practitioners to juxtapose numerical recipes with spreadsheets, fostering collaboration with scientists to refine and scale optimization efforts while enshrining practitioner expertise for maximum ROI.
The dialogue concludes with Vermorel extolling the enduring significance of oil and gas within industrial society, projecting profound advancements as AI ushers back-office automation into a new era. Doherty reciprocates sentiments of anticipation for future transformations, wrapping up a session that not only combs through the intricacies of supply chain management but also serves as a testament to collaborative enterprises in pursuit of innovation.
Full Transcript
Conor Doherty: Welcome back to Lokad. As an industry, oil and gas presents several unique supply chain management hurdles. Now today, Joannes and I will discuss those challenges across upstream, midstream, and downstream operations, as well as looking at the methods that we think do and importantly don’t work. Now while you’re here, don’t forget to subscribe to LokadTV and to follow us on LinkedIn. And with that, I give you today’s conversation on oil and gas with Joannes Vermorel.
So first of all, Joannes, thank you for joining me again. First question, I think it really kind of sets the table or rather gives a bit of perspective on how we got here. When people hear, okay, Lokad is involved in the oil and gas industry, it’s a bit of a surprise because most of the time when you say supply chain optimization, people think about supermarkets, retail, things like that. So much like a previous discussion we had on aerospace, and I asked you how exactly did Lokad get into aerospace, what was it about oil and gas that drew your attention first?
Joannes Vermorel: I mean oil and gas is, I would say, the very start of pretty much all the supply chains we have. They all start there. I mean, I don’t think there is like almost any product of any degree of sophistication that doesn’t critically depend on oil and gas. I mean this pen has probably like half a dozen of compounds that are derived from oil. This mic, the foam here, this laptop, pretty much everything, even this table with the paint, barrel of oil over there. And then energy, the world runs on oil and gas, you know, for the most part. The alternative energies that we have, the only one that can even make, I would say, a dent in terms of the raw supply of energies that we need is nuclear. The rest is just like a footprint at best.
Over the years, yes, we at Lokad started with the things that were most visible, so the stores, you know, and then you go back to the supplier, and then the supplier, and then the supplier, and one day you end up doing like we do right now also oil and gas companies that focus on the various stages because when you say oil and gas, the vertical is absolutely huge.
Conor Doherty: Well that’s the thing, because again you just said oil and gas is absolutely huge. Yet I also know off camera you’ve said before that in terms of comp, I don’t want to put words in your mouth, but in terms of complexity it’s not that much more difficult than aerospace or retail. Unpack that a bit.
Joannes Vermorel: I would say, yeah, in terms of complication if you want to be precise. Let’s be. So yeah, I mean the interesting thing is that the complexity is through the roof due to the sheer scale. So that means why do we have so much complexity? Well, if you want to operate an oil rig, it’s like a mini city, and it needs everything you would expect from like a midsize city, which is literally tens of thousands of SKUs of stuff, from literally gloves, you know, to aircraft engines. They have engines, I would say, to produce power on site. They would have turbines that they occasionally use when they want to have a massive supply or local supply of electricity, etc.
You have everything from very small things to very big things, and all of it is kind of necessary if you want to keep your operations. That’s true for what is onshore, offshore, same thing for FPSOs, which are like non-permanent platforms to extract oil and gas. So the complexity is through the roof. The complication, I would say, much less so. So that’s a distinction.
Yes, the scale is enormous. Yes, the diversity is enormous. But in terms of complication, I would say the products are relatively stable. Yes, it’s an industry that evolved, but let’s say what you need to supply and have available, it’s not fast fashion. It’s not like every single quarter you’re going to have a completely new catalog and whatnot. It’s still very stable.
And then it is not like, let’s say, aviation where you have supply chain loops where everything is reparable and so you have a crazy complication with the parts that are currently flying, the ones that are on the ground, and a big constant shuffle between the part that you repair and put back to the plane. Here it is still mostly, I would say, they do have repairs for expensive machinery, but it’s still mostly dominantly a forward supply chain. They need tons of equipment that are consumable, they consume, and in order to make really, really sure that they have no downtime, they need to have a constant supply of tons of things.
So yes, in terms of complications compared to the scale, it is not very very complicated, but it is obviously extremely complex.
Conor Doherty: Well, so again just to tease apart there the difference between complication and complexity, how does cost figure into the complexity? So again, if you’re talking about parts, again you made the example of, well, parts are stable in the sense that you know at worst here are the parts that I will need to fix a thing. It’s known. The BOM is known. The quantity you may need or when you need it you might not know, but you know at some point I will need these parts for this purpose. However, the cost of not having those parts are quite different if you’re, let’s say, working in a repair facility on land and working in a repair facility 200 kilometers out at sea on an offshore platform. So is that the complexity, how does that fit into yet another dimension?
Joannes Vermorel: You have complications, it’s whether it is a tight mathematical puzzle where just to think about it it’s super hard. I would say no, it is not. Complexity is just about the sheer number of stuff, you know, the fact that you have many, many SKUs, many parts, many sites, many everything. But fundamentally it’s just more, more, more, more. It’s again, you’re not adding complications.
And here we are talking of yet another dimension, which is all the stakes are super high. And that I would say would be a third dimension. I would refer to that as criticality. So criticality is super high. But as far as a supply chain vendor, supply chain optimization vendor like Lokad is concerned, if you tell me that the downtime cost $1 an hour or a million dollars an hour, this is just a parameter. It is obviously very different for the business, but it doesn’t change neither the complexity nor the complications.
It’s just that the stakes, because then you start thinking in terms of dollars, are so much higher. But fundamentally, this is yet another completely orthogonal concern, which is also indeed very, very important in oil and gas is that the stakes are through the roof.
When you have downtime for pretty much even a small site, we’re talking of something that extracts oil from the ground, yeah, exactly. It’s the downtime, we’re talking of something that is $1 million a day, you know.
Conor Doherty: Some sources as a baseline. I mean that range is quite high. And again, it really is worth just unpacking that a little bit because earlier we talked about aerospace, and I know that like in automotive and aerospace sectors the cost of downtime for an hour is analogous. So I know automotive can be about $2 million an hour, aerospace no reason to think it’s any different than that given the complexity. But again, those typically occur on land.
When you start talking about an offshore platform, you have to also factor into the consideration — it’s not just buying the parts at the last second. You’re also talking about, okay, I need either a boat or a helicopter right now to transport those things. And it’s not just the helicopter I need at the last minute, I need a pilot or a captain. So again, I need a person with a very unique set of skills to be available when I need them for essentially a crisis price. You’re paying through the roof because every second you’re not pulling oil out of the ground you’re losing the value of the oil.
So again when you talk about the financial stakes, I mean there’s first order, there’s second order, there’s third order, there’s fourth it’s enormous.
Joannes Vermorel: Yes, the stakes are extremely high. But again for us, you know, from the perspective of supply chain optimization, the fact that you tell me that plan A is get the parts through a boat and low-cost shipment, you know, it will take the time it takes; plan B is have an emergency helicopter that comes to your offshore platform, and the fact that plan B costs you 100 times more than plan A, it’s again just a matter of settings of the economic modelization.
Fundamentally it is not very different from a perspective from just a multisourcing setup where you have one supplier in your country that is slightly expensive and a more distant overseas supplier that is cheaper. Fundamentally multisourcing, for Lokad, from our perspective, it’s not a very complicated challenge.
But yes, the stakes. What makes really oil and gas very specific is that the stakes, the criticality is like through the roof, and in terms of dollars you have just very frequently two zeros or sometimes three more than in most other industries just because it is so massive.
Conor Doherty: Well I really like the categorization here. I’ve just written down basically three C’s: the complication, the complexity, and the criticality. Those are again three dimensions you just described for how Lokad views, or you and Lokad and our supply chain scientists view the problem.
Joannes Vermorel: Yes.
Conor Doherty: Please contrast that with your take on what the traditional approach in terms of supply chain management in oil and gas is. I mean, are they thinking about it at that level of granularity and dimensionality?
Joannes Vermorel: I mean when you think of it, when you use a classical method like ABC analysis, what you’re saying is that you’re segmenting everything on criticality. You’re saying, okay, whatever is my A — you know, most critical items — I will have let’s say one planner per 100 items so that one person will be dealing with 100 SKUs because those are super critical and I want this person to be able to really closely and daily monitor all of that.
And then the B’s I would say, oh maybe I’m going to have a thousand. So one person, a thousand items, lower volume, lower stakes. And then the C’s, I’m just making those numbers up, you know, 10,000. And those are the stuff that are less critical, cheaper, etc. etc.
So you see, in essence, the traditional way to look at it is just tackling through pure criticality. But the thing where Lokad does not really fit is that we want automation. So we want a numerical recipe that will just deal with everything frontally. And machines can do tons of calculations, so we don’t have to starve in terms of analytics whatever are the infrequent items.
You see the idea that, why don’t you have one demand planner, supply and demand planner, for every 100 SKUs? Why do you only keep this ratio for the A’s, where you can, for the A’s items you can have like one person every 100 SKUs? The answer is because you would need a mind-blowingly large army of supply and demand planners if you were to do that when you have oil and gas companies to manage 50,000 distinct SKUs to keep your operation running.
That’s that. If we were at 100 items per employee, we are talking of 500 employees just to manage the inventory of one extraction site. It’s obviously crazy, you’re not going to do that.
But now if you enter the realm of automation and you have smart software logic, then those constraints are completely moot and this perspective is completely moot. So what you want is just to have something that is very good all the way from obviously your very top most critical SKUs, but also back to the long tail.
Because that’s the thing with this long tail: even if things are rarely needed, they can still contribute to downtime.
Conor Doherty: Again just because I want to be mindful of not using too much technical jargon for people who might not be familiar — when you say long tail you’re talking about low probability events on a distribution?
Joannes Vermorel: Yeah, exactly, exactly.
Conor Doherty: 0.1% chance but that could be $25 million of impact.
Joannes Vermorel: Yeah. That can be, you know, and those long tail situations can either be: we need something that we normally don’t need. Or it can also be that something that we normally supply in one week, easy peasy, takes six months. Why? Because, well, lead times vary or other things.
So you see, all the sources of uncertainty — your continued operations depend daily on decisions that are made now, expecting certain future conditions of the market. We expect that we will be consuming this amount of parts. We expect that the suppliers will be able to supply that in that time frame, etc. etc.
So there are multiple sources of uncertainty and due to the complexity, you can be hurt, meaning you can face downtime just because there is something that had only one chance in 10,000 of happening.
You would say, well, one chance in 10,000 is not much — except if you have 10,000 of stuff like that. So you’re rolling the dice over and over and over and at the end once you roll the dice tens or thousands of times, even things that are very improbable still manage to happen.
Conor Doherty: Well again, this is the idea of probability theory. It’s worth dropping, planting a flag here because again I had a conversation with Simon Schalit, COO at Lokad before. And again, we were talking about aerospace in terms of complexity. The example is valid. And he was talking about how people understand probability theory is very different from what probability theory actually is.
So for example, we gave the example — and you will in real time correct my maths — but if you needed 100 parts to complete a schedule, so I want to fix a thing, I need 100 parts, I need them all to be available to do it efficiently. I need them all available simultaneously. And if you set 99% service level for all hundred of these, the reaction might be, “Oh well then I have a 99% probability of having all of them available.” And it’s actually closer to like 65% or something like that, like two-thirds.
Joannes Vermorel: Yes, yes. Assuming dependent probability distribution. Yeah, absolutely.
Conor Doherty: But the whole point being of course what you think is likely to happen — and even if you start to think the initial steps of probabilistic thinking — it’s still a little bit more complex than you would think.
But I do want to build, because you did say, you know, the long tail events, maybe one out of 10,000, this situation occurs and it will be financially catastrophic. Okay. But the orthodox perspective already recognizes that, like okay, I might have a long tail event. I might be missing this $1 screw. Okay, I’ll keep a million of them on hand and I’ll tie up a million dollars in inventory that might never get used.
So in reality, I mean, people are already kind of aware of this. They already have approaches to it, which is buffer stocks. What’s wrong with that?
Joannes Vermorel: I mean if your industry is super rich, there is nothing fundamentally wrong. I mean, it’s wasteful because stock costs tons of money. But if you have very nice margins, you can tolerate that.
But the reality is that oil rigs and FPSOs have limited storage space. So you see, at some point, if you just want more, it’s going to be on land no matter what. And then that’s far away. It’s far away because again, your FPSO is a very large ship, but it has finite capacity. And same thing for an oil rig. It is like a mega-structure at sea. Yes, it’s fairly large for a man-made structure, but again, your capacity to just store stuff locally is limited.
You have to make sure that you’re really, really making the most of the storage capacity that you have. And generally the problem of saying, “Oh we can just have, you know, sky is the limit for the inventory,” is that if you have this approach you will most likely end up being very distracted with all the dead stock that you carry.
It is not necessarily such a great strategy for high uptime. Because if you do that, you will create huge piles of needless stock a little bit accidentally. At some point, someone will manage, someone will say it’s very wasteful, and then you will be very distracted for months in the process of trying to liquidate this dead inventory.
And being distracted is also a contributing factor to just miss on other stuff and thus face ultimately an accidental downtime because there are stuff that you need that you don’t have.
My observation is that it’s very rarely when people have usually too much stock they usually have also less than ideal uptime. You know, it is very rare that super high uptime is achieved by just having a mind-blowingly large amount of stock. Usually having too much inventory creates so many problems that in the end you’re not even that good in terms of uptime.
Conor Doherty: Okay, well to follow up on — and again I just jotted this down — there’s a lot to cover, but one of the things you mentioned was your ability to tolerate excess stock. I mean, depends on if you’ve got really generous margins.
Well the thing is when we’re talking about oil, those prices fluctuate quite a bit and in real time. So we’re recording this on April 10th. Brent oil is now $65.48 a barrel. In the current moment that could rise to $100, it could drop to $25. How do price fluctuations in terms of the actual value of what’s coming out of the ground affect companies’ abilities to absorb the hundreds of millions of dollars that they’re just keeping on the offshore platform and in warehouses and renting airplanes, helicopters, and boats to ferry all of that? How do price fluctuations factor into the supply chain management process?
Joannes Vermorel: So generally, different sites have completely different extraction costs. You have places in the Arabic Peninsula where extracting oil is very, very cheap, and so at $65 they’re just fine. And there are some places where, let’s say, Canada with the — I forgot the name — it’s the Athabasca Basin.
Conor Doherty: Yes.
Joannes Vermorel: Yes, where at $65 it is really on edge on what is actually profitable.
Conor Doherty: It’s oil sand.
Joannes Vermorel: Yes, Athabasca Basin, those oil sands. So my take is that first, the driving force is that depending on the price point, you will keep certain sites operating or not. That’s like a mega-force that defines what can operate. If you’re at $100 a barrel, then you have way more sites that can be operated than if you are at $50. And I would say this is the mega thing that is on top. Then…
Conor Doherty: It still comes at a cost. You’re basically saying like, yes, because you’re so wealthy you can afford to be wasteful. It’s not a good practice inherently.
Joannes Vermorel: If you can extract the oil for, you see, $10 a barrel and then you sell it at $65, the fact that you have a few extra dollars per barrel of inventory cost floating around is something that you can tolerate.
But if the price is at $65 and then when you integrate all your costs you end up producing at a cost of $60 — when you integrate everything, everything, everything — then you realize that suddenly having plus or minus $1 a barrel of just inventory cost is very, very significant.
But still, the thing is generally — as I said — the problem is that it is not really saving inventory per se. It is more like whatever you can do to have very high uptime.
And in a sense, usually having too much inventory is the enemy. It would be a fallacy to think that just inflating inventory will give you high uptime. That would be a sort of safety stock perspective where you just keep increasing. But the reality is that recklessly increasing so many products that you have to keep in stock creates so many mundane problems for storage and whatnot that in the end you end up with a lowered uptime, not a higher one.
So really the question becomes very quickly, goes, reverts back to having the right inventory, taking into account all the constraints, so that the uptime is really, really maximized.
Conor Doherty: Well I kind of want to summarize this little section of the discussion, but I want to do that in a quote that I’ve written down and then I’ll bounce it off you. Tell me how accurate it is.
Would you say that the oil and gas industry overall, or at least from the supply chain management perspective, does not think financially — and by financially I mean the Lokadian meaning of the term, like maximizing ROI for every dollar that is invested?
Joannes Vermorel: I think, you see, morally that’s what they do. In the fine print, that’s not what they do.
Conor Doherty: Tease that apart.
Joannes Vermorel: So if you ask managers, they would say, “Yes, that’s obviously what we do, obviously.” But then when you look at the fine print of the calculation, it’s actually people doing it by hand.
And you see the problem is that when you have a decision-making process that has people tweaking numbers in a spreadsheet by hand, there is no optimization possible. It fundamentally boils down to a horde of clerks who are hand-down… again we are talking of sites that are large.
We are talking of teams that have to manage tens of thousands of SKUs for every site. I mean this is very complex. And so when you have that many people in the loop and the decisions are made manually, you end up with things that are not very optimized.
Even if the top management thinks in terms of financial optimization, if the base layer of execution of the supply chain are people tweaking spreadsheets, you will not get optimized results. Ultimately, yes, the top manager thinks, “I want maximal bang for bucks. So per dollar, give me, invest in priority for every dollar that we invest in the one thing that will increase the most the uptime.” That’s the logic. Everybody agrees on that, no problem.
But then people at the end of the process, when you go at the bottom of the pyramid, you end up with people with spreadsheets. This is the inside they handed over, but then they have to deal with a spreadsheet. How do they do that? And the reality is that they will do something very, very crude. Especially if you end up with this sort of ABC setup, where there is a person that has the A’s and 100 SKUs, and the person with the C’s that has 10,000.
So my take is that you can only start to optimize if you have a numerical recipe. That’s the first thing — that optimization is not really possible if you do not have a numerical recipe. That will be your baseline.
And then what will be better than this numerical recipe? The answer is another numerical recipe. And because it’s two numerical recipes, you can run them side by side and do a benchmark.
A numerical recipe is a collection of algorithms that goes from the raw historical data and all this sort of extra data that you provide to the final decisions — which is, what exactly do I need to keep, for example, on this FPSO at this point of time, so that I can have the best inventory to maximize uptime.
Conor Doherty: I want to follow back on something, and it comes back to, I guess even before I joined Lokad, if you’d said to me — and I mean no disrespect to anyone who works in bakeries or anything like that — but if you said to me before I joined Lokad, how does that small independent bakery run its supply chain or manage inventory, I’d say, “They use a spreadsheet.” I’d say, okay, sounds about right. It’s like two people, limited products, that’s totally normal.
If you’d said to me enormous aerospace companies or oil companies or gas companies just use an Excel spreadsheet and are not accounting for the kinds of things that you just described, I’d have said, “That can’t be right. That’s way too futuristic and complicated a process to just be using something like Microsoft Excel.” Not that there’s anything wrong with that, but to accomplish the kind of things you’re describing, you need to go beyond that.
So my question is, what exactly in your opinion is it that’s holding back an enormously consequential and profitable industry from embracing the tools that you just described — again, numerical recipes?
Joannes Vermorel: In this regard, oil and gas is just like most of the other verticals when it comes to supply chain optimization. Systems of record will never deliver fancy decision-making processes. So all those companies have systems of record.
What are systems of record? That’s the ERP, WMS — that’s the thing that keeps track of what you have, the inventory movements, the payments and everything. Records, just raw data. That’s like a glorified ledger that contains more than an accounting ledger, but still it is basically a glorified ledger.
Those things, all the vendors who provide systems of records have been claiming and failing for decades that they can automate decisions on top of that. It turned out that it’s a very bad idea to even try to do that in a system of record. The software architecture of those systems is absolutely not suited to do that.
So vendors end up, I would say, engineering stuff that proves very underwhelming in practice. Pick any ERP on the market. Those ERPs should have been named ERM — Enterprise Resource Management Systems. Pick any ERM on the market, and you will see they all have capabilities like demand forecasting, inventory optimization and whatnot. They all have — at least on paper — that.
And yet people do Excel spreadsheets. Why? Because those capabilities are crap, do not work. People just try them, they see that it’s not even close to addressing the problem, and so they fall back on their spreadsheets.
Oil and gas doesn’t have a truly exceptional story in this regard. The same problem has been happening in many other verticals. That was the same problem in retail, same in manufacturing, same in aviation. The problem was really ubiquitous.
Those systems of record are fundamental, because that’s how you have the electronic counterpart of what is happening in your supply chain, what are the operations taking place — so that’s fundamental. But those systems will not, and will probably never, be extended as systems of intelligence.
I define a system of intelligence as something that is entirely geared toward automating a decision-making process. And here that’s just not happening. It didn’t happen. Those systems of records have been floating around since the late ’70s. People have been trying to automate decisions with those since the late ’70s.
Most ERM vendors have at least half a dozen iterations, sometimes failed iterations, on their websites about the various stuff they have tried and failed. And it’s not coming from those people.
So now the question is: if companies in oil and gas want to get serious on that, they need to consider the fact that the solution for optimization will not come from those systems of records. It will come from something on the side.
And by the way, it already does — the Excel spreadsheet is already something on the side. So the reality of their operation, the tooling to support the decision-making process, is already on the side, because the system of record itself is just not suitable for that.
Conor Doherty: Well, this is the thing because this is a very important point, but just to frame it, I did in the background do a very quick search. And I literally just typed the words “ERP smarter decision-m” and there are endless results.
And the first one — and I’m not going to give any names, it may be gone by the time this is aired — but “Seven financial ERP software solutions for smarter decision-m.” Next result, “Smarter decision-m with ERPs and business intelligence.”
Now that’s actually the second category once you get into BI tools. That’s system of reports.
Joannes Vermorel: Yes.
Conor Doherty: So again, you’ve nicely laid out I think the distinction between a system of records and a system of intelligence, and then system of reports in the middle — like analytics of your raw data.
But what you have, I think, yet to maybe just unpack a little bit is what is it about that third class — the system of intelligence — that makes it irreconcilable with an ERP or system of records?
Because again, you said, well you’re trying to make your ERP do something it can’t do, and it’s like you can’t get blood from that stone essentially. Why is that? Because it seems like people are being misled, is what I’m saying.
Joannes Vermorel: Yeah, I mean because you’re not tolerant and intolerant to the same thing. If you have a system of record and there is a calculation that is $1 off, just $1, the accountant will go nuts. It is not allowed.
Even if it’s a $1 million payment and it was rounded to $1,000,010, is it a big deal? The accountant will go nuts. It is not even possible — unthinkable. You do not round things even by $1.
So in a way, when you’re dealing with systems of record, you want complete purity on a long, long list of things where you’re like a maniac. It takes a lot of time, a lot of efforts because you want to have absolute purity in many of the data manipulations that you do.
And in terms of latency, you want all of those calculations to be super fast. Why? Because I want to know how much stock do we have right now for this product — I want the information instant. I want to create a new entry — instant. I don’t want to wait. Everything is simple. Calculations need to be perfect and they need to feel instant. Obviously real time does not really exist, but they have to feel very quick.
Decision-making is completely different. Can I approximate something that was $1 million and $10 at $1 million? Yeah, absolutely. Can I just ignore tons of things? Yeah, absolutely. You want to be rough but correct.
When you have so much uncertainty, just imagine you are about to — you’re facing a situation where you have super erratic consumption for a part, super erratic lead time for the suppliers supplying the parts, you’re not even sure exactly which price point you’re going to get the part at, because there is also uncertainty.
So you have like three uncertainties — the demand, the lead time for the supply, and even the price at which you will ultimately source the thing. Now you have to decide now: do you want to trigger a purchase request for this many units of this part? You can see that we are not even close to being 1% precise. What is critical is to be approximately correct as opposed to being exactly wrong.
The system of intelligence is entirely designed in terms of numerical recipes to do that. And that means you’re not even focusing on the same things. You will readily approximate tons of things if those things are inconsequential. In contrast, you will do things that an accountant will never do, which is speculate all the time on things that might happen.
A system of record fundamentally will tell you if you purchased parts in the past, at which price point, what was the price point you paid. It’s not going to speculate on the fact that maybe the price point you’re going to pay will be much higher in the future. That’s not the sort of thing that belongs to the realm of records. But it belongs to the realm of systems of intelligence.
Conor Doherty: So again, if I’ve understood correctly, the way I would try to see the distinction would be — again an ERP — it’s the difference between a record and a decision.
A record is just a reflection: there was one pen on the table, I took that pen off, there’s been an update of that record. Whereas a decision — should I buy a pen? Well, how much does the pen cost, where’s it going to come from, what am I going to sell it for — that’s computationally coming up with a decision. It’s much more involved and sophisticated than just a record of a thing.
Joannes Vermorel: Exactly. And you’re not even paying attention to the same thing. For a system of record, you’re really looking about the past and you want to have perfect accurate recall of the past. Things have to be clean, compliant and whatnot.
Let’s just consider for a second what it means. I need to purchase a part that comes from China. I don’t need it now, I need it in six months. Should I buy now with a crazy tariff, or do I think that the thing will have settled down? Or do I think it will be even more crazy and we will end up with a 200% tariff?
It’s the sort of things that do not belong to a system of record. You’re going to drive your auditors and accountants nuts if you put this sort of stuff in the system of record.
But as far as a system of intelligence is concerned, yes, this is exactly the sort of thing you see. The focus is really different. You’re ready to approximate tons of things that are inconsequential.
You don’t need to have the transactions to the last dollar. You have plenty of tiny costs that are real but can be neglected because they are less than 0.1% of the total cost.
Again, an accountant cannot ever say, “But it was just a $20 spending, frankly we should not even be recording that.” No, no, no. From an accounting perspective, yes it was just $20 but you record that, even if it was just the stamp to get a $100,000 shipment out of the door.
But again, from a system of intelligence, you’re saying, “Okay, I don’t care about this thing. It’s literally inconsequential. It complicates my logic for nothing.”
I really want to have my logic — my numerical recipe — focusing on the big stuff, the stuff that is really consequential. Due to the fact that my numerical recipe should not be a monster — it should not be thousands of lines long, completely incomprehensible — I need to focus on what really matters, ignoring what mostly doesn’t matter.
There is a limit to how much sophistication you can put in this numerical recipe before the thing collapses on itself just due to the fact that it’s not maintainable.
Conor Doherty: Just before we push on, I do want to again drop another pin there. When we talk, we’re not downplaying the value of a system of record — obviously that is critical — and a system of reports, very, very useful.
So the nomenclature is simply just to differentiate. When we say intelligence, it doesn’t mean the others are dumb. It’s just that’s the function. System of intelligence produces decisions.
Joannes Vermorel: But I would argue that it is very dangerous to not have a dumb system of records. You do not want to have something smart and clever. Just ask an accountant and tell him, “You know what, your colleague, he’s doing very clever things in accounting.”
It’s very imaginative and smart. The accountant will be terrified. “What? Imaginative accounting? No, thank you.” I would prefer it to be super dumb and rigid. And yes, there are many rules, but they are super basic.
Don’t get me any kind of fancy creativity here. Again, a system of records has to be the electronic ledger. You really want to minimize the sort of — I mean, sophistication is not something appropriate for a system of record. We want this thing to be super, super dumb, super simple — as simple as you can make it — because it is not the right place to engineer sophistication.
Conor Doherty: Well, if I were to summarize a lot of the commentary so far, it basically reads like risk aversion is very common in oil and gas. But there are two forms: there’s risk aversion when it comes to inventory — which makes sense considering the criticality and the financial stakes you described — but there also, if I’ve understood you correctly and I think I have, there’s also a lot of risk aversion when it comes to even software choices to address the first class of risk aversion.
So I don’t want to lose lots of money because that’s not what I’m in this business to do. So I’ll keep lots and lots of stock on hand. Okay, well here’s a piece of software that is designed to address that issue. No, I’m also risk-averse about using that.
How, in your opinion, do companies reconcile those two forms of what are surely at face value conflicting forms of risk aversion?
Joannes Vermorel: I think the problem is slightly — you know — it’s not framed exactly like that. Oil and gas companies are, I think at heart, engineering companies. Those are companies of engineers. Those are very technical challenges.
The reality is that the engineering problems that oil and gas extraction, transport, distribution represent — the engineering problem — is extremely interesting. It is very complicated.
And so I would say you end up with a situation — which also is a problem faced by some other verticals — where it is very easy to have all your best and brightest engineers doing the things that are the most interesting, which is inventing the technologies so that you are able to extract new sources, think of new technologies to operate everything, to transport everything.
You see, supply chain is also an engineering discipline. But when you have something just next to it that is super shiny and bright and extremely attracting, you can end up with a situation where those companies struggle a little bit to get the talent they really need on the supply chain front.
The problem is caused by the fact that there is a lot of interest for the core oil and gas engineering challenges that are distracting people away — and companies a little bit away — from the supply chain challenges.
Again, that is something that is not unique to oil and gas, but I think it’s something that tends to impact all those tech-heavy verticals. Semiconductors would be the same. To some extent, fashion would be the same — because if you’re in the fashion industry, your interest is in fashion, not supply chain, etc.
That’s typically the sort of verticals where people who join those verticals might be more inclined, if they have a sort of technical inclination, to tackle the core challenges of the domain as opposed to tackling the supply chain sub-challenges of the domain.
Conor Doherty: Well, having said all that and taking all of that into consideration, do you think then it is feasible and/or realistic — so you can choose what you prefer, feasible and/or realistic — to expect that supply chain in oil and gas, oil and gas supply chains, will ever be as agile or as proactive as in other verticals?
Because again, like the ones you listed…
Joannes Vermorel: I mean first, the scale makes them by design more rigid. The bigger you are, the less agile you are. That’s just a given. You can try…
But there are degrees. Amazon is famously known for being quite good at not being a complete bureaucratic nightmare despite being very, very large. But still, it is very difficult when you’re super, super large to preserve agility.
And here, oil and gas is literally — there is no bigger industry. We’re talking of projects starting at $1 billion. Things are extremely large. So in terms of agility, it is okay to… I mean, it is not even a realistic perspective to say oil and gas is going to be as agile as, let’s say, e-commerce players.
It’s just not a reasonable baseline. But they could become a lot more — I would say, considering the proper baseline — they have massive room for improvement. And again, I believe that with the fact that nowadays most of those companies are digitalized — so they have those systems of records in place, they have already done the big investments — to start optimizing.
And the interesting thing is that they very frequently made originally those investments for systems of records thinking that they would get a system of intelligence at the end of the journey. The reality is yes, but typically not with the same vendor.
So you see, the reality is that yes, your system of records is a foundational building block to later do the optimization, but you’re not going to do that in the system of records. You’re going to do that on something else — in a system of intelligence — and most likely it’s going to be a different vendor.
Conor Doherty: If I can just unpack there, so the degrees of agility — what I meant by that question or rather how I would define agility would be with an example.
You have a system of intelligence that is able to react in semi-real time or quite quickly — I think is what we said earlier, quite quickly — to the current state of your supply chain.
Take an example: you’re performing repairs on the offshore platform, the bill of repairs you think you know, but it’s variable because you suddenly find, “Oh, there’s actually a problem I wasn’t expecting.” Okay, it’s there now. Am I prepared for that, yes or no?
I have a schedule of repairs ready. I have the technicians. I have the tools. I have the parts. But what I thought I was going to have to do, I can no longer do because actually that thing, which I don’t have the parts for right now, is broken. What do I do?
In my understanding, an agile supply chain would be, for example, one that has software that could regenerate a system of actions — a schedule of actions. Might not be perfect. And again, we get into the idea that perfect doesn’t exist. But it would be better than something like, “Well, throw our hands up, just go with FIFO, whatever, figure it out.”
Joannes Vermorel: But the reality is that, okay, if you have humans in the loop for your decision-making process, especially if there are many humans involved, it’s going to be slow. It’s going to be slow. Just imagine right now, we have a situation where tariffs in the US have moved considerably. Now imagine you have hundreds of people involved and they will need to update their spreadsheets to take into account the new situation.
If you proceed like that, you know, you will shoot — okay, top management has noticed that — they will shoot an email to everybody and say, “Hey guys, I think you’ve all been reading the news. Here is an update on the new situation, blah blah blah. Just update your practices to reflect the new reality, taking those numbers as inputs.”
But the reality is that it’s going to take time. People have their spreadsheets. They can be complicated to update. It’s not centralized. Some people may not be really paying attention. Some people might already be fighting so many fires that they just don’t have time to deal with that.
So if you generate your decisions with people in the loop, then it’s going to take six months no matter what to bring everybody up to speed.
And when you consider the speed of change of the tariffs in the US with regard to the rest of the world — hour by hour — clearly having a six-month horizon for getting your very diverse teams up to date is just… I mean, it’s just going to be extremely slow.
So yes, that’s one of the reasons why Lokad advocates numerical recipes. You can, in let’s say a day, update the numerical recipe, take into account whatever is the new reality, test, and then it goes to production. And all the decisions that you take onward are now reflecting the modified recipe, which may include the new tariffs or whatever is happening currently.
Conor Doherty: Well again, I realize that while we’ve discussed this, we have implicitly and occasionally explicitly focused on upstreams — we were talking about extraction and offshore platforms — but everything we’ve said here applies across the board, so upstream, midstream, downstream, etc.
Is there anything uniquely different about midstream and downstream supply chains in oil and gas that would make it perhaps more suitable or more amenable to the kind of software interventions you’re describing, or is it just all the same — turtles all the way down?
Joannes Vermorel: Yeah, it is very, very similar. I mean obviously, once you get into transport, there are tons of people who are really traders who are evolved. And by the way, it’s interesting because they don’t consider themselves supply chain, but from my perspective, pricing is part of supply chain.
It’s interesting because those things are completely automatized, robotized already. We have those quants that are dealing with this part. So the interesting thing is that when it comes to speculating on the goods themselves, buying it, and setting the market price, those things are already completely software-driven.
So that’s interesting. That’s the upstream where there is the bulk of the physical complexity, very asset-heavy. It’s still, I would say, under-instrumented when it comes to systems of intelligence. That would be the big differentiator.
Conor Doherty: Well, we’re kind of winding down here. But what I would say is, based on what you’ve just said — you know, under-instrumented, I like that phrase — realistically, no one’s going to go from a completely orthodox, classic approach to managing, let’s just say, upstream supply chain to, “Well, it’s completely robotized from start to finish,” excluding Lokad’s clients.
For people who want to start taking those first steps, what does that look like in terms of software?
Joannes Vermorel: Software — it’s just dual run. The way Lokad approaches those situations is you establish your numerical recipes, and initially people just keep having their spreadsheets. But they just have on the side what Lokad — the numerical recipe of Lokad — recommends.
They can compare, and they can decide whether — which one is the best. And Lokad is not decisions; we also, as part of the white-boxing effort, we provide explanation in dollars that justify. Every decision that we recommend comes with typically half a dozen performance indicators — themselves in dollars — that explain why we think that is needed. That could be the cost of what you’re about to buy, the cost expressed in extra uptime that you’re earning, etc., etc. So you would have half a dozen performance indicators in dollars that motivate the plus and minus, that motivate what we recommend.
And then we iterate. And at some point — that’s the way those numerical recipes graduate to production. The way we think is that when supply chain practitioners say, “Well, today I’ve just validated all your decisions, just like yesterday and the day before, because frankly they’re just good. I don’t see any added value.”
Then we have fixed the problems. It’s just that. And the interesting thing is because you have automation, you’re not asking your teams to work twice as hard — to generate first the decisions with your spreadsheet with a semi-manual process, and then do it in a second system, also with a semi-manual process. That is just a nightmare for the team.
The idea is no — you want the new system to be completely automated, robotized, and then you can run that — do what we call dual run — for as long as it takes. Typically a couple of months until people are very reassured that those things are solid, that they’re good day after day, that they make actually much fewer mistakes than people. When people start to say, “Oh, the system disagreed with me… Oh no, I got the lead time wrong.” Okay, the recipe was correct. Then you just decide to automate.
Conor Doherty: Yep. And also it should be pointed out that any of the supply chain practitioners client-side — they’re able to interact with the supply chain scientists in charge of that account, and collaboratively that does help to improve the numerical recipe to factor in their insights.
Because again, no one’s making the claim that there’s no value to what resides inside the head of a supply chain practitioner — simply that let’s leverage that and maximize the return on that investment through scalability and automation.
I think I heard you say that once, Joannes. Well Joannes, I have no further questions. But as a closing thought, any call to action you want to share?
Joannes Vermorel: I mean, the world of oil and gas is literally the foundation of our industrial civilization, and it’s not going anywhere. It’s not going anywhere. Despite the claim that there is peak oil and whatever — no, it’s there to stay. It will stay for a long, long time.
And even if the world managed to transition to nuclear energy for the pure energy thing, it turned out that there are plenty, plenty of cases where it’s just not very appropriate. If you — for example — electric planes, we don’t even have anything in terms of technology that would make that work. Same thing for, for example, foresting. You need very heavy-duty trucks to do that. Those things are not going to run on batteries. Batteries and more charge. And that’s true for most of the heavy equipment that we use — for agriculture, for mining, for tons of things. Those things depend on oil.
And then there is just also the plastic that we need for tons of things. And unlike what the media is saying — “Oh, we have too many plastics with packaging” — yes, but most surgical instruments are also made in big part in plastic.
So yes, it’s not going to go anywhere. And I think this industry is also having an engineering mindset. If I were to make a slight guess about the coming decades, I would suspect that this industry is also going to take the train of AI and just automate tons of clerical back-office tasks. I mean, that’s what we’re talking about here — just there is this industry employing literally hundreds of thousands of clerks doing back-office jobs. They are absolutely needed because otherwise those companies would grind to a halt. But there is an enormous potential to just mechanize that and free those people so they can do more interesting things.
Conor Doherty: Well Joannes, I share your enthusiasm. And I certainly thank you for your time. And I thank you all for watching. See you next time.