I interviewed Don Creswell who discussed Making Supply Chain Projections.
It’s great to speak with you today, Don, and I’m looking forward to hearing your views today on making supply chain projections. Can you first provide a brief background of yourself?
We come from a company called SmartOrg, which derives out of a long history of decision analysis. Our founders and forbearers have a long trail that started at MIT back with gain theory and things like that. In the 1960s and ’70s, they brought this to Stanford Research Institute—now SRI International—and Stanford University.
The whole idea is to use Beijing probability…” (Bayesian probability theory was developed by the Reverend Thomas Bayes in the 1700’s and further developed by LaPlace. Sir Harold Jeffreys put Bayes’ algorithm and Laplace’s formulation on an axiomatic basis, writing that Bayes’ theorem “is to the theory of probability what Pythagoras’s theorem is to geometry.”) Wow!, try to get your hands around a future where there no facts. Sometimes you find that even the very best big data, whatever you want to look at, while it’s good on the short-term basis and can give you immediate feedback, where things are very unlikely, let’s say a long-range projection of oil price. If oil is at $60 a barrel, you might be losing a lot of money; if oil’s at $100, you make a lot of money. But if you’re projecting something today that’s two years in the future, you really don’t know. How to get your hands around that, that’s where we work. We supply enterprise software to handle this to such clients as Boeing Technology because they’re developing airplane technology that’s decades in the future; oil and gas have a lot of the same problems.
Supply chain where there’s very high volume—manufacturing, particularly; I had a client in southern California who was part of supply chain for a whole bunch of stuff that went into Apple iPads and Apple iPhones, and the demand was subject to a tremendous amount of variation, so how do you get your hands around that when Apple might say, “Christmastime is coming up,” and instead of five billion units, they want seven billion units?
If we have to plan for that in advance, we do sensitivity analysis on our business models and see what the impact would be on our return on investment, for instance, or profitability. If it were to turn out at the lower end and we overproduced and had a lot of stuff sitting in the warehouse on the shelf that we couldn’t move or, alternatively, what would be the impact on us if we didn’t produce enough? And this has happened. I think Apple, a couple years ago, all of a sudden, we hit the buying season, and we ran out of products. We’re really looking at predictions in the long-term where there’s a lot of risk and uncertainty.
Can you talk about what it means to make projections within the supply chain? Do you have any specifics?
I was just talking about this one company that made a component in the supply chain. I’d try and talk about another one; it’s not an electronic supply chain, but we have a client whose supplies involve things like capsules, for drugs, that drug manufacturers have purchased and have to put the compounds in the drugs for ultimate sale and distribution.
Their big problem was that the demand from the big drug companies was very imperfect, and they were saddled with underproduction. The * (3:55—unclear), “How do we plan as best as we can around…if we miss it on the low side or the high side? What impact is this going to have on our ultimate ability to produce or to sell these things and make some money?” In their case, they could not put into the contracts penalty clauses.
I won’t mention any names here, but Truck Company X would say, “I want five million units, and I want them in November.” The company now gears up to produce five million units. October comes and the company says, “Guess what; we don’t need five million anymore, and we’re only going to take four million.” The company is stuck with a million that they’ve made that they can’t use.
We’re talking here about the forward planning, not during the operational, fulfilling the supply chain; we’re talking about planning now. The way to do it is future forecasting about where we want to place our bets, if you will.
Can you talk about how it’s done effectively?
Very few companies do it effectively. Most companies do it very badly. In fact, I’m very surprised how bad some of them do it. They’re watching spreadsheets; they have a tremendous amount of political arguments. What we try to do is say, “Let’s product a couple models.” When you really get down to it, the basic models around these demands, the variables that go into the demand and such are usually not more than fifteen or so top-level variables. Of the fifteen, there’re probably three or four at the most that are really key to whether you have a tremendous problem or whether you’re going to make money or not.
What we do in workshops, we work with these people to understand the nature of the business. We’ll construct a few models; you really need a handful of models. At this level, you don’t model it down to the micro units; you keep it at accurate units. Now, address each one of these factors or variables, as more technical people call them. We say, “Let’s let the experts in our company”—it could be R and D, it could be manufacturing, marketing, whatever—“who has an input to this particular variable, the best information we can get in planning, let’s look at how low it can be, how high it can be for each variable.”
We will put them into the computer. We will run what some people call Monte Carlo analysis; actually, we take it a little bit further than that and use what’s called decision-tree analysis that takes every possible variation of every fluctuation I just talked about, and now it constructs a whole group of curves that say, “This variable has the biggest impact on our final value. The next one has the next value,” and on down.
What you find is that it looks like a tornado. If this thing is at the low end of our value, we can lose a lot of money; at the high end, we could make a lot of money. If it’s cost, you reverse that; if it’s the low end, you make a lot of money; if it’s the high end, you lose a lot of money. Then you rank those down and see where you ought to apply your management at first. Is that the two or three top that can really make us or break us? That’s how this is used. Depending upon the volatility of the industry is how often it’s used; it depends on the industry characteristics. Some of our industries have very long-term between the time they make the plan and the time the plan is actuated; others have very short-term.
The shorter the term, by the way, the better because you can, now, with the data today, get immediate feedback and adjust the model. Have we made a good decision or not? On the other hand, if you’re making plans that aren’t going to pay off for a year from now, which is quite characteristic of some of our clients, then you don’t have feedback that you can now update the model quite quickly because time is going to pass and you don’t get feedback, say, for a year, at that time, you’ll know whether some of your assumptions were on target or not. In those cases, what people do is update this model, say, quarterly or such to make sure the assumptions they’ve made in the model at the beginning are holding, and that’s how they use this.
Is there any more you can talk about where you’ve seen good results?
I’m not going to talk about supply chain specifically. I can talk about investments in product development and R and D and making decisions about which projects and products to invest in. In one instance, we had a very large corporation whose technology people were holding on to their projects much too long. The people who were going to benefit on the commercial side were saying, “Come on, we really need this. We have to get this into the marketplace.
What are you doing?” Management at the top were saying they were going to cut some budgets, and they were stuck in the middle. When we put these kinds of models into place and it was transparent to everybody as to what created value at the end of the stream, the technical people, for the first time, saw they were working on lousy projects. They had no idea before; they were just getting it out. They volunteered to give up projects that should’ve been killed a long time ago or never killed in the past, and they created something like 50 percent additional flow from technology to commercialization within one year. They also doubled the number of projects killed simply by understanding where value was and wasn’t coming from. The head of the portfolio in this case said, “We saved enough money here that was equal to my entire operating budget for one year. What that allowed us to do was take money that would’ve been absorbed by poor projects and allocate it to projects that were going to make a lot more money.” That’s how the uses kind of trade off.
Thank you for sharing today on this topic.
I hope it’s been valuable. It’s a very deep topic, actually, and I’ve seen companies double the value of return in their portfolio within one year.
If you’d like to continue, we can do further interviews if you’d like to go into more detail on this in the future.
I would like to do interviews perhaps more around product development and portfolio management. When I say that, I mean strategic portfolio management.
The words portfolio management unfortunately have two connotations. One is the day-to-day operational portfolio management, the stage-gate management and things like that. That’s all the budgets, that’s all signing human resources to jobs, measuring how many hours you spent on this, how many hours on that; very, very very detailed-intensive and data-intensive. That we do not do. That a lot of people do very well. You’ve got Sofion, you’ve got SAP, you’ve got Oracle, you’ve got IBM; all the big players do that.
We’re one of the few, though, who emphasizes the other side: why we’re doing it in the first place; which ones we should select to put our money in because we can’t do everything; and once we have them in a portfolio, how we manage that portfolio to balance out risk. We want to take enough risk that we can have a chance of making a lot of money on the high end, but we don’t want to take too much risk that we end up losing money.
About Don Creswell
Co-Founder of SmartOrg Inc.