I was on a call with a prospect a while ago whose company is struggling with demand volatility. To address the issue, the company is implementing a statistical forecasting tool. I must admit that it took me some time to convince the prospect that implementing a forecasting tool isn’t going to reduce the volatility. While I accept fully that one’s ability to create an accurate forecast is related to demand volatility, I am adamant that an accurate forecast does not reduce demand volatility.
Demand volatility is an expression of how much the demand changes over time, and, to some extent, the predictability of the demand. Forecast accuracy is an expression of how well one can predict the actual demand, whether volatile or not.
Why do I make this distinction? Bear with me while I go through some personal history as a way of explaining the importance of the distinction. This goes back to my days as a graduate student at Penn State, back when they were still winning national championships!! I had studied chemical engineering as a undergrad and had moved on to industrial engineering and operations research.
Chemical engineering, like most engineering, is a mathematically rigorous course in which complex equations are used to predict the behaviour of complex systems such as a distillation column. The equations are precise and outcomes can be predicted to the umpteenth decimal point. No-one questions the validity or accuracy of the equations used (though there is continued research to “improve” the equations).
In other words there is an implicit understanding that the equations are not accurate to the umpteenth decimal point even though people calculate to that level of accuracy. But the equations are sufficiently accurate to design chemical plants. When it comes to actually running a chemical plant, all sorts of control systems are placed around the equipment to make sure that the plant operates in a “stable” manner.
There are feedback loops and feed-forward loops, and controllers that control other controllers. It is a very interesting study for those, such as me, who like those sort of things. My main point is that these systems are assumed to be highly predictable and that their behaviour can be described very precisely by a set of equations. These are so-called deterministic systems.
According to Encarta, the definition of deterministic is de•ter•min•is•tic [ di tùrmi nístik ] (adjective) Definition:
1. relating to determinism: relating to the doctrine or belief that everything, including every human act, is caused by something and that there is no real free will
2. of knowable outcome: having an outcome that can be predicted because all of its causes are either known or the same as those of a previous event
Clearly, I am referring to the 2nd definition, though the 1st suits my purposes very well too, because I want to bring in the element of free will, which leads to unpredictable behaviour, or volatility. Yet, nearly all of social and business systems are based upon the notion of predictable behaviour. So I had studied chemical engineering and moved into industrial engineering and operations research at the graduate level. A core requirement was queuing theory. If you don’t know what that is, don’t worry you are one of very many.
One of the first things the lecturer told us was that if there was a person checking ID's at the entrance to a bar (I was at university at the time!) and it took about 1 minute to check an ID. If people arrived at the door about 1 person every minute, how long would the queue/line be in front of the person checking ID’s. Fancying myself as quite clever, I raised my hand immediately and replied that on average there wouldn’t be a queue. Sounds reasonable right. After all people are arriving about 1 per minute and it takes about a minute to check ID’s, so the system is balanced. Wrong.
The queue will grow indefinitely. After years of the deterministic world of chemical engineering I just could not accept this. After some "field" research I had to accept it, but I still did not understand it. And finally the penny dropped. Once the person checking the ID gets behind, that is a queue forms, there is no way to catch up. It will still take about a minute to check ID’s and people are still arriving about 1 per minute.
So how is the person going to catch up? The real clue to understanding this is that any time the person checking the ID’s sits idle, because there is no-one's ID to check, is lost forever and cannot be put to productive use. So the available time to check ID’s is actually less than a minute. For those brave enough, here is a reference. So what’s this got to do with supply chains?
Let's be honest, any lead time that is put into an ERP system is an average (at best) or an estimate (at worst). The same is true of production rates and scrap rates. Yet we spend enormous amounts of time and energy fine tuning MRP and APS systems to provide better results, to the point that the results are more accurate than the input data.
(I know that is a contradiction.) But how many times have we heard "garbage in, garbage out" when referring to ERP systems, or other planning systems, and the underlying input data. Well, hello! We’re trying to fix the wrong problem.
There is so much uncertainty related to so many variables in the supply chain that simply having a more accurate representation of the average value of an input variable doesn’t really solve the problem.
I am not questioning the value of accurate information. What I am questioning is the value of spending lots of time and effort to make the inputs very accurate when they are only ever going to be approximations because of the inherent uncertainty in supply chains. On top of it all, so much of supply chain processes – order taking, purchase order issuing, … - is carried out by human beings that we haven’t a hope of creating metronomic repeatability.
Humans are far better at dealing with uncertainty than are machines, but they are also a lot less predictable than machines. Let us embrace their capabilities rather than turning them into machines. Let us give the humans tools in which they can use their judgment, in the face of uncertainty, to evaluate different courses of action quickly and effectively. Which brings me back to the prospect I spoke to a few days ago.
Having a more accurate forecast isn’t going to remove the volatility/uncertainty from the demand. Having a more accurate forecast isn’t going to help the supply side to deal with the volatility on the demand side.
The supply side is still going to have to be agile and flexible to adjust to the demand changes. And I don’t care how much time and effort is put into statistical forecasting, in a dynamic market with lots of volatility, the forecast will always be inaccurate. So the question is where should one spend time and effort. In making the plan as accurate as possible, including all the input data, and forever analysing why the actuals didn’t match the plan? Or in accepting that there is a lot of uncertainty in the supply chain and devising ways to respond quickly and effectively to the change?
Clearly it is important to be able to get a fairly good understanding of the results of alternative decisions, but a quick approximate answer will always be better than a slow more accurate answer, simply because the uncertainty inherent in the supply chain will drown out the “accuracy” of an optimized result. I know there are a lot of Lean and SixSigma people out there who must be frothing at the mouth. So let’s hear your comments and rebuttal of my arguments.
I have to say that I disagree. The demand volatility is a statement of your customers true needs. That you "level out" the volatility is a reflection on your ability to meet true customer demand.
I not advocating that you do not forecast. What I am advocating is that your time and effort will be better spent working out how to respond to demand volatility rather than trying to get an "accurate" forecast or how to "level out" the volatility.
The focus of your blog is the variation between plan and execution. And, I agree with much you said...just not the way you said it. True...forecasts are never precise. However, you fail to mention that the least accurate forecast is no forecast.
Forecasting is only the first step of a strategy to fulfill demand with supply. The next step is actually using the forecast to drive production and inventory operations. And, to have the right product in the right quantity at the right place at the right time while doing so at a competitive cost that allows the enterprise to meet it's financial targets (forecasts?).
FACT: Somewhere along the supply chain someone will carry inventory. As you stated, leadtimes are real and, although not precise, are predictable to a degree. Long lead items are simply items that take longer to produce than the acceptable customer order lead time will allow. Forecasting is most effective when used to smooth demand and drive a level production/inventory plan. No one likes to be jerked around and supply chains are no different. Planning is all about providing the signals for execution based upon lead time parameters. It is mathematically precise even though the master data foundations are hardly rocket science. Even world class planning will drive some inventory (queue) somewhere do to the reasons you mentioned and many more. The goal is to optimize inventory investment at all levels while meeting the objectives stated in the previous paragraph.
True...one organization may be able to use Lean, Six Sigma, JIT, ERP, or some other TLA to supply product based on the demand drumbeat without forecasting. To use an old Southern expression, "Bless their heart."
The total supply chain will never operate optimally without a forecast. Don't be turned off just because your forecast plan doesn't match customer demand fulfillment and execution with rocket science precision. Stop worrying about whether your planning master data is mathematically precise. Does your planning master data produce an acceptable plan that predicts timely execution of production and inventory operations? Yes...almost...OK...get better each time.
Easy quick answers are usually wrong...I'm adamant about that!
We understand that demand is stochastic (the opposite of deterministic: http://en.wikipedia.org/wiki/Stochastic) and determine optimal levels of inventory to meet a target service level based on predicting the probability of demand and also understanding that supply is also variable in terms of new buy and repair leadtimes and yields. Aerospace, high tech and capital equipment companies with a significant aftermarket business must turn to a solution like ours that deals with the uncertainty inherent in the aftermarket supply chain.
Trevor, I agree that Lean is not a cure all in this environment. As we note in our blog (http://blog.mcasolutions.com/Blog/bid/24059/Lean-and-Inventory-Optimization-in-the-Service-Supply-Chain) Lean is valuable in aftermarket service in reducing process variability, but variability cannot be eliminated, which means that very intelligent buffers and planning are required to meet customer expectations in this environment.
SVP Marketing, MCA Solutions
Ultimately, the business that believes the future is predictable using algorithms (or weegee boards) is fooling itself and its shareholders. Supply chain is about balancing the benefit (service, revenue, margin) vs. the risk ( inventory, costs, lost sales.) Funneling vast (yet ultimately limited) resources into forecasting models at the expense of lean manufacturing and inventory optimization tools is a suboptimal solution.
This is a very good discussion. The general feeling here is that I am advocating that we don't forecast. Far from it, though I am sceptical of putting "blind faith" in statistical forecasting methods.
I also want to state that my arguments are mostly focussed on the build-to-order and engineer-to-order segments. I accept that statistical forecasting has a role to play for many make-to-stock industries, though this is less true for consumer electronics and high-tech/electronics.
On the demand side, shorter product life-cycles and reduced customer loyalty are adding a lot of uncertainty to demand. On the supply side, outsourcing and off-shoring has both extended the supply lead time and increased supply uncertainty. Lean and, probably more importantly, the current economic downturn has reduced inventory levels greatly thereby reducing the ability of the supply chain to absorb either demand or supply variability.
It is in this context that I ask the question: Where do you think the next breakthrough will come? From planning better or from responding to plan variance better? Planning is good, but execution is what really matters.
What I am really arguing against is the belief that a plan is going to be realised fully. Or if we only put just the right amount of inventory in just the right place then we can "absorb" all the demand volatility. What if we can't afford the inventory investment required?
Postponement is nothing other than removing inventory as a mechanism of absorbing demand and supply variabilty at the finished goods level and pushing these buffers further up the supply chain, perhaps even to the component level. So how does one respond to demand and supply variability below the postponement level?
I am arguing even more strongly that we cannot "optimize" the supply plan beyond a rudimentary level because:
a) we use deterministic methods to calculate the supply plan when in fact the supply chain is not only stochastic, but it is also highly non-linear.
b) the (deterministic) data we use to calculate our supply capabilities is at best an approximation of our true capabilities
So to what level is the "optimum" accurate? Is it even achievable?
I am not suggesting that we do not plan. We have many customers that plan far out into the future in order to estimate the supply capabilites required within the lead time for building new supply capacity. Or to evaluate the inventory policies to deploy.
What I am suggesting is that we recognise the limitations of the plan in satisfying true customer demand. We should be focussing on the capabilities - processes, network configuration, manufacturing capabilities, systems - that allow us to respond to true customer demand.
I think the breakthroughs will come thru innovation of process and collaboration. My first wish would be to go into a meeting and not just show inventory projections, but the error lines around that future prediction based on historical forecast bias and variability. Collaborating with not just the marketing people on whether this bias is real or fixable, but also with my procurement people on what do we really tell our upstream suppliers (and is there anyway to use this information to build visibility & trust in the supply chain). How this collaboration takes place, how often, and the robustness of it, will be the innovation, I predict.
As far as the demand variability, that's life. I know one product that we were making that had 89% of demand in 3 months of the year with lead-times at > 120 days sometimes. That's not a sustainable business unless you have an employee named Carnac the magnificient, or you have distributed risk to the business over other product lines.
One thing that I've been hearing from smaller businesses is the desire to fire some customers: 'We're just not going to serve customers that have unreliable demand (or monthly hockey sticks)--or give them low-priority'. Customers that help us have even, steady demand are kept. Has anyone else seen this behavior? Is it a recent increase?
Happy new year.
Welcome to a firestorm. Let me see if I can help. In short, folks are confused about demand planning. They try to force tactical demand planning concepts into the shorter-term operational horizon (0-10 weeks) where it does not work very well. This confusion is compounded when there is a volatile demand environment because in these environments, the concepts of Distribution Requirements Planning (DRP) and Materials Requirements Planning (MRP) do not work vey well either. Let me share some insights based on seven years as an industry analyst....
In your blog, you are reflecting the traditional definition of demand planning. Let's face it, demand planning was defined too narrowly in the definition of Advanced Planning Systems (APS). The concept of where companies forecast the future demand based on historic demand, and the forecast is consumed by supply which translates to order fulfillment is not sufficient. The more important role of demand forecasting is in planning.
For that reason, I would argue the inverse to your logic. I believe that as demand volatility increases that good demand planning processes matter more than ever. These are longer term (12-18 months) forecasting processes based on market views. Why? There are five processes that need good long-term demand forecasting processes more than ever with increased demand volatility:
-Building and executing commodity strategies: What positions should we take on commodity buying? What do we hedge?
-Pricing and go-to-market positioning: Which markets do we enter when? What is the price elasticity for our products in these markets? What is the potential for these products and services at these price points?
-Long-term asset planning: What assets do I need to have when? And, where?
-Strategic relationship management: Which contracts do I need to have with which partners?
-New product launch: What is the market potential for the new products that I am bringing to market?
My logic is not solely tied to supply chain execution. The processes of order or contract fulfillment are too limiting. Unfortunately, all too many supply chain executives, were lulled to sleep by the powerpoints of APS vendors, and have largely missed the greater role of demand planning in driving supply chain excellence.
When the recession hit, the companies that did the best job of managing demand and translating demand into source, make, and deliver processes used econometric modeling and what if analysis on market drivers. They understood the impact of the market shifts before they happened and had developed contingency plans.
In make to stock environments, these companies also did not confuse demand planning with demand sensing. Longer term demand planning processes cannot be confused with shorter-term demand consumption or demand execution processes. In volatile environments, the concept of rules-based consumption is too limiting. Instead of using rules, to break the demand plan into time-phased requirements, supply requirements need to be modeled from the outside-in (from the market-back) using statistical modeling to forecast what is needed in the short-term horizon at a distribution center or for a plant to make. Just reacting to order demand will minimize opportunities (cycle stock management, transportation opportunities, etc), because customer orders do not represent true demand.
In make-to-order environments, short-term demand is driven by the configuration of the item. Too few companies have built good systems to translate tactical demand planning into specific material requirements. Configuration systems largely lie in isolation.
Enough people have interpretted that I am arguing against the importance of forecasting that I can only conclude that I have not communicated my ideas very well. I am not, though I am focussing more on execution than the tactical aspects of forecasting. I have no question that forecasting adds a great deal of value in all the tactical planning areas you list.
However, I don't see how this addresses the title of my blog. Having a 100% accurate forecast does not reduce the demand volatility one bit, and it is the volatility that causes all the problems on the supply side. Forecast accuracy is not the same as demand volatility, though high demand volatility does make it more difficult to create an accurate forecast.
The point I am really trying to address is the following: So we have done our best at forecasting and planning, now we need to execute. And guess what, in environments with high demand volatility, the customers are not ordering the items and quantities in the periods we anticipated and supply has not lived up to their commitments. What do we do now? Simply record this as a low order fulfillment KPI? Or a forecast error KPI? I don't think this is good enough.
I am arguing that the divide between planning and execution is in fact a chasm. This chasm will only get wider as demand volatility and uncertainty increases for all the reasons many analysts have written about in the past. There is a great deal of value to be extracted from the supply chain by being able to bridge this chasm.
This is a great dialoque and allow me to share some additional thoughts.
In my view, the need for accurate forecasting has meaning within the context of the business model that the supply chain must support. That may be different for make-to-order or make-to-stock. If product demand is mature and highly predictable, than a forecasting process will certianly pay dividends.
However, in today's dynamic business world where the forces of most efficient supply chain meet the more empowered customer, business models and product cycles are becoming more dynamic, resulting in more demand volatility. Thus, your argument that accurate forecasting may not matter, does have some merit for today's more dynamic supply chains.
My belief is that when demand volaitility increses, planning and execution processes must merge together. A few years ago, I helped SAP develop a vision framework termed "the adaptive supply chain". Putting the marketing intent and ultimate execution aside for a moment, the concept and tenets were that that supply chain planning and execution processes had to come together, far different than the classical MRP/DRP models defined in texts so many years ago. Planning would shift more toward insuring that long-term capacity and critical supply needs were resourced appropriately, along with the ability to perform continuous what-if scenario analysis. Execution would shift to a combination of predicting and sensing of product demand, coupled with highly responsive demand and supply response capability. The adaptive supply chain era has now arrived, and planning and execution processes must now adapt.
Executive Editor of Supply Chain Matters
I do not understand your terminology. You said "Having a 100% accurate forecast does not reduce the demand volatility one bit" Are you talking about the variation in demand even when it is predicted 100% accurately? Ex: 2,100,5,250 as monthly demand? Any one can mange this demand pattern when it is 100% accurate. The problem we face is forecast accuracy not being 100% and also the execution of the plan not being 100% (due to supply variation)
Yes, indeed I am talking about is how the demand changes, even if your forecast is 100% accurate. I use the term volatility to represent the amount demand changes from period to period.
I do not agree that "Anyone can manage this demand pattern when it is 100% accurate". (Well, perhaps theoretically they can with high inventories.) This is absolutely key to my posting. Supply chains are highly non-linear and non-deterministic. The models we use are approximations and the data values are more or less correct. This is the point you make in your last sentence.
The question I raise in my posting is if your forecast isn't 100% accurate and your execution of the plan isn't 100% accurate, what are you going to do about it. The prospect with whom I was talking was making the assumption that having a 100% accurate forecast would remove all demand volatility, and he wouldn't need to worry about the supply side.
There is no doubt that having a forecast that has an accuracy of less than 50% will create a lot of turmoil in your supply chain, so getting a better forecast is a good thing in these situations. Ultimately thoug, the breakthrough in performance will come from responding better to demand and supply disruptions, not from better planning.
Google "Coeeficient of Variation" on Google for more discussion on demand volatility and how this differs across industries.
Pardon my lateness in joining the conversation. I couldn't help but to point out a couple of salient issues with this post that are left unsaid.
First, it is very well-known that bad forecasting induces demand volatility upstream in the supply chain (i.e., bullwhip effect; c.f., Lee, Padmanabhan, and Whang 1997). For instance, in a dyadic supply chain node, the buyer attempts to forecast its own demand. Due to inflated errors in forecast, the buyer generates an inaccurate order to its supplier. The supplier, in turn, observes inflated demand variance. So that while yes, forecast accuracy does not necessarily impact the demand volatility observed by the forecaster, but it will most definitely impact what is observed by the firm upstream.
Second, your post treats demand volatility in the supply chain, or the total variance of demand, as wholly stochastic. To simply put, total demand variance is composed of both deterministic and stochastic components (c.f., Bray and Mendelson 2012). While you are entirely correct that it's a fool's errand to attempt to predict the stochastic variance, but there is inherent value in attempting to forecast the deterministic component. For instance, it is well known that demand for groceries generally spike on the first and the fifteenth of each month. Equally as well known is that sales of bottled water is highly seasonal, as is canned soup. Accurately forecasting the deterministic component of demand variance allows the supply side to better smooth operations. There is a very long line of scholarly studies and industry evidence in support of production smoothing. While yes, explicitly separating the stochastic and deterministic components of demand is difficult, but approximation of each can be done through various statistical treatments.
Lastly, while I completely support your advocacy for more agile supply chain capabilities, it's important to recognize that they are responding to, rather than anticipatory of demand volatility. In other words, it is exactly as you said, humans are inherently more unpredictable than machines. Therefore in formulating managerial responses, behavioral idiosyncrasies may introduce further variance into a supply chain. For example, when demand spikes, supply side immediately needs to consider how much to produce. In the perfect world, a firm has the capability to seamlessly transition its production lines from one product to the next with minimal cost. Yet we also know that this is highly unlikely if not altogether impossible. Even assuming a highly modular product design were in place, questions still surround the optimal inventory to stock for each individual part in anticipation of responding to change in demand. Thus we are back to forecasting all over again!
In short, forecasting itself may not reduce the demand volatility for the forecaster. However, forecast accuracy can most definitely help to reduce demand volatility of the firm supplying the forecaster, which in turn assists the upstream firms to reduce their costs and increase their capacity to serve. The real question is how to share the benefits? Because increased demand accuracy requires accurate demand information. A long line of literature calls for increased supply chain visibility and increased information sharing, yet we also know that companies tend to guard their own demand data as secrets and are thus reticent to engage in such sharing. The real question is not whether accurate forecast reduces demand volatility, but how to share the benefits for an entire supply chain due to increased forecast accuracy.
Great post. The more volatile the situation gets, more focus should be on improving the system's capability to respond to volatility than to focus on improving forecast accuracy. But the general tendency is to focus on improving forecast accuracy when volatility increases.
Of course forecast has a role to play. But it plays a secondary role in volatile conditions. To give an example, in a rainy season, don't look at the daily forecast to decide to carry an umbrella or not. But the umbrella was bought little ahead of the rainy season based on forecast that it is soon going to be a rainy season. So on a given day in a rainy season it may or may not rain but its better to carry an umbrella.
Like in the example, if we know we are in a volatile situation, we have to focus on our umbrella than to look at the weather channel.
Yes, I agree with your idiom of the umbrella. Where the idiom of the umbrella fails is that a company has to decide on timing AND volume. In other words if the rain is going to be really heavy perhaps I also need jacket and wellingtons. But if it is a light rain I only need an umbrella.
Have you read 'The signal and the noise' by Nate Silver? It's well worth the money.
"Prediction is very difficult, especially about the future." - Niels Bohr, Nobel physicist
No I have not read the book you mentioned. Will try to read it soon.
Also Nassim Taleb in general advocates against trying to forecast too much (especially in highly volatile situations) and instead focus on systems to make them respond well when the uncertainties strike.
My realizations are: If volatility is less then we can rely on forecast as future is more or less like the past. But when volatility is high, then we should start less relying on forecast directly and change focus to building better response systems and use forecast information as secondary input for planning.
To illustrate the point, I will give an analogy for a typical situation where some stock quantity has to be maintained for a profitable business.
Let us compare the stock to fuel or gas in your car.
You have to go from one city to another. The route represents lot of uncertainties. So you have to do some forecasting/estimation of the amount of fuel you need.
1. Volatility is less scenario: You know that route has many many fuel stations. So you can actually fix a min level (re-order level) and you can be sure that before you run out of gas you will reach a refill station.
Here you can fix the re-order level based on a forecast of how much gas is needed between two gas stations.
2. Volatility is more scenario: Here you do not have ANY clue about the number of fuel stations and the distance between them. Here the behavior must change. The focus must be keeping the fuel tank full. So when you spot a refill center, go straight in and fill the tank to full. Here the uncertainty is making us focus on our system (is my tank full?) rather than trying to forecast how much fuel I need before I get to the next refill station. In an extreme case, even full tank might not be enough, but given the constraints you did best.
And yes, this "full" tank level is based on a forecast/estimation but this system is better positioned to deal with uncertainties than the ones which are fully exposed to forecasts. Also analogy assumes that it is important to have fuel and keep going. If its okay to not have fuel from time to time, then no arguments.
Hope this analogy adds some clarity to the topic.
Leave a Reply