A Guide to Implementing the Theory of Constraints (TOC)





Next Step



Bottom Line


Supply Chain

Tool Box



& More ...



Lead Times

Finished Goods


& Distribution

& Marshalling

& Healthcare



Competitive Advantage – Just Sufficient Stock

We have seen in the previous section how we can reduce lead time in a make-to-order environment and in particular one where there is an external sales constraint.  If we can provide the same product or service as our competitors but with a shorter lead time then we have a distinct competitive advantage.  However, many organizations are not clear-cut make-to-order entities but rather have a component of make-to-stock in their process or they are purely make-to-stock.

In this situation something is completed ahead of the customer requirement – even if that customer may in fact be internal to the process.  The business has made a decision to decouple the customer from the process lead time by the use of inventory.  Indeed the only valid reason for making to stock is that the process can’t make to order fast enough to satisfy customer expectations.

If we have excess process capacity relative to sales and therefore quite a lot of finished goods; what is the problem?  Well, let’s have a look.  In a number of places we have addressed deviations from the plan (1).  Deviations arise from;

(1)  Not doing what was supposed to be done for the constraint.

(2)  Doing what was not supposed to be done for the constraint.

Here the constraint is sales, and in order to exploit sales we need to make things that customers do want to buy in the near future – we deviate by failing to make these things and thus produce stock-outs.  We also need not to make things that customers do not want to buy in the near future, or at least not make them in excess of the customer’s requirements – we deviate by making these things and thus producing excess stock.  These are the issues that we need to address for finished goods.

We can gauge the amount of overstock by the weeks or months of supply we have for each item relative to recent demand.  We can gauge the amount of out-of-stock to some extent by the need to expedite some items even though we have more than sufficient capacity.  The severity of the problem will differ depending upon whether we are in a make-to-stock environment, or a make-to-forecast environment, so we will deal with each of these in turn.

The whole situation will be further exacerbated if there is a distribution system attached to the process.  However, we will leave that for a page of its own, because in order to understand the solution we must first understand a new concept for replenishment.  We will deal with replenishment on the next page.  Here we will look at the two older concepts; make-to-stock first, and then make-to-forecast.


Let’s begin by saying that in some instances it should be quite possible for a make-to-stock process to migrate to make-to-order if lead times are sufficiently reduced.  To do this we would follow the suggestion in the previous section on lead times; drying the system out, that is reducing the excess work-in-process, and thus removing substantial amounts of the queue time from the system, and addressing batch size issues.  These together will substantially reduce process lead time.  Many businesses will, however, still find that because of their product range (maybe many thousands of standard items) or maybe large seasonal variation in demand, they must manufacture to stock first, and then draw later from stock for sales.  We need to address these types of business.

Finished goods stock in a make-to-stock environment is the result of total work in process (manufacturing lead time) and batching policy.  This is primarily driven by;

(1)  Excessive queue time/work-in-process.

(2)  Batching of product.

(3)  Batching in time.

Lets’ look at each of these in a little more detail.

Excessive Queue Time/Work-In-Process & Batching In Time

Essentially the same mechanics apply here as was the case in the make-to-order environment.  We must dry the system out by reducing the input into the system while holding the output constant using drum-buffer-rope as the logistical system.  Let’s repeat the diagram for good measure.

This will reduce the overall work-in-process and thus lead time.

Batching Of Product

Batching of product, especially batching policy, is of critical importance in a make-to-stock environment irrespective of the lead time.  You may want to check the production section on batching once again to understand the differences in batch type.  In a make-to-order environment we have seen that; either transfer batching – the splitting of batches into sub-batches while in the process, or process batching – the release of smaller discrete batches more often allows for considerable reduction in lead time.  However, process batch reduction has a much greater direct contribution to finished goods reduction than transfer batching does.  Let’s investigate this further in some detail; it is important to understand this differentiation.

In the following diagram we can trace the inventory levels for a particular product as finished goods in a make-to-stock environment over several cycles.  We start with a new batch of goods arriving into the finished goods store from manufacturing or processing.  Over time the inventory level decreases as the stock is drawn down for sales.  At some stage a reorder point (ROP) is triggered and a new batch is released to manufacturing.  Inventory continues to be drawn down until it is replaced with the new batch.

Variation in the rate of inventory draw down and the variation in lead time required to produce the new batch is taken into account by the allowance of some safety stock.  The average stock level over time will be half the maximum stock level.

Let’s have a look then at the effect of introducing transfer batching.

Now, in a slightly exaggerated view, we can see the original batch replacement arrives into the finished goods store as two transfer batches.

If we have already removed much of the queue time from the system, then using two transfer batches will halve the waiting within the run time at each section.  The overall manufacturing lead time should be reduced proportionally.  As a consequence we can supply the batch to the finished goods store faster; therefore we can wait longer before triggering the reorder point.  The reorder point is therefore lower.  Also we don’t need as much safety stock and therefore the safety stock level can also be lower.  However, the average finished goods level in most cases does not substantially decrease.  This is important, failure to reduce finished goods quality exacerbates our deviation of producing too much of things that we don’t need.

Let’s then have look then at the effect of introducing process batching.

Now we can see the original batch replacement arrives into store as a sequence of two discrete and separate process batches.

Again, if we have already removed most of the queue time from the system, then using two process batches will halve the waiting within the run time at each section.  The overall manufacturing lead time should be reduced proportionally.  Once again the reorder point is lower and we don’t need as much safety stock as we initially did.  Now, however, the average finished goods level has also decreased in proportion to the decrease in process batch size.  We can maintain the same level of service with half the finished goods!  All we did was change a batching policy.  In the production section on batching we showed how to achieve this using a “truck and trailer” analogy.  The cost of additional setups to the system is not great – and in any case as sales is the constraint we already have some excess capacity in processing to do this with.

This situation would seem to be a lean manufacturer’s – and any financier’s – dream.  Short lead times, reduced work-in-process, and reduced finished goods stock.  In fact, you can see that there is considerable cash to be liberated from the system that was previously tied up.

The reduced lead time will allow the system to much more closely follow the sales trends and therefore reduce the number of stock-outs.

Process Batching Is An Antidote To Excessive Warehousing

Lets return to a not uncommon situation that occurs at the start of a drum-buffer-rope implementation in a make-to-stock environment where there is a lot of finished goods and there are also significant stock-outs.  Once the constraint has been identified and exploited we will have more capacity to produce more than we did in the past.  There is thus a potential for some companies to make the error of using this additional capacity to chase their stock-outs – they build even more finished goods inventory in the mistaken belief that this will reduce stock-outs, when in fact stock-outs are caused by long lead times and/or forecasting.

The effect is the same as burying cash in the system.  Liquidity dries up, and bankruptcy can become a real option.  This is a cautionary tale for the need to involve everyone; finance, sales, marketing, as well as manufacturing at the earliest stages.  Moreover, it is important to recognize that the right solution to the right problem will avoid this situation from the start.  In this case the right soluiton is process batching rather than transfer batching.


In make-to-stock our finished goods level oscillates between some minimum value protected by safety stock and a maximum value, usually separated by one “batch-size” worth of material.  However, this is not always sufficient to protect our customers especially where demand is highly variable or there is a considerable degree of seasonality in the demand.  The alternative is to make-to-forecast.  Let’s have a look at that.

When do we forecast?  In the following situations;

(1)  We make-to-order and we have insufficient forward orders

(2)  We make-to-stock and our lead times are long/and or demand is variable

The first case is tragic and really concerns make-to-order – we are trying to be “efficient” by keeping people busy.  We don’t need to consider that case any further here but it is worthwhile to point it out.  The second case, however, does concern us.  Many companies build stock because their manufacturing processes aren't fast or agile enough to build to order.  But the stock level targets are usually determined by forecasts, which are based on historical trends, best-guesses about future trends in market preferences, known or anticipated product introductions, planned sales or marketing campaigns, speculation about what the competition will do, and the reading of crystal balls, bones, and chicken entrails (2).”

Chicken entrails aside; forecasting introduces a new and additional factor into our consideration of finished goods levels.  Let’s look at this.

Finished goods stock is now the result of total work in process (manufacturing lead time), batching policy, and forecast policy.  This is primarily driven by;

(1)  Excessive queue time/work-in-process.

(2)  Batching of product.

(3)  Batching in time.

(4)  Forecast Accuracy

In fact, we often have a vicious cycle operating prior to the introduction of drum-buffer-rope; batches are large, work-in-process is high, and lead times are very long.  This almost begs the use of forecasting to “see” beyond the lead time horizon.  We know from our discussion on make-to-stock and previously on make-to-order here how to break this cycle for all but the new entity – forecast accuracy.

It doesn’t seem to be difficult to obtain an average estimate for a forecast value, but to know the variability inherent in that estimated value is much more difficult to determine – and it gets worse the longer the time horizon.  “There are two reasons why we don't see the forecasting error very often.  First, when forecasts are based on the intuition of the marketing people, the concept of error simply doesn't exist.”  “Second, even when the forecasting error is available, and computerized algorithms are able to assess it, the error may be so large that the whole notion of forecasting seems dubious (2).”  So it is not necessarily the use of a forecast per se but rather the noise in the system that gives rise to most of the problem.  Let’s use a generalized example to illustrate this further.

What is the accuracy with which you can estimate next months’ total sales?  Is it of the order of plus or minus 5%.  Sure it’s probably padded for safety, but if you had to put money on it falling either way, would +/- 5% be sufficient?  Now consider a product line within total sales, what would the accuracy be on that?  Twice the previous value perhaps – plus or minus 10%?  We might sell 90% of budget or maybe 110% for a particular line.  How about a line item within a particular product line?  Would you be surprised to see plus or minus 20%?  Customers, the economy, and everything else are just so fickle.

Let’s look at this same problem from another perspective.  If the accuracy of the estimate for next month’s sales is of the order of plus or minus 5%, what is the accuracy for a quarter?  It would be more than 5%, maybe plus or minus 10%.  What then of the accuracy for 2 quarters out?  Maybe plus or minus 20%?  The actual numbers aren’t so important.  Making our intuition explicit is more important.  There seems no doubt that the less aggregate the forecast the greater the potential for forecast error, and the further out that we push the forecast horizon, again the greater the potential for forecast error.

An unavoidable consequence of this is that sometime we are going to produce too much and sometimes we are going to produce too little.  Consequently we carry stock that we can’t immediately sell and we miss immediate sales that were certain.  How can we get around this problem?

We can reduce the impact by reducing lead times substantially and we can reduce the impact by reducing finished goods stock substantially.  Indeed, we may be able to migrate from a make-to-forecast system to a make-to-stock system.  However, to really do much better in volatile markets we must consider replenishment – replenishing our stocks as fast as our market variability.

Be careful; forecasting can be a cop-out.  A cop-out by manufacturing to marketing and sales.  And a cop-out marketing and sales cop-out to the market.  No one needs to take responsibility for making things we can’t sell and not making things that we could sell.  This need not be the situation.  We just have to be as agile as our fickle market and for that we need replenishment.


In an external constrained make-to-stock or a make-to-forecast environment it is easy to make goods that can’t be sold, and not to make goods that could be sold.  The net effect either way is to reduce potential throughput.  The primary antidote to this is to use smaller process batches and thereby reduced total finished goods inventory while at the same time increasing service levels.  However, in a supply chain environment we can go one step further.  We can introduce replenishment.  Let’s have a look then at replenishment.


(1) Goldratt, E. M., (1990) The haystack syndrome: sifting information out of the data ocean.  North River Press, pg 146.

(2) Schragenheim, E., and Dettmer, H. W. (2000) Manufacturing at warp speed: optimizing supply chain financial performance, pp 212 & 67.

This Webpage Copyright © 2003-2009 by Dr K. J. Youngman

Description: web counter