We all know that forecasts are always wrong.  So why do we make them, then?

Here’s a story about some decisions made in 1992 still relevant 13 years later, based on forecasts made then.

The question was “What freeze drying facilities should be provided to cater for lyophilization demand over the next decade?”

It’s always best to be flexible about plans for the future. Business changes, new products come and old ones go. If you are lucky, you can cope with most capacity requirements by hiring more people, working more shifts, renting some offices or using simple easily reproduced manufacturing processes. However, the chances are that you have got some processes that require large capital expenditure, take a long time to put into place and can’t easily be changed once you’ve built them.

Lyophilization is one such process. Great for ensuring long term stability and activity of reagents of biological origin, but requiring special equipment, elaborate validation and special facilities. So, when you are planning lyophilization facilities for product lines lasting more than a decade, you face some severe forecast challenges that boil down to “how many and what size should the lyophilizers be?”  Errors in judgment can easily result in running out of capacity too soon, requiring additional capital expense, not to mention a probable space problem (where do we put it?), along with having to rush an unrushable validation program.

Alternatively, you will have to justify surplus capacity “just in case” under the spotlight of the annual budgeting round.

A forecast is needed of the likely products and batch sizes, of course. What is needed more, though, is a recommendation that is least sensitive to this forecast being wrong.
Here’s an example. The author was asked to answer this question in 1992. In a new facility, how many and what size lyophilizers are needed? For the needs of the next 13 years!

There are a range of variables: 

• A new range of products under development
• Unknown new components to be lyophilized
• Volume growth or decline in each
• Batch sizes dependent on shelf life and demand
• Unknown marketing initiatives in the future
• Changes in lyophilization cycle time
• Regulatory rules as to what can and can’t be lyophilised together
• Existing components still in design that might or might not need lyophilizing

Forecasts produced a chart for total lyophilization volume as follows:

Fig 1

With the minimum forecast volume 20% of the maximum, there was at least a conviction that zero was not an option!   The business had faith in its long term future.

So, what were the options?

The latter options were tested first: avoiding capital expenditure is always welcomed at board approval level. However, it was impossible for R&D or marketing groups to imagine a future where nothing was lyophilised, even if as a contingency. Experience with the biological components used in this industry had to rule here.  Subcontractors were sought out.  However, even those that could possibly have helped in the short to medium term could not be relied upon to provide long-term security of supply. Here was a strategic process that needed to be kept in-house in order to maintain that security. The subcontractor option was ruled out, at least for the “most probable” forecast.

Now the original question had to be answered: How many and what size should the lyophilisers be?

The forecast had produced an aggregate volume from a listing of current and projected components.  What it did not do was to clearly identify what batch size each would be made at.  This is crucial. If you want to lyophilise 20,000 vials a year, you could do it with different sizes of lyophiliser as follows:

• 20,000 vial capacity and do the task once a year, using approx 1/360 of its capacity with a drying cycle around 1 day.
• 5,000 vial capacity and make 4 batches in a year, using 1/90 of its capacity
• 60,000 vial capacity and lyophilise 3 different concentrations in the same cycle [if the product was supplied that way]

This provides many options.  What if product performance, at some time in the future, requires that a component must be made 4 times a year [perhaps its shelf-life has to be reduced, for example]? If you had specified the 5,000 vial capacity machine, then this would make no difference.  If you had specified the 20,000 vial machine, then you would still have to use it 4 times a year, but each time only filling it ¼ full.  Its capacity utilisation would be the same as the smaller one.  In this situation, why buy the larger one?

Multiply these contingencies by the full product projected product range, and you can see that building large to provide vial capacity does not necessarily provide capacity that can be used. At the other extreme, if you build too small, you will be able to use the capacity better, but at a price.  Sometimes you will have to make batches smaller than you really want in order to fit the lyophiliser space.  Making more batches to do this will incur the batch related expenses of doing so, e.g. more formulations, more QC testing.

So, how best to resolve this?

Using a spreadsheet to calculate the required number of batches across all the predicted products, the effect of lyophiliser size on the number required can be seen (Fig 2).

 
Fig 2

The exercise can be repeated with maximum and minimum estimates to give a spread. The number and size can be chosen to cover the worst case, without providing an excess.

In this case study, this was sufficient to make the choice and demonstrate its worth.
If necessary, the analysis can be extended to include balancing the capital cost of larger lyophilisers against the extra revenue expense of making and testing the additional batches that the smaller one will demand.

What Actually Happened?

After 13 years, the actual volume was within the forecast band.  During that period, volume had dropped off dramatically, whilst at the end it was on an upward trend.  Other completely unpredicted things happened too. New components were introduced (as had been expected), but these were only needed in very small quantities. The effect was to demand many more lyophilisation cycles, but each one only containing a small quantity.  With hindsight, these components could easily have been dried using a much smaller lyophiliser instead.  However, the machines were built and installed and once there could be used for anything. Such has been the case.  Capacity to still sufficient to meet current demand, and is certainly not too much.  The forecast has worked out right!

Now What?

At time of writing, the existing facility will be needed for several more years, but then may be replaced.  The 1993 question has been asked again.  It has been answered in the same way.

by James La Trobe-Bateman, reMODEL Consultants International Ltd

Internal Competition for a New Healthcare Product

How do multi-nationals decide where to make new products? In this case, the choice was between existing facilities in the USA or the UK.  Each existing site made a product similar to the new one. Each had an established technology that would be used at the chosen site. The company also had a long-standing strategy to rationalize facilities in the two countries.

There was a prima facie case for the US site. It was local to the market for the new product, the time to obtain regulatory compliance was shorter and the current unit manufacturing cost of the existing product range was 20% lower.

Other considerations favoured the UK site.  Activity based analysis of the existing product lines in the two countries showed that marginal costs were much lower in the UK.   The UK used a process technology that was more modern and inherently more flexible.  Further, the existing US product was in decline, whilst the UK made one was still growing fast.  Capital plans for expansion could accommodate the new product without excessive new expenditure.  On top of this were financial advantages from tax benefits and local grant-aid.

The process of reaching the decision was effectively a competition by each site to bid for the new product.  Costs, inventory, responsiveness and capacity for both current and predicted new operations were presented.  This levelled the “fact” playing field. Each party had to assimilate the knowledge of the other over a range of issues such as regulatory matters, process reliability and operational efficiency.

They then had to commit to improvements to come up to the performance of the other. This commitment was more important that the final decision.

by James La Trobe-Bateman, reMODEL Consultants International Ltd

This example shows how reducing the response time of the supply chain reduces the variation in supply and provides better control.

The example product was made in a single factory and shipped worldwide through a number of distributors to supply a market with a steady daily demand.  In other words, the true demand was completely flat.  The stock and ordering policy of the distributors adds variation which is reflected in variation of the factory’s output.  At the start of the period shown in Fig 3, the manufacturing response time was 18 weeks which coincides with the time between peaks of supply.  It can also be seen that the supply quantity varied by +/- 30% representing quite a strain on factory management (feast and famine).  Batches of the same product were made every 6 weeks and manufacturing lead time was 12 weeks for some of the components.

Fig 3

During the middle of this period, the factory was changed so that lead time for the most-awaited component was reduced to 8 weeks and batches made for shipment every 3 weeks. The right part of the graph shows that the peak to peak time reduced to 11 weeks (the new response time) and the amount of variation reduced to +/- 15%.  In this new regime, everyone was much happier and felt more in control.  A fringe benefit (actually probably ultimately worth more in Sales) was that the customers received fresher product and could keep it for longer before it reached the end of its shelf-life.

A further observation is that the reduced variation did not happen as part of a conscious effort on anybody’s part. It seems that the distributors “sensed” the greater responsiveness of the factory and adjusted their ordering and stock policy unconsciously.  In other words, it is human psychology that puts the variation into the demand (“I don’t trust the factory to deliver when they say, so I will order a bit more, in case”.  Followed a few months later by “I seem to have far too much, I’d better cut back drastically on my orders”). Greater responsiveness leads to greater trust and a more realistic ordering pattern.

Fig 4

It is worth also mentioning that manufacturing cost was held neutral, whilst inventory came down.  The chart shows on the same timescale how the inventory level of 3 of the components reduced as implementation progressed.

by James La Trobe-Bateman, reMODEL Consultants International Ltd

In most industries, demand varies in the short, medium and long term. Such variability is why demand forecasts are inaccurate. Fluctuations in demand are notoriously difficult to predict and this has driven a trend towards shorter lead-times and “make-to-order” production. If demand in a factory was stable, its output could be set to guarantee satisfying that demand. Even when demand changes, such a factory can respond within its capacity if the forecasts are accurate. In both situations, the time it takes to process material or the interval between making batches would have no effect on the success of the factory in meeting demand.
However, in reality, demand does change, new opportunities for sales present themselves and existing customers change their minds about product mix.  These situations are impossible to forecast accurately. So, in real life, the factory must respond to those demand changes.  At those moments of unexpected changes in demand (and only then), it becomes apparent whether the factory responds like a supertanker or a jet-ski. Various aspects of product and process design influence responsiveness, but their effect can be reduced to two measurements: the interval in time between batches and the time it takes to process a batch.  “Response Time” is a measure of the ease with which manufacturing can change its output to meet changes in demand.

Definition
Response Time = Batch Interval + Process Time

An Analogy
It is helpful to think of it by analogy.  If you were planning to fly from A to B, how long should you allow to be sure of getting there?
 

Fig 1

i.e. Allowed Time = Interval between flights + Flight Time
 which equals 1 + 2 = 3 hours in this example.
If you arrive just in time to catch a flight, it will take you 2 hours, but if you just miss one, it will take 3.  In the absence of a flight timetable, you will need to allow 3 hours to be sure.
It is possible to talk about the response time for intermediate components too. However, it only makes sense if the intermediate is stocked: “planes must land at the airport”
 

Ways of Reducing It
It is clear from the definition that response time is reduced by:
• Making batches more frequently (reducing batch size)
• Reducing process lead time
However, the most common form of response time reduction is by the use of stocking points, in particular, finished goods.  Sales can then be in units (=batch size) of 1 and processing time is typically just the shipping time from distributor to customer.
Of course, buffering against sudden changes in demand by adding raw material, intermediate or finished goods reduces response time at the expense of inventory. So response time reduction should not be viewed as an end in itself. It needs to be done without increasing inventory or cost to be a credible improvement to the business.

Fig 2

A response time can be calculated for each set of processes which end in a buffer stock. The complete supply chain is thus an assembly of such links each with its own response time.
 
Thus there is generally more than one figure to be looked at, although the final one (nearest to the end customer) is the one most often scrutinized.

The Result
A more responsive factory:

• Is less likely to default on customer orders
• Has fewer disruptive changes to production plans
• Is in less of a panic

Fit with Management Strategies
Whilst not stated explicitly, Response Time reduction is the goal of many manufacturing strategies such as “Lean Manufacturing”.  However, there is nothing new about it as a concept.  Henry Ford can be thought of as an early advocate when he established a car production system that smelted 1500 tons of steel a day to turn out a car every 49 seconds in the shortest possible elapsed time.

It is a common sense concept. If you want to satisfy the customer, you must respond to their demands.

By James La Trobe-Bateman, reMODEL Consultants International Ltd

see Case Study for an Illustration of the Real Effect of Reducing Response Time

Why Is Batch Size an Issue?
Batch size is a big issue in healthcare manufacturing because there are some good reasons to make them large and other good reasons to make them small.  It is a potential battleground.  To avoid a war, your organisation needs to understand where it should stand on this issue and agree about your policy. At stake are end user  satisfaction, the ability to maintain supply and operating expense.  Whilst the end user does not care about batch size per se, they are affected by batch size policy.  This means that batch size is often critical to business success.

Why Do Some People Like Big?
Large batches are perceived to mean (although it may not necessarily be true) lower cost, fewer quality control issues, and consistency of product performance.

Why Do Some People Like Small?
The main advocates of small manufacturing batch size are those working with the supply chain.  Here the main issues are being able to supply the customer and shelf life constraints. Finance people are always keen to reduce the amount of capital employed (i.e. inventory) and smaller batches will naturally lead to this. 

Another group strongly in favour of small batches are internal and external productivity improvement consultants who have studied modern manufacturing methods (e.g. the Toyota production system and lean manufacturing).  The perceived benefits are first Timeliness, followed by Inventory reduction, but always with a strong hope of cost reduction, too.  Where product innovation is rapid, then small batches will mean avoiding scrap of product made obsolescent by the introduction of new variants.

What Batch Size Means to Users
End users want to buy what they need and no more, so their preferred batch size is the order quantity. 

Furthermore, in healthcare, consistency of performance is very important and so users often like to know that when they place repeat orders the product will be the same.  In other words, they favour large manufacturing batches, as long as they don’t have to store it all. Users also want material that is “fresh” when they buy it, and are put off by the idea that the material they just bought has been lying around in storage “collecting dust” for a long time. Overall their requirements are not really batch size related at all, even though they may hold strong views about it depending on their experience and the emphasis that they place on delivery or technical performance.

What Batch Size Means to Distributors
Manufacturing batch size will affect the number of different lots that the distributor might hold at one time. Often in healthcare the lot number is important and so holding stocks of different batches adds administration effort and makes allocation of stock to specific orders a bit of a jigsaw puzzle. 

Distributors would thus prefer large batch sizes.

What Batch Size Means to Manufacturing
Manufacturing is where the batch size battles tend to be fought. This is where the advocates of modern manufacturing generally work, it is where the costs are felt, where the delivery performance “buck” stops, where product consistency is measured and controlled, and where product design quality manifests itself in technical support groups responsible for “keeping the show on the road”. All these influences impact batch size choices.

The main drivers can be summarised as:
· Cost drives batch size up
· Product performance drives batch size up
· Capacity drives batch size up
· Delivery performance drives batch size down
· Shelf life drives batch size down

What Batch Size Means to Purchasing
Purchasing are mostly measured on the discount that they can get from suppliers.  Bigger orders command bigger discounts, meaning that purchasing prefer to buy large batches of raw materials.  Difficulties with raw material quality might also drive purchasing to buy as much “good stuff” as they can while they can.

They do not normally influence manufacturing batch size.

What Batch Size Means to Quality Assurance
Quality Assurance, including Regulatory Affairs, would prefer larger batches in order to reduce their workload, because much of it is batch related and because they are often challenged by maintaining product performance batch to batch.  Whilst it is quite possible to have quite significant variation within a batch, it is batch to batch variation that is first noticed by end users.  When this is unacceptable, customers complain and for regulatory reasons QA are then involved.

Who Do You Listen To?

Which way the decision should go, depends on which of these influences is significant and how much they matter. 

For example, if shelf life guarantees and delivery performance are vital to achieve sales, then these must override other concerns, especially when quality concerns are minimal, and capacity is affordable. The bias can be moved by satisfying the business need without resorting to changing batch size. For example:
· Responsiveness can be improved by reducing manufacturing lead time instead of reducing batch size.  Improving delivery performance this way does not demand smaller batch sizes. 

You might have to take a long term view to resolve the conflict. 

For example:
· You can remove shelf life constraints by development program to extend shelf lives. This will allow larger batch sizes. Within a given product line, there may be a range of products, some of which are high volume and some low. 

The batch size policy may well have to be different for each.

by James La Trobe-Bateman, reMODEL Consultants International Ltd