One question that everyone faces in the livestock industry is what is an optimal production run? Are long production runs a more effective way of production and what are the right products and the right time to produce them? These are a few of the questions encountered recently in our webinar on “Big Data”. There was a considerable demand for delving into this subject, which is the reason why we are putting that to the test in this blog, especially since we know from experience that up to 25% additional output can be generated from the factory simply through good planning! Talking about a good ROI.

By Dennis van Lankeren - Head of Sales North America, KSE Process Technology B.V.

Putting It to the Test
Since this subject contains considerable, theory some explanation is required on how the result was achieved. Not all model choices are explained, here since that would be too voluminous. The main principles must be clear in advance, however. Nasty surprises can be avoided with proper and clear needs planning. Daily handling of rush orders and delayed orders can often be prevented through a thorough analysis and by forecasting ordering patterns, thereby reducing the cost of failure for inefficiencies and errors.

When we talk about the costs of failure in this sense, we mean: Needless Transport, Excess Stock, Unnecessary movements and relocations, Waiting, Overproduction, Over Processing (performing too many actions for the same end result) and Defects. This is based on final products to be produced. The supply of raw materials with associated order levels and optimisation thereof is another issue, which is not discussed in this blog since it is worth exploring that on its own.

The Analysis
To put this to the test an analysis was prepared with data from an actual factory, which included all of the articles demanded and production over a period of one year. Orders from the previous 2 weeks were examined each time in order to identify a pattern. The desired stock level was then calculated based on this, along with a border when new production is required. It is assumed that it is undesirable to produce stock products in multiple runs per day. This means that the run size must at least be equal to the demand for a single day for that article.

Step 1: Determining Articles
A Glenday Sieve was first of all prepared of all the articles to be produced. This is a type of Pareto analysis to identify runners within production. It was found that 22 products were “A” articles, while 161 were “B” articles and the remaining were C or D articles. Runners need to be kept in stock, which are the minimum “A” category items. Afterwards we determined the remaining potential capacity of the silos and based on that the “B” products to be kept in stock were determined. It is also possible that some “B” products must be kept in stock for a different reason (different die size, long changeover times, etc.). These must be added manually as stock items.

Below is an example of the analysis with the number of articles in each category compared with the theory. For example, 6% of the articles always accounts for 50% of total sales, which means in this case 22 articles are responsible for 52.05% of sales. This fits in well with the theory.

Step 2: Demand Pattern and Order Level
The demand pattern for potential stock items was then analysed with the space available in the silo park. This was done based on statistics with normal distribution. It was assumed that a minor risk could be taken of stock running out (<5% probability). A calculation was then made of the desired safety stock levels and production levels. Variations in order volumes and order numbers by week/day were also taken into consideration.

In order to accommodate start-up losses with press runs during production it was assumed that only complete batches (6 tons) are to be produced, with a minimum threshold of x batches (x is product dependant for different runners of 4-10 for batches). These numbers were completed based on experience and no underlying model is provided.

Minimum stock levels to be kept were calculated based on these statistics. Order models that are normally reserved for suppliers were used. In order to apply this theory, it was assumed that the factory supplies to the silo (the silo is the end customer). The delivery time is one day because it must be decided 1x per day whether to produce this article or not and the quantity to be produced. The BS model is used with an order level that is determined every day, while the volume to be produced is done in a minimum number of batches + the number of full batches (depending on open orders and demand). This amount is based on the demand and variation in demand over the two previous weeks.

Step 3: Comparing Theory and Practice
A comparison of the actual production of Article X with the theoretical approach for Article X would look as follows:

Reality

 

Theory

Or if we zoom:

Reality                                           Theory

What is immediately noticeable is that the upper graph shows a much more erratic course. The theoretical factory provides much more stable production. In reality stock levels also ran out twice, while this never happened in the theoretical model. The total stock was higher, however, with 52 tons in reality and 85 tons in theory. The question then is whether the silo park could actually hold this volume, which can be verified later in the model. I want to emphasise again that this is actual data. This factory neglected optimisation and lost production as a result!

Step 4: Comparing “A” Products
There are a few interesting values to consider in order to get a true assessment of the advantage or disadvantage of maintaining theoretical production runs and inventory levels. This can be done by comparing the actual and “simulation” versions and then putting that opposite each other in a matrix in order to determine what the added value would have been:

	Practice	Theory
Number of products in the analysis	16 (6 units impossible, no inventory articles)	16, the same as in practice
Number x without stock	80 x	1 x
Total average stock	75,805 kg	70,798 kg
Number of runs required	2451	1820

16 of the total of 22 “A” items were analysed (“B” items were not considered for this blog). The remaining six are not always produced on stock during the year and a real comparison can therefore not be made. It is remarkable in itself that runners are not always available in stock, while the silo park had adequate capacity to this end. This may be due to seasonal influences that were not included or new articles. It would be worth the effort to look into this, however. It is immediately noticeable in the comparison that in the theoretical model the number of times that an inventory item was ordered and where stock levels appeared to be empty is considerably less. This shows that rush production was required in reality or that the item was not sold to the client since stock was not available.

The total production for the required inventory levels was also produced in 631 fewer runs! This does not directly show that there were larger runs for all of the articles, but the balance in the production of these articles was much better. Average stock levels were also slightly lower, which is an indication that the silo park is adequate. Maximum stock levels were not included. In order to perform a complete analysis the difference in stock levels must be considered, along with the number of silos that should be allocated for this purpose. “B” articles can be kept in stock based on the number of silos available subsequently.

Step 5: Conclusions
The theoretical model provides considerable advantages in production. There are fewer start-up runs, fewer urgent requests and waiting time, but a similar total stock level. This is an indication that working with theoretical production quantities would provide added value. Another advantage is that there are more breaks in production time, which simplifies scheduling and which makes it easier to plan for urgent requests. Efficiency and productivity also increase further, simply because employees have time available to analyse instead of being caught up ad-hoc problem solving.

The number of cases was not taken into account in the model and in this analysis. A shortage in raw material was for instance not included, which could be encountered in reality. Articles that change during the year between stock/order products and products with seasonal influences were also not taken into account.

The yields to be gained are also partly dependent on the selected minimum run size. This has a direct relation to the required number of runs required and the total stock levels. If the run is bigger stock rises and the number of runs required is reduced. For purposes of the calculation it was assumed with a certainty level of 95% that it would not be desirable to produce stock products more than once on the same day. The choice of how often a product is to be produced has a great influence on the result of the theoretical model. Much can be gained, however, from optimising and analysing demand patterns. It is simply a matter of doing in order to reap the benefits!