The Formulation of Optimal Diets for Poultry

Rick Kleyn, SPESFEED (Pty) Ltd, South Africa

Abstract

Formulating poultry diets using Linear Programming (LP) software is a well-established methodology. This mathematical technique simultaneously considers the specifications of the diet, and the available feed ingredients. It does this such that the cost of the diet is minimised. However, the objective of commercial animal production is to maximise profits, and LP does not do very well in this regard. It fails to take into consideration the type of diet (diet specification) that should be fed at any time during the production cycle, or for how long a diet should be fed or a production process be allowed to be continued. Lastly, it does not consider what economic strategies should be used to maximise the profitability of the feeding operation. This paper will give a broad overview of formulation of optimal diets for poultry. In order to achieve this goal, it is important to have a very clear idea of how to measure profit in the poultry industry. It will examine the manner in which both economic forces and time will impact on the profitability of a production system. Most importantly, it will examine those aspects of protein and energy nutrition that lead to the maximisation of profit. In short, it will highlight those aspects of feed formulation that go beyond the simple provision of ‘least cost’ diets of a fixed specification, and will discuss methods that can be used to maximise the profitability of broiler production.

Introduction

Feed formulation is the means by which nutritionists apply their nutritional knowledge in practice. Some authors have gone so far as to call effective diet formulation an “art form”, but this is an overstatement as most of the scientific information required for cost effective feed formulation is currently available to us. In essence, feed formulation consists of three components: the animal itself, the cost and availability of the ingredients that make up the diet, and a relatively simple mathematical technique called linear programming (LP) which enables us to evaluate all parameters simultaneously. In Figure 1, this set of interacting factors is illustrated. LP allows for the optimal allocation of scarce (therefore costly) resources and was originally developed by George Danzig (1952) to optimise the logistic function of the US military. Nutritionists were quick to realise the power of LP and in 1967 Dent and Casey published their book entitled Linear Programming and Animal Nutrition. More recently, Pesti and Miller (1992) have published a book entitled Animal Feed Formulation: Economic and Computer Applications.  Most modern textbooks on poultry science and poultry nutrition give the topic of feed formulations scant attention (Leeson and Summers, 2001; Bell and Weaver, 2002).

From the perspective of both nutritionists and the producers, computers and computer software programmes have had a major impact on the field of animal nutrition. This is mainly because the magnitude of the calculations required effectively precluded the use of much of the available information prior to their introduction. As our understanding of nutrition improves, for example the realisation that amino acid digestibility may be used to improve the standard of feed formulation, the complexity of these calculations increases. Not only are we now able to formulate diets considering hundreds of nutrients and /or ingredients simultaneously, we are also able to examine the effects of an ingredient over a production site or even a whole company in a matter of seconds (Format International, 2004).

There are, however, a number of questions that LP feed formulation software simply cannot answer. These all have to do with the profitability of the animal feeding system. This paper will deal with some of these aspects. .

Measuring Profit

The principle objective of any animal production enterprise is to make a profit. Having said this, it is essential that we have a very clear idea of what constitutes a profit, and how this should be measured. Many producers measure the technical efficiency of their operations, be it via average hen housed production, Production Efficiency Factor's (PEF), feed conversion ratios or any other of a number of different measures. Emerson (2000) makes the point that prior to 1995, at which time industry wide cost data became available through the implementation of the Agristats service, US broiler integrators also measured performance characteristics. After 1995, they moved to cost-driven analysis of results. This, in itself, has shortcomings and he correctly points out that in the future, all producers will have to measure performance in terms of returns (profit) rather than in terms of performance or costs.

In order to measure profitability, we need to evaluate our performance in terms of basic production economic principles. As previously mentioned, the measurement of costs alone has some shortcomings. The measurement of profit is, however, not complicated. It is simply the difference between the return from the sale of any product, minus the fixed and variable costs of the enterprise. Variable costs would include items such as the feed, and the day old chick or point of lay pullet. It is essential that in the calculation of a fixed cost for an operation, all costs and not just those associated with the poultry site be considered.  For example, administrative overheads should be correctly allocated to each house or flock.

An aspect of measuring profitability which is often forgotten, is that of time. We borrow capital on the basis of time (interest per annum) and pay tax on the same basis, so it stands to reason that we also need to measure our profit in terms of per unit time. Where capacity is unlimited and the production process is not time-dependant, emphasis should be placed on maximising profit per animal produced. In situations where capacity is limiting and time is at a premium, the emphasis would then be on maximising profitability per unit time. In essence, capacity is limiting in nearly all poultry operations.

It is well and good to keep a flock of broilers or laying hens for an extended cycle length as your return per bird will, in all likelihood, increase. However, it needs to be borne in mind that you may well be losing the opportunity of replacing this flock with a younger, more efficient one. The impact that time has on a broiler operation can clearly be seen in table 1.  Note that you will gain 1.5 crops per year by slaughtering at 38 days of age with a 7-day cleanout period, as opposed to slaughtering at 42 days with a 14-day cleanout. This is a little unrealistic, as in practical terms one should work in weeks. All the same, a 38-day cycle length with an 11-day cleanout would lead to an extra crop per year.

Table 1: The effect of length of grow-out period and down time on the number of broiler cycles per year

In order to express profitability, we need to express our earnings in a manner that is meaningful and comparable. It may well be possible to express profit on a per bird basis, but it could be argued that it is the poultry house that is in fact the unit of production and not the birds housed in it. For ease of comparison between houses of different sizes, it is proposed that all profit should be reported on a return per unit of house space. 

From what has been discussed thus far, it is clear that our objective in broiler production is to measure profit per production unit (in this case per square meter of house space) per unit time. It is proposed that the measure be defined as Unit Profitability or UP. UP can be described by a simple equation, which is shown below:

UP = [(Income from birds/m²)-(All Costs/m²)]/Cycle length

The application of UP can be as far-reaching or as simplistic as the user wishes. For example, it is possible to use the income from the sale of live birds at the farm gate as input. Alternatively, and more correctly, it would be possible to realise the net realisation of the birds sold to the supermarket, which would take the cost and value added by any processing, into consideration.

The most important economic driver on any poultry production system is the feed price ratio, which refers to the kilograms of feed that can be bought with the proceeds of the sale of 1 kg of product (meat or eggs). Not only does this ratio broadly indicate the potential for operational profitability in the poultry industry but it also has a major impact on the manner in which we ought to manage our flocks. This will be demonstrated by an example.

Nutrient Requirements 

The first major issue this paper will deal with, will be the determination of optimal feed specifications. LP generates a ‘least cost’ diet for a certain pre-determined feed specification.  Specifications are derived through an initial determination of the birds’ nutrient requirements. Expected nutrient requirements, recommended allowances and ultimately feed specifications are available from many sources. These include the sets of tables published by in the American NRC (1994). Increasingly, commercial companies such as Adessio (2003) are also publishing values. The publication of tables of nutrient requirements is perhaps an unfortunate trend, as tables (although derived by factorial methods) fail to present working and logical sequences of quantified steps that allow for a practicable scheme for deriving circumstance-unique allowances for farm animals (Whittemore, 1983). Very simply, the question that needs to be asked is not “what are the requirements?” but rather, “what is the target response needed for maximum returns?” Also, how is that response satisfied in terms of nutrients? Or put another way, what recommendation as to the daily allowance of nutrients should be made to achieve this target?

Before entering into any discussion on the determination of the nutrient requirements for poultry, it needs to be remembered that in the case of poultry we are feeding a population of birds and not simply an average individual. Birds will respond to the input of a nutrient, whether it is balanced protein or a single amino acid, in a typical manner. Each bird will respond such that its genetic potential is met, at which point no further production will be realised. The response shown by a flock on the other hand will generally show a smooth, diminishing response to increasing input. This response has been classically described by a model known as the Reading Model (Fisher et al., 1973). By comparing the marginal value of the eggs produced with the marginal cost of the amino acid, it is possible to determine an optimal amino acid allocation for the amino acid under consideration, for a particular circumstance. Although highly specific, this publication is of importance because it represents a watershed in the way in which nutritionists approached feed formulation. In order to apply this methodology, production functions that predict the response to a nutrient (performance) need to be generated.

The first example that will be dealt with here is the manner in which broiler chickens will respond to protein. Broadly, birds will respond when given more protein or amino acid by increasing the growth of body protein, reducing the growth of body fat and reducing feed intake. In production terms, we expect a heavier and leaner bird, with an improved food conversion ratio. Meat yield, which strongly correlates with body protein growth, will increase as a proportion of body weight (Anon, 2000). Koch et al. (2002) illustrate this very clearly (figure 3).  By feeding graded levels of an Ideal Protein to Ross broilers, they were able to measure the various performance criteria during both the grower and finisher phases. The Ideal Protein concept suggests that the ratios between the essential amino acids and lysine (reference amino acid) remain constant while the quantitative amino acid requirement is affected by many factors, including genotype. It was concluded that current broiler breeds have a high performance potential that can be optimised, according to current knowledge.

Using the data, and by carrying out repeated runs using an LP feed formulation package, it is possible to calculate the level of lysine in a grower diet for male broilers that results in maximum return. This is illustrated in figure 4.

Y = 1202 + 302(1-e^{-.141(Lys-9.1}))                             y= 1.871 – 0.387(1-e^{-.214(Lys-9.1)})

Figure 3: Effects of increasing amino acid levels that were balanced according to the Ideal Protein concept, on weight gain and the feed conversion ratio in male broilers aged 14 to 34 days (Koch et al., 2002)

The determination of the energy level of poultry diets is perhaps the most important decision that has to be made by the nutritionist. Energy contributes approximately 60 to 70% of the cost of a broiler diet, making the selection of an energy level that will maximise profit all-important. It is widely accepted that nutrient requirements should be expressed in terms of grams of nutrient per unit of energy contained in the diet. By deriving functions of broiler response to energy density, it is possible to determine the optimum energy level of a diet. Fisher and Wilson (1974) summarised a large number of trials, but it is fair to say that the results may no longer be valid, as the genotypes of the birds a quarter of a century ago may no longer be relevant to current conditions. Fortunately, Saleh et al. (2004) and Guevara (2004) have both studied the effects of nutrient density on the modern broiler. Guevara fitted a simple polynomial model to his data, which could be compared to a model that was fitted for the data of Saleh et al. (2004). This is illustrated in figure 5 below.

Figure 5: Response in body weight gain in male broilers to incremental levels of nutrient density, after Saleh et al. (2004)

There is reasonable agreement between the two data sets, and when the models were used to predict broiler weight response using the two models, a correlation coefficient of 0.88 was calculated for the predictions, as can be seen from table 2.

Table 2: Predicted body weight response using the models of Saleh et al., (2004) and Guevara (2004)

Whilst these data are useful, care should be taken by any practitioner who wishes to use them.  Both data sets were determined using very low stocking densities. This effectively means that the data may not apply to commercial conditions, as under these circumstances birds are able to respond adequately to low-density diets. Under commercial conditions, birds are often not able to achieve adequate feed intakes. It is suggested that where stocking densities are higher, as is the case in many countries, the expected growth on lower density diets may well be over-estimated. It is of interest that Saleh et al. (2004) reported that there was no increase in mortality or leg disorders when feeding high-density diets. Abdominal fat was not adversely affected by increasing nutrient density when protein was maintained in ratio to energy. Breast meat yield and percentage remained constant as the nutrient density changed. As was the case regarding the response to protein, it was possible to calculate the return at different energy densities using the data at Saleh et al. (2004), and this is shown in figure 6.  

Figure 6: The profit per broiler at differing energy densities using the data of Saleh et al., 2004.

Phase Feeding  

The decision as to which diet should be fed during each stage of the production cycle cannot be made using traditional feed formulation methods. Yet, the choice of diet at any particular stage in the production process can have an important impact on the overall profitability of a broiler operation, both in terms of input costs and technical efficiency. Figure 7 below illustrates how as birds grow, so their energy requirement increases relative to their protein (specifically lysine in this case) requirement. Feeding different diets leads to over or underfeeding of these two critical and expensive components of the diet.

Emmert & Warren (2000) compared the NRC (1994) recommendation to three diets formulated on an Ideal Protein basis to meet the weekly requirements for a flock of broilers. The specifications used in these diets are summarised in table 3 and the results are shown in table 4. Both of these tables are below.

Table 3: Summary of specification of diets used in experiment (Emmert & Warren, 2000)

Table 4: Results of Experiment 0 to 21 days (Emmert & Warren, 2000)

As can be seen, phase feeding had an impact on not only FCR (which was not significant), but also on lysine utilisation. Although not shown, any reduction in lysine usage would ultimately lead to a reduction in cost. Further experiments conducted on birds of different ages showed improved performance in addition to improved amino acid utilisation (Pope et al., 2004).

Table 5 shows what the impact of changing not only the number of phases, but also the manner in which the different phases are offered to the birds. The simple expedient of feeding a Withdrawal Diet (no medication or premix) for the last 4 or 5 days of the cycle would lead to a further saving of 6 cents per bird. From table 5 it can be seen that the total nutrient allocation to each bird remained effectively the same. It is important to point out that in this worked example, it is assumed that the growth and feed conversion ratio remain the same. The work of Emmert et al. (2000) illustrates that it is likely that FCR will improve and practical experience has shown us that body weights generally improve when more phases are fed. 

These data illustrate how both the variable costs and the technical efficiency of a broiler flock can be improved by simply managing a more effective phase feeding system.

Figure 8: The relationship of increasing levels of balanced dietary amino acids and different grow out phases on weight gain and feed conversion in male broilers 1 to 37 days of age (After Koch, et al., 2002).

Some Case Studies

In order to demonstrate just how important it is to make the correct decisions regarding any changes to either the cost structure, the level of technical efficiency or time, two exercises were carried out. The first example was based on the results of an experiment carried out at the Ross research farm in South Africa (Kleyn, 1999). In this trial, three different dietary regimes were used (high-energy diets, medium energy diets and low energy diets) as shown in table 6.  The results that were achieved are shown in table 7.

Table 6: Energy level and cost of diets

From table 7 it can be seen that there was a 4.1% difference in body weight between the HE and LE fed birds. The HE diet out-performed the LE diets by about 16% in terms of PEF. When looking at costs, a different picture emerges with the HE diet being only 2.1% cheaper than the LE diet. Interestingly, the LE diet was a cheaper way of producing broiler meat than the ME diet. If one looks at the return per square meter of chicken house, it can be seen that the difference in return between the HE and LE diets is in the order of 8%. As the cycle length was identical for all treatments, there is little or no effective difference achieved through the introduction of the time factor.

Table 7: Technical performance and financial return (Rand) for diets containing three energy levels at 42 days of age (year 2000 prices)

The results in table 7 were calculated using the feed costs at the time and a selling price of chicken of R 7.00 per kg. What, then, is the impact of the current feed price (assuming that chicken is sold for R 9.00 per kg)?

Table 8: Financial return (Rand) for diets containing three energy levels at 42 days of age (year 2004 prices)

UP is a more important tool in the measurement of broiler profitability when the feed price ratio is low. PEF and the feed cost per kg of chicken are then inadequate.

In the second exercise, an attempt has been made to show what effect a change in technical efficiency will have on profitability. Using data published by North and Bell (1990), it is possible to see what effect increasing the stocking density has on broiler performance. 

Table 9: The effect on changing stocking density on profitability of a broiler operation

These data illustrate a number of important points. Firstly, if the feed price ratio drops the only way to remain profitable is to improve technical efficiency, either by reducing stocking density or improving management in some other way. When the feed price ratio improves, the decision as to the correct strategy, changes. If one were to look at the PEF for guidance, or even the profit per bird, we would be inclined to use a lower stocking density.  However, if we were to use UP, it is clear that the higher stocking density would result in almost twice the return.

Practical Solutions

In order to optimise the feeding process we need to go beyond the formulation of a single ‘least cost’ diet.  By now it should be clear that the feed formulation problem is multi-dimensional. The first of these dimensions would be the ingredient availability, quality and price, which are adequately dealt with by traditional LP.  A second dimension would be the way in which birds will respond to protein, while a third could be the manner in which birds respond to increasing nutrient densities. A range of genetic, commercial and environmental issues will determine the manner in which a flock of birds will respond to these critical components.  The final dimension that needs to be dealt with is time:, both the number and time each feed (phase) should be fed.

It is possible to determine an optimal nutrient density for a diet making use of repeated iterations of a standard LP feed formulation programme.  Kleyn and Gous (1988) proposed a mixed integer programming method, by which non-linear functions such as a response curve can be fitted into a standard feed formulation model in such a way as to maximise profit.  Guevara (2004) utilised the response data that was generated to develop a non-linear programming model which was able to optimise the performance response to energy density in broiler feed formulation. This model determines the optimum energy density for a specific set of financial circumstances and represents a valuable tool for nutritionists to use when formulating diets. However, both these methodologies are only able to consider a single phase at a time, which is a severe limitation.  Roush et al., (2004), have proposed that mixture designs and models could be used to study phase feeding of broilers.  These mixture models can be used to find the balance of the provision of diets that will provide optimal performance and processing performance.  This methodology is limited by the fact that the number of phases needs to be predetermined and also that it does not consider different nutrient densities for the various phases to be fed.

A way by which the shortfall of the methods discussed above can be overcome, is to make use of simulation models. Simulation models take into account the three major stimuli operating on a system. These are:

Simulation models - which are able to predict bird performance under a certain set of environmental and management conditions- are combined with a feed formulation module and what is known as an optimiser module, which allows for semi-intelligent selection of the next step to be carried out  (Gous, 2004).

 Conclusion

It should be clear by now that there are no hard and fast rules for maximising profit in the poultry industry. Changes in production system, bird genotype, the feed price ratio and many other factors will impact on the strategy that producers and nutritionists should follow in order to continue to make a profit. Before any formulation strategy can be decided upon, it is essential that profitability is measured correctly, and it is proposed that the UP system is the best way of doing this. The PEF system gives little or no guidance in terms of profitability.  From a nutritional perspective, it is clear that higher density (more expensive) diets tend towards higher profitability on the farm. These diets may not always lead to the lowest costs per kg of chicken produced and nutritionists and producers need to look beyond costs and rather measure profit. It is important that we consider the whole poultry production process, and not just the on-farm performance and farm gate prices when measuring profit.

From the examples shown above, it is clear that the only real option that the producer has to improve profits is to improve technical efficiency. It is also clear that as the feed price ratio decreases, so technical efficiency becomes even more important. There are numerous ways of improving technical efficiency. These often include more attention to detail in terms of basic management such as house cleanout, effective vaccination, improved chick quality and litter management. They would also include capital items such as nipple drinkers rather than bell drinkers, and pan feeders rather than tubes or chain feeders.

When margins are tight, it is tempting to try to improve these by saving on costs achieved through measures such as the use of lower density diets, which can, and often do lead to a drop in technical efficiency and a reduction in profit. By applying relatively simple techniques, such as the implementation of the correct phase feeding system or diets that have been optimised to contain the correct levels of protein and energy for each phase, a significant improvement in UP can be achieved.

It is hoped that this article has shown just how important it is that each and every producer does accurate calculations and correct calculations i.e. a calculation of UP. In other words, always measure the return per unit of floor space per unit time.

The nutritionist is ultimately responsible for determining which calculations the computer does by determining the values used in the ingredient matrix, as well as which feed specifications should be used.  This is not a simple task, as the ideal solution will vary with the prevailing economic circumstances, the management system used on the farm, the commercial  objective of the client and the genotype of bird being fed.

References 

Anon, (2002). Broilers: Protein and profit.  Rosstech 00/39 Aviagen Scotland.

Anon, (2003). Rhodimet nutrition guide. Adisseo, Anthony France.

Bell, D.D. and W.D. Weaver, (2002).  Commercial chicken meat and egg production. Kluwer Academic Publishers.

Dent, J.B., and Casey, H.,  (1967). Linear programming and animal nutrition. Crosby Lockwood London.

Eits, R.M., Kwakkel, R.P., Verstegen, M.W.A. and Emmans, G.C., (2003).  Responses of broiler chickens to dietary protein: effects on early life protein nutrition on later responses. Br. J. Poult. Sci. 44: 398-409.

Emmerson, D., (2000). Evolving performance in the US broiler and turkey market. WPSA Conference Montreal.

Fisher, C. M., Morris, T. R. and Jennings, R.C., (1973). A model for the description and prediction of the response of laying hens to amino acid intake.  British Poultry Science 14: 469-484.

Format International, (2004).  www.formatinternational.com

Gous, R. M., (2004). Optimising the feeding programme of growing pigs. Proceeding Second Joint Congress of the Grassland Society of Southern Africa and the South African Society of Animal Science.  Page 76 (Abstract).

Guevara, V. R., (2004). Use of non-linear programming to optimise performance response to energy density in broiler feed formulation. Poultry Science 83: 147-151.

Kleyn, F. J., (1999). Unpublished research data.

Kleyn, F. J. and Gous, R.M., (1988). A mathematical model for the formulation of optimal amino acid and energy concentrations in feeds for laying hens. Agricultural Systems 26: 65-76.

Koch, F., Wijtten, P. J. A, Lemme, A. and Langhout, D. J., (2002). Impact of a balanced amino acid profile on broiler performance. Veterinarija Zootechnika 19: (41).

Leeson, S. and Summers, J.D., (2001).  Scott’s nutrition of the chicken. University Books, Ontario, Canada.

North, M. O. and Bell, D. B., (1990). Commercial chicken production manual. Forth Edition. Van Nostrand Reinhold New York.

NRC, (1994). The nutrient requirements of poultry, 9^th revised edition. National Academy Press.

Pesti, G. M. and Miller, B. R., (1992). Animal feed formulation: Economics and computer applications. Van Nostrand Reinhold Company.

Pope, T., Loupe, L. N., Pillai, P. B. and Emmert, J. L., (2004).  Growth performance and nitrogen excretion of broilers using a phase-feeding approach from twenty-one to sixty-three days of age. Poultry Science 83: 676-682.

Roush, W.B., Boykin, B., and Branton, S.L., (2004).  Optimization of phase feeding of Starter, Grower and Finisher diets for male broiler breeder by mixture experimental design: Forty-eight-day production period. Poult. Sci. 83: 1264-1275.

Saleh, E. A., Watkins, E. A., Waldroup, A. L. and Waldroup, P. W., (2004). Effects of nutrient density on performance and carcass quality of male broilers grown for further processing. International J Poultry Science 3: 1-10.

Warren, W. A. and Emmert, J. L., (2000). Efficiency of phase-feeding in supporting growth performance of broiler chicks during the starter and finisher phases. Poultry Science 79: 764-770.

Whittemore, C. T., (1983).  Development of recommended energy and protein allowances for growing pigs.  Agricultural Systems 11: 159-186.