Concept and Implementation of an Integrated Decision Support System (IDSS) for Capital Intensive Farming

P. Wagner and F. Kuhlmann(1)

ABSTRACT

During the evolutionary process of developing software for management tasks the need for integration became more and more obvious. This paper discusses, how integrated information processing can be accomplished to support the managerial functions.

Based on the concepts of control theory principle schemes of comparison possibilities and deviation analysis are shown.

The philosophy behind the design of an Integrated Decision Support System (IDSS), the implementation on a farm and the integration problems of hardware and software are discussed. The applied IDSS consists of several planning and controlling models. These models and the linkages between them are described in detail.

1. INTRODUCTION

The need for more and better information on which to base decisions is not a new problem. However, in recent years this problem has become even more important, particularly for capital intensive farming in industrialized countries.

Information is required for different levels of farm management, reaching from very short term decisions such as applying an insecticide or not to very long term decisions, such as building a hog-barn or not.

In addition, the information needs for capital intensive farming are deviating from those of extensive farming. Capital intensive farming is characterized by high sales volumes in comparison to the generated net value added. E.g. layer hens or feeder pigs where the monetary input is high in relation to the sales volumes in contrast to range cattle, where the monetary input is just a small part of the sales volume. Thus, in capital intensive farming small changes in input-output-coefficients and/or prices can cause the net income to switch from positive to negative. Due to these facts, the inputs and outputs need to be monitored and controlled much closer than in extensive farming. Therefore the information required for capital intensive farming needs to be on higher levels with respect to quality as well as to quantity.

Providing farm managers with better information has been an evolutionary process. These efforts include developing electronic data processing systems, such as linear programming, management information systems (MIS) and currently, decision support systems (DSS). During this development process, the need for integration became more and more obvious:

Because DSS's should have more emphasis on human effectiveness than on machine (computer) efficiency, idealistically the data should be placed in one comprehensive data base which can be accessed by various models, which are placed in a model base. This conceptual design has been proposed by SPRAGUE & WATSON (1983, pp.22). As will be shown later, there are different ways of going from the idealized integrated DSS (IDSS) on the one side and stand alone (independent) programs, on the other.

At this point a principle question arises: If a user is able to use an IDSS and the relevant data are available for processing, he will get the information he is looking for. But how to continue when the information indicates that something is going wrong? What should be done next? How to take influence? How to regain control?

This paper tries to answer those questions, therefore it first deals with the value of information and ways of processing and using it. Hereafter, the role of integration will be discussed and finally the implementation of an existing IDSS on microcomputers on an experimental farm will be presented. The emphasis in this paper is placed on short term controlling. Long term controlling applications will be mentioned, but not be discussed in any depth.

The group of real-world decision makers to which the described IDSS is addressed to has been mentioned already: These are farmers, practicing capital intensive farming. Due to the fact, that the principal planning and controlling needs for capital intensive farming almost always are similar, the type of farm is not really important. That means, that the IDSS can be used by a family owned dairy farm for example as well as by more complex agricultural firms with employees and/or part time workers. This is due to the nature of an IDSS: All of the models or a subset of them can be used independently. However, as a general rule can be stated, that the more intensive and diversified a farm firm produces and the more profit centers a farm firm comprises, the more benefitable the use of an IDSS will be.

2 THE ROLE OF INFORMATION

2.1 Information

The more one knows about alternatives of actions, their likely consequences and restrictions, the more successful one will be in general. In other words: The right information at the right time is the key to success.

The basic problem therefore is the scarcity of information. Information does not exist per se, information has to be produced. Information is obtained from data. The manager typically has a large amount of data, but a limited supply of information.

As shown by CONNOR and VINCENT (CONNOR, VINVENT, cited in HARSH, CONNOR, SCHWAB, 1981, pp.15) information itself can be descriptive, diagnostic, predictive, and prescriptive.

2.2 Production and Distribution of Information in a Firm

In systems theory firms can be defined as open dynamic systems (BAETGE, 1974, pp.11). For the purpose of this paper they can be characterized further by splitting them into a basic subsystem and an information subsystem, as shown in Figure 1 (KUHLMANN, WAGNER, 1986, pp.410). The information subsystem refers to the basic subsystem. Figure 1 shows that the basic and the information subsystem, can both be subdivided in a production and a distribution subsystem. The two distribution subsystems are connecting the production subsystems with the environmental system.

Figure 1: Information subsystem and basic subsystem of a firm

The basic subsystem represents those parts of a firm, where real and nominal goods are transformed. Here the production inputs are purchased, shared and transformed into products. The products are sold. The flow of nominal goods (money) is induced by these processes. In the basic production subsystem the production of real goods takes place. The basic distribution subsystem keeps contact with environmental system: Goods are transformed over distance and time.

The information subsystem behaves in analogy to the basic subsystem. The information distribution subsystem first takes or receives data as informational production factors from the environmental system, the basic production subsystem and/or the basic distribution subsystem. The data then are stored or transmitted to the information production subsystem, where they are processed into information. The information gained is transmitted to the environmental system, e.g. in form of orders, advertisements, etc., or to the basic subsystem, e.g. in form of instructions and results.

The information subsystem of a firm is the place where the DSS is located. The information distribution subsystem holds the database and the information production subsystem holds the model base. Therefore the information subsystem will be the further object of consideration.

3. SUPPORTING MANAGERIAL FUNCTIONS BY INFORMATION PROCESSING

3.1. Management Functions in a Firm

Goal oriented management needs to use the cybernetic concepts of open- and closed-loop control. Prior condition for the use of those concepts is the installation and application of decision support systems. The DSS contains models of the system or models of parts of the system, where the system is the firm to be managed.

However, the concepts of open- and closed loop control shall not be discussed here because this material can be found in other publications, e.g. KUHLMANN, WAGNER, 1986, pp.413 or KUHLMANN, BERG, HARSH, 1984, pp.21.

3.2. The Managerial Process in a Firm

The managerial process may be subdivided into six subprocesses:

  1. definition of goals
  2. planning (observation and analysis)
  3. decision
  4. implementation
  5. controlling
  6. evaluation

Other studies may give other definitions of the managerial process, depending on the theoretical point of view. In order to define the managerial process according to an IDSS, the above definition is sufficient.

The six steps of the managerial process are to be understood as an iterative process rather than a one way sequence, as shown in Figure 2.

Figure 2: Controlling as function of operating a firm

The process starts with the definition of goals. The definition of goals takes major influence on the outputs of the planning process. The plan defines the objectives to be realized by the system, for which the reference vector must be transferred into the control vector. The real system, now acting with the control vector, delivers actual values (output vector) after a certain time lag. Those values then need to be compared under different aspects, like the preset objectives by the means of a pre-post comparison or any other, e.g. external values by a post-post comparison. Probably there will be some deviation or error. The error vector itself provides three different kinds of information:

1. It will tell the controller (the model of the system) that something went wrong. The controller has to decide what to do to bring the real system back on course: The controller must define a new control vector.

2. In the case of fatal errors or long lasting errors of the same direction, the plan should be considered for correction, because defined objectives could have been wrong.

3. The errors must be evaluated to learn what the reasons for the errors were. This may happen periodically and is the case particularly if the actions 1. and 2. have not lead to the desired results. The evaluation may show that the goals have not been well defined. Either the reference vector turned out to be miscalculated or the goals are simply not to realize with the given assets or production capacity.

It has to be said at this point, that the values of the reference vector are future values. They can only be estimated and not predicted with complete certainty. However, there are ways to compute those values with a higher certainty than it has been done in most cases in the past, as will be shown in chapter 5.

In the decision and the implementation steps there are some differences concerning the actor. This depends on whether to optimize the structure or a process of a firm. Table 1 shows this setting.
Phases of Management Action taken by
Structural optimization Process optimization
Definition of goals manager manager
Planning manager/model manager/model
Decision manager/model manager/controller
Implementation manager manager/actor
Controlling manager/controller manager/controller
Evaluation manager/controller manager/controller

Table 1: Actors in the Phases of Management Depending on Structural Optimization or Process Optimization

3.2.1. Structural Optimization versus Process Optimization

The optimization of structures in general is a design problem and has two dimensions. The first one cannot be directly influenced by the manager, for example new varieties of cash crops, which require fewer inputs or have higher yields and other factors which are influenced by technical or biological progress. The second dimension, which Table 1 refers to, can easily be influenced by the manager. For example the structure of a firm, the crop rotation, the variety of dairy cows. These are factors which "have to be lived with", at least for a certain period of time. This does not mean that there is no room for corrections or changes, but the decision on a certain crop rotation for example can only be changed when the period of vegetation is over, for example.

The nature of process optimization is a different one. The process is predefined by the structure, but the actions to be taken in the process may vary and can be adapted if necessary, often very quickly in very short periods of time. Feeding programs for pig fattening may serve as an example.

In a complete IDSS models must be implemented to manage both problems.

Referring to Table 1 the two examples above shall be taken to illustrate by whom actions may be taken in order to optimize structures or processes.

The optimal configuration of a firm for example can be found by a simplex algorithm or just by trial and error (simulation). In any case the manager has to define a goal, e.g. maximize the overall gross margin. The planning phase can be accomplished by the manager himself or by any LP program. In the latter case the result is normative, there is no decision because the information is prescriptive. In the simulation case the manager has to decide what to do. The implementation of a chosen production program is left to the manager. He or a controller (which would be a computer program) may then control the success of the implementation of the plan by comparing the results with the predefined objectives. Finally the obtained results have to be evaluated, either by the manger himself or by a program, designed for this purposes.

In the second example (pig fattening) the goal again must be set by the manager. A possible goal could be increasing each pigs weight by about 700 grams per day. How to reach that? Again, this can be planned by a computer program or by the manager. In any case decisions have to be made about the feed ratio (in the first step), the combination and the amount of ingredients, or changing of a ratio (in the second or any further loop in case of error). The adjustment of ratios, i.e. computing new ratios, can be performed by a controller (model of the system). The implementation can be accomplished either by a feeding computer (actor) or by hand (manager). The controlling as well as the evaluation phase can be depicted as in the first example: The manager or the controller may control the success of the implementation of the plan by comparing the results with the predefined target values of his own farm and/or actual values from successful comparable pig fattening farms. The error values which are needed to adjust the ratios are then generated.

The two facts that

show that process control and optimization can be accomplished much easier automatically than structural optimization.

3.2.2.re-post Comparison versus Post-post Comparison

In the following the concepts of pre-post and post-post comparison as the two major controlling tools of Figure 2 shall be explained in detail. The principle scheme of both is shown below (Table 2 and Table 3).

Table 2: Principal schemes of comparison possibilities: pre-post comparison
SUM JAN FEB MAR APR MAY ... DEC
efficiency ratio 1
target value (cum.) 2400 200 400 600 800 1000 ... 2400
actual value (cum.) 780 180 340 540 760 x ... x
deviation (absolute) 1620 20 60 60 40 x ... x
deviation (relative) -67.5 -10.0 -15.0 -10.0 -5.0 x ... x
efficiency ratio 2

.

.

.

Table 3: Principal schemes of comparison possibilities: post-post comparison
reference values actual values
unsuccessful farms average farms successful farms farm to compare
(abs.) (rel.) (abs.) (100%) (abs.) (rel.) (abs.) (rel.)
efficiency ratio
1 400 80.0 500 100.0 550 110.0 530 106.0
2 1900 95.0 2000 100.0 2200 110.0 2008 100.4
3 360 102.9 350 100.0 325 92.9 350 100.0
4

.

.

.

The pre-post comparison compares the development of several defined efficiency ratios over time. The objectives (target values) originate in the planning process. In the example they are generated in monthly steps at the beginning of the year. The actual values have to be put in whenever they occur. Hence, the comparison shows the absolute and relative errors per month and to what degree the plan is fulfilled. Pre-post comparisons indicate deviations at an early stage, and thus permit early corrective actions. The values may be cumulated over time to get smoother time series.

The post-post comparison is used to provide information about the result of managing a firm in relation to comparable firms, such as farms of the same region and the same kind of production system (horizontal comparison). This type of comparison should explain the reasons for deviations. The corresponding data of other comparable farms (reference values) may be collected by the extension service or by book accounting bureaus. The time horizon of such a comparison may be a year.

These two ways of comparing data are only examples, though there are many other ways of doing this job, e.g. time series analysis of several efficiency ratios of one firm (vertical comparison). This may show the firms development path.

The basic idea of such comparisons is to learn. The learning process starts with recognizing a deviation. It continues by analyzing the causes of the deviations (the evaluation step) and by learning, how to do it better.

3.2.3. Deviation Analysis by Means of Controlling

Basically, deviations between the desired and the actual state are caused by three effects. These are deviations in

In addition, each of these three different sources of deviation may be caused by internal or external factors, that means controllable or non-controllable variables. To bring the system back on course, the reasons of the deviation must be analyzed first. In general, short term adaptation can be accomplished by trying to influence quantities and/or prices, whereas in almost all cases long term adaption stands for manipulating the structure of a firm.

However, when emphasis is placed on short term process control, price and quantity effects must be considered first. Analyzing gross margins may serve as example. A procedure of analysis is shown as a flow chart in Figure 3.

If, as shown in the example of Figure 3, deviations between actual gross margins and target gross margins occur, one has to find out first whether they are caused by prices (or quantities, the order is not important at this point). In order to determine the reasons for price deviations, the actual prices of the considered farm have to be compared with the target prices as well as with the reference prices, which may be obtained from comparable farms. A similar procedure may be necessary to repeat for the quantities.

3.2.4 Ways to Help Entrepreneurs in Decision Making

Figure 3: Analyzing volumes and prices

If the farmer is faced with the problem of taking actions, he needs to decide what to do and what not to do. Therefore he needs the ability to perform his managerial tasks efficiently. There are commonly four methods of improving that ability (OHLMER, NOTT, cited in POLYAKOV, KUHLMANN, OHLMER, 1981, pp.103):

(1) Providing the farmer with information about relevant data (e.g. available facilities and services), about problematic situations and about analysis and planning methods. This kind of help utilizes written material and broadcasting, which is directed to many farmers.

(2) Increasing the farmers knowledge and managerial skills, so that he will be able to perform the management task on his own. This means education about the situation (problem), relevant information, analysis and planning methods, available facilities and available services. Each activity within this kind of help is directed to a group of farmers.

(3) Face to face service, i.e. an extension or commercial agent helps the farmer in doing a part or all of the managerial tasks.

(4) Providing facilities which the farmer can use by himself and then be able to perform the management task. Each of these facilities is used by a single farmer, although the facilities can be mass-produced.

The implementation and usage of an IDSS is grouped under point 4. This does not mean that all problems are solved only by implementation of a system. This can be only one step in the process of improving the farmers ability to perform his managerial task. In order to get some benefits out of an IDSS the farmer needs to be educated and informed. Using an IDSS requires knowledge about the implemented methods and tools, knowledge about the quality of data available and the capability to evaluate the results. On the other hand, it may be easier to perform the managerial task with the support of an IDSS. A modern IDSS can help by selecting and using the appropriate tools and methods for a given problem as well as by interpreting the results and recommending the actions to be taken. This will be discussed in chapter 5, where the general structure of an IDSS will be presented.

4. THE ROLE OF INTEGRATION

Before an implemented IDSS will be described in chapter 5, it shall be discussed, why a DSS needs to be a somehow integrated system and what integration means in the context of a DSS.

Integration of information processing can be seen at least at two basic levels:

The integration of hardware deals with compatibility and is mandatory for integrating software. In regard to hardware integration, most problems are of technical nature, such as the design of interfaces, the kind of handshake or just the compatibility of magnetic tapes or, more important nowadays, disk formats and sizes. Therefore, hardware integration problems arise if computers need to be connected. It isn't to worry about integrating hardware, if all the used software is to run on the same machine and the data are entered into the computer via keyboard. However, this is almost not practicable, as the later example will show, because a substantial portion of the needed data are registered via sensors and stored in devices, which may be called "process control computers". This will become even more important in future capital intensive farming.

The second level of integration mentioned above is software integration. At least three different sublevels are to be considered here, the enumeration follows in ascending order of practicability and ascending level of possible complications:

(1) No direct linkage between programs. The output of one program must be reentered in another program via keyboard. Integration here means matching units, e.g. an output in hectares of one program cannot be used as input in acres in the other program. Aside from this there will be no particular problems.

(2) Indirect linkage between programs. Data between programs can be transferred via specific connection modules (these are programs, too), but at this level there is no common database which would be accessible to the communicating program. The communication needs to be performed by file operations. Problems arise if programs or data structures need to be modified, if, for example, the units do not match as described above, and last but not least if the process of transferring data from one program to another is not automated, the user may simply forget to start the transferring program before doing some analysis with the program which needs to receive the data.

(3) Automatic linkage of all programs. That means, any input is to be entered exactly on time. Every other program requiring the same input will check in a common database if the data were already entered. Outputs are written into the database as well, so that other programs can refer to the results of preceding ones. The problem at this level of software integration is to be seen in defining the data structures and managing the database. Even more complicated is how to "tell" each individual program whether to ask the user for data or to look in the database first, and, if the data were found in the database, to decide whether they meet the needs of the special application the user wants to run. Due to the fact that software is not "static" (this holds true for IDSS, too), new programs may be added, old ones may be dropped or existing ones changed. Some cases may require independent usage of a program. All these things have to be managed by the data base and the model base manager (see chapter 5 for explanation) automatically. It cannot be managed by the IDSS user.

This last and most sophisticated level of integration is the goal to be achieved if an IDSS is to become user friendly and widely accepted. It is not the level of sophistication IDSS meet at present.

5. CONCEPT OF AN IDSS IN AGRICULTURE

5.1. General Structure of an IDSS

Figure 5: General concept of an IDSS. Source: SPARGUE, WATSON, 1983, pp.22 (modified)

The general structure and the components of an IDSS are described in Figure 4. This theoretical approach is presented by SPRAGUE & WATSON (1983, pp.22) and modified here. The major components are the data base, the model base and the user support base. In addition, but not less important, there are management systems for all three bases, a user interface and the decision maker himself. Each of the components will be examined independently.

Database and Database Management System. A database system is used to store classes of data which have been collected for various purposes such as financial data, production data, statistical data, and so forth. The data can be generated from the firm itself or from external sources. Because the sketched IDSS meets the conditions described as the third level of integration in the last chapter, the various databases need to be consistent within the overall structure and need to be shared across functional needs. This means for example, that the accounting data is not stored using a different system than the production or statistical data.

Model Base and Model Base Management System. The model base is closely connected to the database. The model base contains several kinds of models, some of which are used for strategic planning (structural optimization), others for supporting tactical and operational decisions (process optimization). The model base management system performs the same basic tasks as the database management system. It is charged with retrieving the appropriate model needed for a specific purpose and then requesting the necessary data for the model from the database management system and/or the user interface.

User Support Base and User Support Base Management System. The user needs to get information about the data/database and/or models contained in the model base. That information is kept in the user support base, which could be called "information base" as well, but the word "Information" is already used in another sense (see chapter 2.1). The user support base management system e.g. could be a powerful hypertext system. Besides the model base, this is the most important part of an IDSS when it comes to user acceptance. Not every user knows about the possible applications of every model contained in the model base. Of course, there are other ways to help the decision maker here, as depicted in chapter 3.2.4, but in the short run the user support base must contain all necessary information about the usage of models and data of the IDSS.

The relevant models for short term controlling are contained in the operational part of the model base. Therefore, the emphasis on the following description of an implemented IDSS will be concentrated on this part.

5.2 The Concept of the Applied IDSS

Figure 5: Concept of the applied IDSS

The design of the applied IDSS is illustrated in Figure 5. The IDSS has been developed at the Institute of Farm Management at the University of Giessen. It presently is implemented and tested on the experimental farm of the Institute, MARIENBORN.

The models can be subdivided into two major groups: planning models and controlling models. The planning models are generating the reference (target) values, the controlling models are processing the actual values and the pre-post and post-post comparisons, respectively.

In the following the linkage of the models will be discussed first, after that the models themselves.

The Planning Models:

QUANSET generates the reference values for the quantities, PRESET those for the prices and PROPLAN those for the structure of the firm. All three deliver their generated data to CASHPLAN. CASHPLAN stands for cash flow planning and control.

The Controlling Models:

CASHPLAN is both, a planning and a controlling model. The defined efficiency ratios can be compared by means of a pre-post comparison. It provides the reference values for CROPCONTROL, PIGCONTROL and COWCONTROL. The three programs control the crop production, pig fattening and milk production. They give their actual values to FARMDATA, a program which provides an overview over everything that happens on the farm, except financial transactions. Those transactions are handled by CONAC, a book keeping program. CONAC is connected to PRESET and CASHPLAN. PRESET needs past time series of prices for the prognosis of future prices (reference prices), CASHPLAN gets the actual values for the pre-post comparison from CONAC. COPRA, a cost accounting program, gets the reference values from CASHPLAN and the actual values from FARMDATA (volumes) and CONAC (prices). It provides a pre-post comparison for every single production process. USTAT and FARMEXPERT compute post-post comparisons. They compare special efficiency ratios of the farm with those of comparable farms.

5.3. The Physical Linkage of the Models

The above outlined linkages are of logical nature, the physical linkages describe how the data actually are transferred from one model to another. The transfer between the planning programs is via keyboard. That means the outputs of QUANSET, PRESET and CASHPLAN are printed on paper. The figures as input for the connected programs have then to be reentered via keyboard by the user. This seems reasonable, because the target values have to be generated just once a year. The linkage between FARMDATA, COPRA and CONAC as well as USTAT and FARMEXPERT also has to be performed manually. There is no other way, so far. It would be worth thinking about a transfer program, because the amount of data to be plugged in is fairly large.

Figure 6: Information linkages between the farm levels

The most challenging task, however, is the linkage between the programs which require actual data as input, that is to say, the programs which stand in close connection to the real world (the production processes). Gathering and entering all of the data from the production processes by hand would make the programs unacceptable and would be too labor intensive. So, this process has to happen automatically. How the linkage between the process control computers, and the programs CROPCONTROL, PIGCONTROL, COWCONTROL and FARMDATA has been realized is shown in Figure 6.

At the level of the three production processes the control is left to the process control computers. The computers are responsible for feeding the pigs and dairy cows. Furthermore the process control computers are gathering the data via sensors, such as the daily milk production or how much of fertilizer has been used on the fields that day. Some of the data, such as the names of the fields for example, have to be entered via keyboard.

The process control computers are connected via cable to the process management computer. At this level a lot of software and especially hardware integration problems arose during installation.

The bottom-up link between the process management computers on which the three controlling programs are installed and the farm management computer with FARMDATA is via diskettes. This could be replaced by a network as well. The top-down link, that means the connection of CASHPLAN and the controlling programs, is via keyboard, because, as already stated, the reference values for the controlling programs only have to be entered once a year.

Finally, on the farm management computer the files of the control programs are joined by a transfer program which makes the data compatible to FARMDATA.

5.4. Description of the Models Contained in the Model Base

In the following paragraph each of the models contained in the model base of the IDSS will be presented. The widely used programs such as CONAC, a double-entry bookkeeping program, are just mentioned in order to sketch the whole system, but not described in any detail.

5.4.1. Models for Planning Purposes

5.4.1.1. QUANSET

QUANSET is a program to generate quantities. The basic idea is, that for planning purposes one needs detailed information about the amount of inputs for the production processes to be planned. The conventional way is to get those values out of tables or to simply estimate, how much of fertilizer, seeds or working hours for example are needed to produce wheat, barley, sugar beets on one hectare or what kind and amount of input are necessary to produce 400000 liters of milk or 2000 fattening pigs per year.

QUANSET creates such figures for the user. The user has to define a desired production level for the particular process to be planned, e.g. winter wheat. That could be about 50, 70 or 90 decitons per hectare. The program then will come up with a proposal of required inputs. The user now can modify the suggestions according to specific circumstances of his farm or just accept the given values. It is important to note that the proposal includes the points of time, when the inputs have to be applied.

5.4.1.2. PRESET

The counterpart of QUANSET is PRESET. PRESET forecasts prices for inputs and products, based on past time series by means of a method called "Adaptive Filter" (RÖHRIG, 1989, pp.69). Thus, PRESET comes up with a prognosis for the price per unit for the desired product or production factor in monthly steps for about one year in advance in the same fashion as QUANSET. The user may accept the projection or correct the prices, if he feels that something could happen that would influence the prices in a different direction. Either way, the information is of predictive nature (see chapter 2.1). The program also compares the farm prices with the market prices in its statistical part to show deviations, in other words it analyzes price effects in detail (see chapter 3.2.3).

5.4.1.3. PROPLAN

PROPLAN basically is a linear programming model based on gross margins. It generates the reference values for the structure of the farm. That means, how many hectares of wheat, barley or corn should be in the crop rotation, how many dairy cows are most efficient etc.

Joining the quantities of QUANSET, the prices of PRESET and the optimal farm structure obtained from PROPLAN gives a complete preliminary budget for the farm, in other words: The plan for the upcoming production period.

5.4.1.4. CASHPLAN

The necessary combination of structure, quantities and prices is accomplished by the planning part of CASHPLAN. Here the projected prices per unit and the quantities per unit are matched and calculated for a complete production budget by multiplying them with the farm's resources. The budgets can be defined for all production processes of the farm.

The production budget is projected for about one year in advance in monthly steps.

Multiplying prices, quantities and the resources (described by the structure of the farm) on a monthly base yields monetary inflows for sold products and monetary outflows for purchased production factors over time. Thus, the control over liquidity is maintained. At the last step an expected balance sheet and profit-loss account is generated.

The results of CASHPLAN show, what, when, where, how and by whom to do.

The reference vectors are set by this. Thus, CASHPLAN generates prescriptive information (see chapter 2.1).

5.4.2. Models for Cost-performance-control and Pre-post-control

5.4.2.1. CASHPLAN

CASHPLAN delivers a pre-post-comparison of free definable efficiency ratios on farm level in monthly steps. The target values are generated in the planning module of CASHPLAN as described, the pre-post-comparison is computed in the statistical part of it. The actual values for that comparison are the results of FARMDATA, COPRA and CONAC.

Table 4 shows as an example some selected efficiency ratios of a pre-post comparison by CASHPLAN. The figures represent the actual state of the planning period until october. Because the figures are cumulated (see chapter 3.2.2) it is to see at a glance, how much is left to reach the final target at the end of the planning period.

Table 4: Selected figures of a pre-post comparison by CASHPLAN, Marienborn, Planning period: 01.01.89-31.12.89

Month --> Jan. Feb. Mar. Apr. Mai Jun. Jul. Aug. Sep. Oct. Nov. Dec.
Formula 3 label: Gross income of farm
target value (C) 112460 191660 300100 480520 602880 688080 782620 916180 1027015 1118190 1222985 1378210
actual value (C) 93092 192055 265050 409899 516735 649940 757926 887563 968318 1152901 1152901 1152901
deviation (A) -19368 395 -35050 -70621 -86145 -38140 -24694 -28617 -58697 34711 -70084 -225309
deviation (R) 82.78 100.21 88.32 85.30 85.71 94.46 96.84 96.88 94.28 103.10 94.27 83.65
Formula 5 label: direct expenditures
target value (C) 54096 130359 197789 271978 337552 399938 456725 503661 553237 616905 675775 721640
actual value (A) 20427 98559 154141 244271 312690 354105 417039 478046 548320 627610 627610 627610
deviation (A) -33669 -31800 -43648 -27707 -24862 -45833 -39686 -25616 -4917 10705 -48165 -94030
deviation (R) 37.76 75.61 77.93 89.81 92.63 88.54 91.31 94.91 99.11 101.74 92.87 86.97
Formula 6 label: Overhead expenditures
target value (C) 17350 30700 42050 62900 74650 108000 121350 146700 164050 175800 191050 234800
actual value (C) 10279 22688 38892 67473 85976 112564 148201 167462 188642 219153 219153 219153
deviation (A) -7071 -8012 -3158 4573 11326 4564 26851 20762 24592 43353 28103 -15647
deviation (R) 59.24 73.90 92.49 107.27 115.17 104.23 122.13 114.15 114.99 124.66 114.71 93.34
Formula 8 label: Wages
target value (C) 21760 42533 66131 85632 107948 127764 146343 164263 190438 213505 244985 265505
actual value (C) 24385 44166 61593 80053 100663 121821 149330 165018 184965 216177 216177 216177
deviation (A) 2625 1634 -4537 -5579 -7285 -5944 2987 756 -5472 2672 -28808 -49328
deviation (R) 112.06 103.84 93.14 93.49 93.25 95.35 102.04 100.46 97.13 101.25 88.24 81.42

(A) absolute; (C) cumulative; (B) relative

In order to analyze the reasons for registered deviations, many other efficiency ratios have to be considered, which are provided by CASHPLAN as well. Thus, as already mentioned in chapter 3.2.3, each deviation of quantities and prices becomes obvious and the reasons for deviations in more aggregated efficiency ratios and can be retraced from here.

5.4.2.2. The Process Control Programs

All three of the following control programs provide detailed pre-post-comparisons. Each of those comparisons consists of comparison of quantities and costs separately. This enables the user to filter effects caused by quantities and/or prices in case of deviations.

The three controlling programs produce descriptive information (see chapter 2.1).

5.4.2.2.1. CROPCONTROL

CROPCONTROL is a program for managing crop production during the vegetation period, it is an enhanced computerized field record system. CROPCONTROL contains

The actual data as input for CROPCONTROL have their origin directly from the field, the target values are provided by CASHPLAN and may be more specified for the controlling purposes of the program.

5.4.2.2.2. PIGCONTROL

The management of feeding pigs is the objective of PIGCONTROL. With PIGCONTROL the farmer is able to keep track of about every group of pigs from the time they get into the barn until they leave. It includes:

The actual data for each group are stored in the feeding computer and transferred daily to the process management computer. The target values, again are provided by CASHPLAN and may be more clearly specified for the controlling purposes of the program.

5.4.2.2.3 COWCONTROL

This program helps manage milk production. For this purpose each dairy cow is individually registered in the program and the process control computer, which automatically transfers the actual data to COWCONTROL. Each cow wears an identification tag, so that the feeding ratio can be determined with respect to the individual performance of the cow. In the same manner, each cow's daily milk production is stored by the computer. This makes it possible to obtain:

5.4.2.3. FARMDATA

FARMDATA is the aggregation module of the actual values delivered by CROP-PIGCONTROL and COWCONTROL. The values of the single production processes are brought together on the farm level. FARMDATA itself provides the aggregated data for COPRA.

The most important outputs of FARMDATA are

5.4.2.4. COPRA

COPRA as general cost and result accounting contains

5.4.2.5. CONAC

CONAC is a double-entry bookkeeping program. Here the realized "on farm" prices are stored for later use by PRESET.

5.4.3. Post-post-comparison Models

5.4.3.1. USTAT

USTAT and FARMEXPERT are the programs for comparing a considered farm with other farms, but in a very different way. USTAT furthermore enables the user to compare his farm vertically. The program uses free definable efficiency ratios to compare the entire farm, production branches or whatever the user is interested in. Precondition for the horizontal farm comparison are external data at least at the level the user wants to compare his farm with. Thus, the major outputs of USTAT are horizontal and vertical comparisons for the farm.

The external data have to be chosen and entered by the user, the farm data are delivered by CONAC, COPRA and FARMDATA.

5.4.3.2. FARMEXPERT

FARMEXPERT is an expert system. Its only output is a horizontal farm comparison. In contrast to USTAT, the outputs here are not just raw figures. Instead of simply producing statistics, FARMEXPERT analyzes deviations and explains the reasons for the deviations. The program analyzes retrospectively price, quantity and structural effects, as they are mentioned in chapter 3.2.3. Furthermore, it contributes conclusions about the profitability of the farm.

The outputs generated by FARMEXPERT are an example for diagnostic information (see chapter 2.1).

The data of the comparable farms are not to be entered by the user, they are provided by a database according to comparable regions and farm types so, that the farm in question can be compared with a group of similar farms.

The actual data of the farm in question are provided, as in the case of USTAT, by CONAC, COPRA and FARMDATA.

6. CONCLUSION

The presented IDSS is the result of rather comprehensive research activities, carried out within the last three years. This holds true especially with respect to the planning models and the physical integration. Due to that fact, the IDSS depicted is presently being used and tested on just one farm. In general the appropriateness and efficiency of the applied IDSS shows encouraging results. On the other hand it had to be recognized pretty soon, that the amount of data to be entered via keyboard is unacceptable for practicing farmers. Nonetheless some of the models like PROPLAN or CASHPLAN (150 copies sold within 18 months) are well accepted by the farmers as stand alone programs. Obviously, the reasons are:

(1) The two programs represent well known approaches such as linear programming (PROPLAN) and budgeting (CASHPLAN). Therefore, it is possible to the user to reason about the expected benefits beforehand.

(2) PROPLAN and CASHPLAN are to be used for planning purposes. There is no need to run them every day, the user may decide when and how often to use them. This is different from the controlling programs, where data have to be entered almost every day. For those cases, where not even a connection to a process control computer is available, the time for entering the data is not seen as acceptable by most farmers. This is even more of a problem as the final reason of all the efforts is cost accounting, an approach which is not widely spread and accepted by farmers so far.

(3) The interdependences between so many programs as in the above described IDSS are of very complex nature. Thus, the "black box" simply becomes too big for the user to be accepted without any doubts. Therefore many users prefer stand alone programs.

To enhance the acceptability of the IDSS as well as parts of it, at least three things should be done:

(1) Despite of the level of integration already achieved, the need for better linkages between the single programs is obvious. We are working on it.

(2) Apart from the model compilations, the selection of appropriate efficiency ratios for pre-post-comparison and post-post-comparison has to be accomplished carefully to avoid information overkill as well as insufficient information.

(3) The farmers knowledge and managerial skills have to be improved as already mentioned in chapter 3.2.4. This can be done by extension services or better in earlier stages, e.g. vocational training institutions and universities. Therefore we are providing those institutions with our IDSS or parts of it in order to make it available for educational purposes. In this case we fill in the data base with default data, so that the students will enjoy using the programs and approaches behind them, without being trapped by frustration due to the boring job of entering data. In this way it is possible to increase the transparency of existing problems and to show how to solve them by using an IDSS.

Recent discussions with farmers showed, that there is an increasing sensibility and awareness about the upcoming techniques of information processing by means of IDSS's.

REFFERENCES

BAETGE, J., 1974. Betriebswirtschaftliche Systemtheorie. Opladen, 277 pp

HARSH, S.B., CONNOR, L.J. and SCHWAB, G.D., 1981. Managing the Farm Business. Englewood Cliffs, N.J.,384 pp.

KUHLMANN, F. and WAGNER, P., 1986. Zur Nutzung der Informationselektronik. In: Berichte über Landwirtschaft, Band 64 (1986), Heft 3, Hamburg, Berlin, pp.408-440.

KUHLMANN, F., BERG, E. and HARSH, S.B., 1984. On Decision Support Systems Using Adaptive Control Procedures. In: Agricultural Markets and Prices. IVth european congress of agricultural economists, Kiel, F.R. of Germany, September 3-7, 1984, pp.13-34.

LANG, N., 1989. Pigcontrol. In: Gießener Schriften zur Agrar- und Ernährungswirtschaft, Heft 18. Frankfurt. 251 pp.

MÜLLER, H. and KÜBLER, H., 1989. Computergestützte Steuerung und Kontrolle in der Milchviehhaltung. In: Ergebnisse landwirt-schaftlicher Forschung an der Justus-Liebig-Universität Gießen, Heft XIX, Gießen, pp.105-110.

POLYAKOV, M., KUHLMANN, F. and OHLMER, B., 1981. Computerization of farm management decision aids. In: JOHNSON, G. and MAUNDER, A. (Editors). Rural Change, The Challange for Agricultural Economists. Proceedings of the Seventeenth International Conference of Agricultural Economists, 3rd-12th September 1979 at Banff, Canada, pp.102-112.

RÖHRIG, C., 1989. Preset. Ein computergestütztes Modell zur Dokumentation, Prognose und Kontrolle von Faktor- und Produktpreisen. Dissertation, Gießen. 216 pp.

SECK, M., 1988. Zur computergestützten Führung in der Pflanzenproduktion. In: Gießener Schriften zur Agrar- und Ernährungs-wirtschaft, Heft 16. Frankfurt. 183 pp.

SPRAGUE, R.H. and WATSON, H.J., 1983. Bit by Bit: Towards Decision Support Systems. In: HOUSE, W.C. (Editor). Decision Support Systems. New York, pp.15-32.

WAGNER, H., 1983. Computergestützte Ist-Kosten-Leistungsrechnung für landwirtschaftliche Betriebe. Dissertation, Gießen, 273 pp.

WAGNER, P. and LANGENBRUCH, F., 1987. Die Anwendung eines Datenbanksystemes im landwirtschaftlichen Betrieb. In: Berichte über Landwirtschaft, Band 65 (1987), Heft 1, Hamburg, Berlin, pp.140-168.

(1) Institute of Farm Management, Justus-Liebig-University, Senckenbergstraße 3, 6300 Gießen, FRG