The relationship between the Production Research Department of Guinness in Park Royal and and Statistics Department, run by Stella Cunliffe, was somewhat problematic. The key question was getting recognition, in the dynamic context of process control, for the 'Philosophy of the Unplanned Experiment' in a context dominated by the classical statistical philosophy of 'experimental design' as it had emerged in a static agricultural 'experimental design' situation.

There was a rotary filter with which the 'wort' (ie unfermented soluble malt extract) was drawn off from the mash. The mash was then 'sparged' with hot water, to rinse out the last of the wort. If you sparge too much, the resulting wort ends up too dilute. The wort, after boiling with hops, had to end up at a specific gravity of 1.0459, so that measurement of 'gravity' was crucial. In the traditional process one does this with a hydrometer, a relatively simple task, but the object of the development was to make the process continuous, so this meant on-line instrumentation and a control system. My task was to try to get on top of this instrumentation problem. The wort was then boiled with hops; this was done in small batches in a sequence of vessels in rotation. An attempt was made to compensate for the variability of the hops by measuring a key component of the extract, a biochemical known as 'iso-humulone', and a member of the team, George Philpotts, had attempted to develop some instrumentation for this purpose, using auto-analyser technology.

The hopped wort, at the desired gravity, was then declared to the Excise in a batch vessel, from which the continuous fermenter vessels drew. I spent most of my time at this end of the process, though some of the instrumentation was relevant upstream, and we ran a sort of test-bed for gravity-related instrumentation of all sorts in the neighbourhood of the rotary filter.

We developed instrumentation to measure the oxygen level in the growth vessel, and to measure the gravity in the fermenters, and to keep track of the amount of yeast that came out. The rule was, everything must go downstream; no recycling of yeast. This latter feature dominated the Coutts and Lebatt patents, but Guinness wanted to maintain their own system distinct from this, the key idea being to start with pure yeast culture under laboratory control, and not allow a population of wild or mutant yeasts to build up, as would happen under a recycling regime.

We did the best we could, but the system was wildly unstable, and would have required a sophisticated feed-forward control system, based on totally reliable instrumentation. There is a non-linear relationship between the gravity of the fermenting beer and the extent to which the yeast flocculates. The attempt to control yeast concentration in such a way as to achieve the target beer gravity at the outlet, while staying within the 'no recycling' constraint, was doomed to failure.

In this situation, Stella Cunliffe in the Statistics Department was trying to impose on us a pattern of meticulously planned experiments, derived from the procedures used in agriculture for evaluating barley and hop production. This however was wildly out of phase with the requirements; it was all we could do to keep the process going, with the resulting beer fit to add to the 'added beers' (after all, excise had been paid, and to dump it would have been a mortal sin!). So what we did was treat the plant as an ongoing 'unplanned experiment', keeping records of the parameters at the various stages of the process, and working over the data afterwards to try to draw some conclusions. We were able to do this, and we wrote reports, the results being meaningful experience, and the general message being that continuous fermentation without yeast recycling is intrinsically unstable and Guinness should not pursue this road, a useful negative result. This experience of an 'unplanned experiment' in a continuous process however was an early example of the 'operations research' approach to industrial process control statistics.

BIOGRAPHICAL NOTE:
Dr RHW Johnston

Dr. Johnston graduated from TCD in 1951 in physics and mathematics He worked for 2 years at the Ecole Polytechnique, Paris, and then at the Institute of Advanced Studies Dublin until 1960, on basic research in experimental physics. From 1960 to 1963 he worked with Guinness (Park Royal) on process development. Since 1963 he has worked with Aer Lingus applying mathematical techniques to decision model building, mainly related to long-term investments in the Economic Planning Department.

He is a council member of the Operational Research Society of Ireland, has lectured in the Graduate School of Business Studies, TCD and has published in the international operational research literature.

THE COMPUTER ORGANISATION WITHIN THE FIRM

INTRODUCTION

The installation of a computer represents a major technological step for any company.

No one is surprised when they discover that the transition from workshop to assembly line in mechanical engineering involves structural changes in the company. Yet, too often the computer makes its first appearance in the company under the Chief Accountant and its main job is to do the payroll. Extra functions are added from time to time, and the computerisation of the firm proceeds without strategy, co-ordination or principle. The parallel between the organisation and the human organism suggests an analogy between this type of computerisation and growth which has taken place outside the control of the central nervous systems despite the fact that evaluation of information is the main function of the latter.

If on the other hand, computerisation is to be an extension of the basic organism and is to contribute to it by speeding up and refining its information network, it must do so under the control of the central nervous system. In other words, top management must get involved, to the extent of taking a hard look at the goals of the company, the company structure, the departmental goals and the indicators which are used to indicate success or failure.

It is then necessary to set up the computer in the organisation in such a way as its influence can consciously be channelled into the problem areas according to a strategy which is dictated by the overall goals of top management.

THE COMPANY AS AN ADAPTIVE GOAL-SEEKING SYSTEM

I refrain from giving this; basically it is a representation of different types of information flowing between various levels: policy, planning, control, operations, and the outside environment.

All firms possess a basic structure like this, whether implicitly or explicitly.

The information system absorbs raw data and feeds out information at the level of abstraction necessary for the various functions. Prior to the computer, the information system was not necessarily an explicitly defined function, except in those firms whose process or environmental variability problems had dictated the setting up of a specialist statistical department.

I had here in mind the Guinness experience

The coming of the computer has introduced a qualitatively new element into the technology of information and has, therefore, increased the pressure for explicit recognition of information-processing as a specialist function related to the needs of the company as a whole.

FUNCTIONAL ANALYSIS

This can be done on a two-dimensional chart with axes:

• Decision time,
• Degree of abstraction from production technology.

Note that because this is a growth industry there is a natural division into a planning group and a operational group, to the right and left of the 1 year ordinate. The Planning Group reflects all company functions and lives in the imagined world of the future, basing itself on the real world of the past as reflected in statistics abstracted from the operational group and the environment, by means of the information system.

Again I won't attempt to reproduce Fig 2, but the X-axis is a log scale, with points defined as day, week, month, year, 3-years, 10-years. The Y-axis is labelled 'degree of abstraction from the production process'. In the day-week column we have processes like operations and sales. In the week-month column we have processes like maintenance, schedule control, sales control, promotion. In the 1 to 3 year column we have fleet planning, capacity planning, manpower planning, route profitability analysis, market research. In the above description, named processes are regarded has having a degree of 'abstraction from the production process' which increases from left to right, as expressed by their y-axis co-ordinates.

The Information System

Alas Fig 3 is missing from the paper, but the separate areas were related to the day-to-day sales function (the reservations system), and to the traditional data-processing areas like payroll, routine accounting and stock control.

The main reason for the abuse of the computer which has occurred sometimes... is failure of top management to guide the process firmly from the start.

If the process is left to be defined by the lower administrative levels, the need to provide for the link with the planning functions is likely to be ignored. They will plan for reports suitable to be produced to their short-term needs and will ignore the need to abstract and make available planning statistics in computer-accessible form.

This has very often happened and is responsible for the frightening percentage of computer installations reputed to be "failures".

It is necessary to build a system in which the accounting and statistical functions are clearly distinguished. In other words the design of the information system should be used as an occasion of introducing standard costing procedures.

Thus the accounting element of the system should be used as the source of a set of unit-costs and unit-revenues based on the unit-event of the business. These change slowly and form a link with the planning function. This link has to be manually constructed if it is not designed into the information system. To do the latter is easy, but it requires that it be included in the specification, and this will not happen of its own accord.

All the input data which we used in the economic planning department, for the techno-economic modelling which we were innovatively developing, had to be abstracted by hand from the basic traffic statistics reports and suchlike, these being on computer printout paper.

The statistical element of the system consists in keeping a record of unit-events, and abstracting this to form the basis of future planning (demand profiles, maintenance standards, etc). Again, the abstraction process will not be done if the system specification is left to the immediate users.

The budget control element consists in comparing "budget" and "actuals" and taking decisions to increase or decrease production in the light of short term trends. This is the driving-seat of short-term middle management. The main needs here are speed of response and knowledge of efficacy of measures adopted. Thus the adoption of an '(event-count) times (unit revenue minus unit cost)' as the measure gives a rapid response, as the signal does not have to wait until the accounts balance.

The abstracts form the basis for the use of the computer for long-term decision models. The most elementary of those is the Budget for next year's activities. With a properly planned information system all these are on disc or tape and no data need be manually handled.

PLANNING A SYSTEM

To the left is the "data processing world" which has grown up under the control of technician-programmers and system analysts. To the right is the "management science world", which has evolved by the interaction of graduate scientists with the thinking of top management The principal use of the computer here is the decision model.

If the information system is built 'from the left' the division is likely to be a 'wall'. To get a 'membrane' it is necessary for those connected with the right hand side to be in on the design.

The basic operational unit in system design is a team consisting of at least one member from each user-area at each level of abstraction from production. Thus the design of the operating statistics and costs system should include someone from the planning area to define the unit-cost needs.

The ''system analyst" is a key man here, in that he can translate needs into specification. He tends to be recruited internally and needs a period as a programmer so as to know what can be done. But the specification of the system should not be left to him alone, as.he is unlikely to have a feeling for the top-management view of the company, or any insight into what the significant control variables really are.

The actual programming is done to the system analyst's detailed specification, by programmers who work in a language appropriate to the job. There are three distinct levels of language: 1) machine-oriented language, suitable for "mass-production work", efficient as regards machine time, requiring sophisticated and committed technician-ship; 2) intermediate-level language suitable for extracting raw data from files in various ways; 3) high level language suitable for decision models and statistical abstraction.

The programming team should have people familiar with all levels. The lower level specialists require the system analyst to come between them and the problem. The high level language programmers however can talk directly to the problem formulator; the latter should preferably have enough command of the language to hand over a draft program containing the basic solution. The function of the programmer is to tidy up and make it conform to the program library standards.

Where the main system is committed to a rigid production-type job, there is a case for handing over the decision-model and statistical analysis work to a separate more flexible computer, working uniquely in high-level language.

The link between the two systems, which can be disc, tape or cards, is however still basically necessary, as it would be absurd to limit the second computer to manual input when the necessary abstracted.information is available from the 'mass-production' computer. This, however, must be provided for and that is why the design must not be left to technicians and middle-management.

CONCLUSION

The feedback from the IMI seminar attenders, I gather, was deeply divided; some of them loved it and were very impressed; others hated it and thought it a total waste of time. Regrettably I have no record of the attendance. None of it was derived from any text-book; it represented a totally original analysis of Aer Lingus experience. I like to think that maybe this pioneering 1968 paper was a foreshadowing of subsequent best practice, though many of the problems I identified are still with us.