1 4 0 3 3 4 2 1 1 5 S W P 15/90 S T A T E - O F - T H E - A R T D E V E L O P M E N T S IN E X P E R T S Y S T E M S A N D S T R A T E G IC M A R K E T I N G P L A N N I N G P R O F E S S O R M A L C O L M M C D O N A L D a n d H U G H W I L S O N Cranfield School of M a n a g e m e n t Cranfield Institute of Technology Cranf ield Bedford M K 4 3 O A L (Tel: 0 2 3 4 7 5 1 1 2 2 ) Copyright: M c D o n a l d a n d Wilson 1 9 8 9 STATE-OF-THE-ART D E V E L O P M E N T S IN E X P E R T S Y S T E M S A N D S T R A T E G I C M A R K E T I N G P L A N N I N G by Malcolm H.B. McDonald and Hugh N. Wilson A B S T R A C T T h e paper describes a case history of the development of a n Expert System in Strategic Marketing Planning c o d e n a m e d E X M A R . It traces the evolution of the system from the formation of the DTI club two years a g o to the launch of the prototype model. T h e paper outlines the technical a n d domain-specific obstacles encountered e n route a n d h o w these were overcome. A n u m b e r of conclusions are drawn from the project. T h e principal o n e is that there is a bright future for expert systems in the field of strategic management. Professor Malcolm McDonald, o n e of the authors of this paper, is the principal expert to the club. H u g h Wilson, the other author, is a senior consultant with Artificial Intelligence Ltd. H e was also the knowledge engineer a n d the project manager. - 1 - -c - STATE-OF-THE-ART DEVELOPMENTS IN EXPERT SYSTEMS AND STRATEGIC MARKETING PLANNING by Malcolm H.B. McDonald and Hugh N. Wilson 1. INTRODUCTION After nearly a quarter of a century of Expert Systems, virtually no progress has been made in the domain of marketing and there are few products and no on-line systems available’ O2 There are no shortcuts to building good Expert Systems. It takes a considerable amount of skill, patience and several years of effort to develop an Expert System in a new area and get it in the field3. -.. During the 198Os, Japanese activity in the field of Expert Systems and related technologies prompted the EEC to give birth to the ESPRIT programme in an attempt to integrate European efforts. This in turn led to the DTI sponsored ALVEY and IED programmes, and other initiatives. An outcrop of these is a new DTI - sponsored club called EXMAR. EXMAR is a club of ten major British companies. Formed in 1987, its objectives are to investigate the possibility of computerised assistance for strategic marketing planning by the development of a prototype, and to spread awareness of expert systems in club member organisations. It is funded by contributions from the member companies, and by the Department of Trade and Industry. The club’s primary source of marketing expertise is Professor Malcolm McDonald of Cranfield School of Management. Marketing experience within club member companies is also being tapped. - 2 - T h e involvement of Artificial Intelligence Ltd. of Watford, with the club as technical contractor b e g a n in the second half of 1988, w h e n AI Ltd. conducted a n analysis phase, followed by production of a demonstrator a n d a n appraisal of the way forward. T h e demonstrator was built in the Interlisp programming environment a n d the Loops object system o n Xerox 1 1 8 6 workstations. T h e Requirements, Functional a n d Design Specifications for a prototype, to b e used for experimentation a n d evaluation by club m e m b e r s in their o w n organisations, were completed in August 1989, a n d work is n o w in progress o n the implementation. T h e prototype will run o n IBM-compatible 3 8 6 machines, using Smalltalk- a n d Analyst, a n d was completed in January 1990. Pumose and Structure of PaDer T h e purpose of this paper is to outline the progress of the E X M A R project, a n d to draw conclusions about appropriate computer support for marketing planning. In the next section, the approach taken to the analysis phase at the start of the project is outlined, a n d the system objectives that were derived are described. T h e nature of the logical model that e m e r g e d is discussed, a n d the demonstrator system based o n it is described, emphasising the nature a n d style of the support to the user provided by the system, h o w this reflects the logical model, a n d h o w this meets the system objectives. Section 4 describes the feedback from club m e m b e r s o n the Demonstrator a n d h o w this is influencing current work o n the Prototype. T w o review sections 5 a n d 6, discuss the nature of the marketing planning domain. a n d the deductions that can b e m a d e about appropriate computer support. - 3 - Finally, s o m e conclusions are drawn o n the appropriate approach to systems development in such “soft” areas as marketing planning. - 2. P R E V I O U S W O R K A N D E A R L Y O B S E R V A T I O N S T h e initial requirements analysis produced a n u m b e r of interesting problems for the project, which were to sow the seeds of expensive a n d time-consuming delay. These problems can b e summarised as follows:- .- - (8 (ii) (iii) it b e c a m e clear that not m a n y of the m e m b e r companies were particularly a u fait with the methodology of marketing planning. This led to the problem of setting clear objectives for the project. the diversity of c o m p a n y industry types, ranging from capital goods to service industries, meant that n o subsequent system could possibly b e suitable for all circumstances. problems a n d subsequent proposed objectives ranged from “T o support a formal planning framework to improve discipline during the planning process” a n d “T o support further understanding of the effects of currency fluctuations” to “T o promote discipline in pricing control”. For these reasons, it was decided to focus o n the process of marketing planning itself rather than o n any situation-specific system. A firm of software consultants was appointed project m a n a g e r a n d a knowledge based systems house was appointed principal contractor. - 4 - -. ,- The systems house began a series of twelve half day interviews with Professor McDonald in order to develop a formal paper model as a basis for computerisation. Unfortunately, although taped and transcribed, they were largely unfocussed due to the inexperience of the interviewers and little progress was made towards formal modelling of the marketing planning process, in spite of very specific guidance given by Professor McDonald to the interviewers. The problem centred around lack of proper project control by the project managers, confused expectations by members of the club based on marketing planning naivety, the inexperience of the knowledge engineers, and the passive role of the domain expert, which was necessary in view of the nature of the project. Several attempts on Professor McDonald’s part to guide the system were brushed aside as politically inexpedient. The result was that the paper outlining the tasks to be performed by the computer system targeted the whole marketing planning process rather than any subset, and because of this breadth, the process to be computerised was not documented in any detail, nor backed up by any substantive models and interrelationships. Other specifications required by the development methodology in use, such as financial requirements, system structure and so on, were never produced. At this point, the project manager appointed new software consultants to take over the feasibility study and the delivery system. The new contractor set about finding some common requirements among end users in order to outline the domain model, with a boundary definition showing which parts of the model would be tackled by the computer system. They set about establishing the following areas: . n scope n constraints n organisational impact - 5 - # - rc- -- - n maintainability n extensibility n technology n time scales n risk a n d cost versus quantifiable benefits Artificial Intelligence Ltd., the n e w software consultants, drew various conclusions about the appropriate technical approach: n Need for focus T h e previous work h a d b e e n o n a broad front, involving analysis into all aspects of strategic marketing planning. This is a vast topic, tackling m a n y of the most fundamental problems inherent in business activity, a n d progress was therefore slow. There was a n e e d to focus o n a subset of the overall problem. n Feasibility and utility to be established T h e very title of the club, “Expert Systems in Marketing”, suggested that the use of expert systems techniques in this area was possible a n d appropriate. This assumption of feasibility was based o n the observation that there existed demonstrable expertise, but why this might imply a classic rule-based expert system h a d not b e e n addressed. This was a doubly large assumption as n o previous systems (or work towards systems) were known in this application area. There was a n e e d to address this early, a n d the related issue of h o w any system would b e of use to the marketing planner. n Modelling and representation It was decided that the appropriate first step was to carry out analysis in a closely scoped subset of the problem, with the emphasis o n modelling the area using whatever formal techniques were appropriate. A n example of the choices deliberately not m a d e at the start was whether any modelling of expertise adopted the - 6 - rr- “low road” of embedding the expertise in data structures and code, the “high road” of an explicit, “deep” representation, or the “middle road” of an explicit but heuristic representation4. In this modelling work, the emphasis would be on representation rather than computation, as the essential first step towards any computer system. n The marketing swamp Marketing will be referred to later in this paper as a swamp of intuitive, experience- based practice with the occasional rocky peak of formal techniques. In the experience of Artificial Intelligence Ltd. the best place to start when modelling such “soft” domains was often on the boundary between the soft area and neighbouring more readily formalisable areas. In this case, that meant starting with the established formal techniques and working out from there. - 3. RESULTSOFANALYSISWORK. ANDDEMONSTRATOR Several analysis sessions were held with Professor McDonald, and with marketing practitioners in club member organisations. This resulted in an overall EXMAR system objective, an outline model that was used as the basis for a demonstrator system, and a list of areas where further work was required. The overall EXMAR system objective was defined to be: “To provide assistance for the marketing planning process in such a way as to spread knowledge and further understanding of how and why the multifarious factors of the market interact and serve to define the parameters of the business activity.” The remainder of this section describes features of the model, and how these were exploited in the demonstrator. The structure is an interleaved description of the two: - 7 - - L - - each subsection describes a model feature, a n d the relevant aspects of the demonstrator. 3.1 Assistance to aid in interwetation and understanding T h e model covers the data manipulated by a marketing planner w h e n developing a strategic marketing plan, a n d structures the marketing planner’s task. M a n y of the individual subtasks or processes of this task involve modelling by the user of the business context, or interpretation by the user of the information entered. There is m u c h that a computer system based o n the model cannot d o for the user, a n d it b e c a m e increasingly clear that its most appropriate aim is to assist. T h e objective of the demonstrator was therefore to provide a n interactive system that supports a marketing planner by providing tools that help the user to represent the state of the markets a n d products under consideration; to interpret this information so as to gain a n understanding of the markets a n d o n e ’s place within them; a n d to determine a course of action based o n this understanding. 3.2 Model of the Drocess of eeneratine a marketing nlan A hierarchical breakdown of the process the marketing planner should adopt to generate a marketing plan was defined. Encouraging the user to adopt this process is of value in itself, as the process incorporates m u c h experience that helps avoid c o m m o n pitfalls: for example, the n e e d to arrive at a n appropriate understanding of the current situation before setting objectives for the future. T h e demonstrator uses this hierarchy as a basis for the user’s navigation round the system. T h e initial screen display is shown to illustrate this (Figure 1). Also sholvn is a window for m o r e detailed navigation round a particular stage. - 8 - - II E X M .4R. demonstrator Araldal Illtd~ U m N e d Prodlet relMlonshlp W I U I M * . W 1 S e t Objectha fof N e w Productl/Markatr Figure 1 Initial screen display, with an example of a detailed browser E a c h box in the graphical browsers represents a stage of the process. T h e user carries out a stage by selecting a box with the m o u s e : the system then takes the appropriate action, which m a y for e x a m p l e b e to present the user with a form to fill in, or to o p e n a m o r e detailed browser of the process for that stage. T o give a n overview of the process: Select/Define Business Unit identifies which area of the business the marketing plan is for, a n d records the purpose of the business area. F o c u s identifies which of the unit’s markets a n d products are of interest. Conduct Audit assesses the current position of the products a n d markets. Forecast predicts the future position of the products a n d markets, a s s u m i n g w e d o not intervene, as a base-line for objective setting. Finally, Set Objectives a n d Strategies sets objectives for the business unit b a s e d o n the information collected, analysed a n d summarised; a n d defines strategies by which the objectives are to b e met. - 9 - - Detailed browsers contain icons showing the nature of the support offered for a particular stage: for example, there are icons for graphical displays of information, for tables of numbers, a n d for free text. T h e Predict Relationshiu with Markets browser is illustrated as a n example, also in Figure 1. - Users will largely g o through the process depth first a n d top to bottom; but they are free to d o otherwise, as there are m a n y cases where they m a y legitimately wish to d o so. l Markof Share F A C T O R S l Price l Sales V o l u m e . 3.3 Generallv aDDlicable. sound data model A data model was developed that captured a n d related the information considered during production of a strategic marketing plan. It has proved essentially sound, a n d of general applicability to the wide range of marketing situations represented by the diverse club m e m b e r companies. A simplified entity-relationship diagram of the model is given below in Figure 2, a n d briefly described. I N V O L V E M E N T IN ‘s c o m o n M w * a n d M A R K E T ~emllscom S N A P S H O T P R O D U C T F O R P R C O U C T S N A P S H O T L P R O D U C T - 10 - - T h e model has three cornerstone entities: Business Unit, the part of the organisation for which the plan is being developed; Product, the products or services offered by the unit; Market, the markets in which it operates. Critical Success Factors model the workings of a market by documenting the factors critical to the success of any product in the market, from the consumers’ viewpoint. They are a n objective assessment of h o w the market works, independently of the Business Unit’s presence in it. T h e matching of products to markets is represented by the important Product for Market entity: a product’s score o n the Critical Success Factors relates to this entity. Market Attractiveness Factors model the priorities of the business unit by documenting the factors determining h o w attractive a market is to the unit. Being a subjective assessment of the business unit’s priorities, the criterion for their correctness is the agreement of key executives. T h e matching of markets to business units is represented by the important Involvement In Market entity: a market’s score o n the Market Attractiveness Factors relates to this entity. Time-dependent information is held in Snapshot entities. For each plan, the demonstrator system holds data structures closely based o n the data model. T h e user’s primary m e a n s of manipulating the data is by using forms, which are illustrated in Figure 3. - 11 - Weight IBL Score Competitor 1 1: Quality 20 9 1.8 4 .8 2: Price 1.5 5 .lS 6 .9 3: Differentiation 25 7 1.75 4 1.0 4: Image 20 7 1.4 3 6 5: support 20 9 1.8 6 1.2 [BE(l[ Strength in Market: IBL 7.5 Competitor 1 4.5 Figure 3 Typical forms for data manipulation - - The top form shows current information about the Food Processing market for the fictional International Bearings Limited (IBL) company, which sells bearings into a variety of markets. The bottom form shows the Critical Success Factors defined for this market, with weights to illustrate their relative importance. For example, while price is important in this market, it is less so than several other factors, such as product differentiation and quality, the product’s image, and the engineering support provided. It also shows a score for IBL and its main competitor against these factors, and a weighted average computed by the system, to represent IBL’s overall strength in the market. This is copied to the top form by the system. The Market Attractiveness score on the top form results from a similar weighted average form for the attractiveness of the market against such criteria as the market’s size, growth and profitability. - - 12 - - - - 3.4 The use of techniaues in the data model The “rocky peaks” with which the analysis work started are “textbook” techniques for analysing an organisation’s markets and products, such as the Directional Policy Matrix, which is illustrated below (Figure 4), the Boston and Porter matrices, and so on. These view different aspects of the data model using differing graphical representations, to aid in interpretation of the data. To extend our analogy, the data model thus forms the bridges between the rocky peaks to enable us to navigate the intervening swamp. M Higt j q,, r”‘roacmt c \.I I p&y- E 1 I 0 T T,4 h I &- A T R s F-7. Auruman ‘! \ T I ~nmmiul V ‘. : E \. N ELow s <_r, s High SI-RENGTH IN MARRET Low Quadrant advice: Maintain market position, manage for cash Figure 4 Data presentation to aid understanding - 13 - .- - - The screen snapshot in Figure 4 gives an example of how the demonstrator exploits these features by showing the underlying data presented in the standard formats. The Directional Policy Matrix plots, for each of IBL’s markets, the market attractiveness against IBL’s strength in the market. The size of the circles is proportional to the market’s contribution to IBL’s revenue (though it could have been set to any useful metric). Different circle shadings illustrate the current, forecast and objective situations for the product/market. (In terms of the data model discussed earlier, each circle strictly represents a Product For Market,) The matrix aids in understanding both the situation of an individual product/market, and the balance of the portfolio of products. An example of the matrix’s interpretation is that in all its markets, IBL is moving downwards and rightwards from the current to the forecast situation. This indicates a general weakening of IBL’s position: the matrix illustrates what IBL intends to do about this for the automotive market by maintaining its competitive position while cutting costs where possible. The demonstrator also provides on request standard, “textbook” advice for a product- for-market in a given position on the matrix, as a guide to the planner in setting objectives. For example, for the automotive market, the system advises that the market position (strength in market, and market share) be maintained, but that subject to this the market be managed for cash to fund development of more attractive markets. This is the only case in the demonstrator where it was felt appropriate that the system should take an active role of giving advice, rather than the passive role of presenting information in differing forms to aid the user in interpretation. - 14 - - - _- - The diagram also shows a “gap gauge”, a bar chart showing the financial gap between the business unit’s target revenue and the sum of the individual objectives so far set for the various markets. 3.5 Less structured information: checklists. free text Some parts of the marketing plan were best expressed in text: for example, the business unit’s mission statement, and lists of opportunities and threats. Also, in several areas, marketing expertise was identified that was not formalised beyond free text in the model. Examples are checklists of common critical success factors; assistance with definition of a business unit’s mission statement; and checklists of possible opportunities and threats to consider. This unstructured information was related, however, to specific points in the planning process, or to specific items in the data model. The demonstrator exploited this by making available text windows at appropriate points with icons on the browsers and elsewhere. This was implemented using the NoteCards hypertext system. 4. DEMONSTRATORFEEDBACKANDPROTOTYPE Initiative is with the user The demonstrator leaves the user to decide what to do next. This was liked by the club members, who felt it to be appropriate for this application. Evidence of utility Club members felt an operational system based on the demonstrator’s ideas could be of significant use in the vital process of strategic marketing planning. This is an example of utility being addressed by the clients rather than by the developers. - 15 - - - - - Communication of the nature of the proposed prototype T h e demonstrator served to communicate the nature of the support that would b e offered to a marketing planner by a fuller computer system, to club m e m b e r s a n d to the primary expert, Professor McDonald. With this innovative system, this was difficult to achieve o n paper. Use in specification of prototype T h e demonstrator has b e e n used effectively in discussions with club m e m b e r s to aid with specification of the prototype n o w being developed. 5. R E V I E W O F T H E V A L U E O F A P P L Y I N G E X P E R T S Y S T E M S T O T H E M A R K E T I N G P L A N N I N G P R O C E S S During the 196Os, attention was focussed o n specific problem-solving applications in scientific fields. M a n y successful Expert Systems have b e e n built, including M Y C I N for diagnosing infectious diseases’, a n d P R O S P E C T O R , a system for evaluating geographical locations for possible mineral deposit8. M a n a g e m e n t problems, however, d o not lend themselves to quite the s a m e precise logic as scientific problems. People d o not solve most of life’s problems by mathematical means, but rather by experience, knowledge a n d intuition. Marketing problems are dealt with in the s a m e way, as most of them are logical rather than mathematical, a n d problem-solving knowledge, whilst available, is incomplete. Decision-Support Systems a n d the like have traditionally used hard facts a n d static formulae which, given the correct data, provide correct answers. They belong m o r e naturally to the logical, black-or-white, right-or-wrong world of computers. But m a n a g e r s in the world of marketing deal with uncertainties a n d often with vague concepts. O n e approach to pinning d o w n the basis for such decisions is to attempt to - 16 - - - model the decision-making process as a set of “rules”, or heuristics, that reflect the expert’s o w n knowledge a n d experience about the problem in question. These “rules” are hard to nail d o w n a n d quantify - partly, perhaps, because the expert’s experience enables him to think in terms of shades of grey, “m o r e or less”, a n d “approximately” - a n d partly because rules are not always a n appropriate or sufficient representation. W h e n h u m a n beings find a path through situations that are too complex a n d a m o r p h o u s for the h u m a n mind to handle in a totally conscious, rational, scientific way, it can b e difficult or impossible to elicit the m e a n s by which they d o so. Most people would acknowledge that in virtually any walk of life, the true expert has built u p his expertise largely from experience a n d a n intuitive grasp of problem- solving in the real world, something which is often referred to as the “University of Life”. Indeed, m a n y of the world’s leading business people acknowledge that they o w e their success not to formal business education a n d text books, but to their o w n experience, flair a n d intuitive g o o d judgement. Donald Schon’ describes this p h e n o m e n o n as follows: “Competent practitioners usually know m o r e than they can say. They exhibit a kind of knowing-in-practice, most of which is tacit”. H e cites a n investment banker, w h o makes his decisions based o n 7 0 to 8 0 per cent instinct, a n d only 2 0 to 3 0 per cent calculable rules. This “gut feel” was a major asset to the bank in question. His point is that artistry is not reducible to discernible routines. H e describes scientific rigour as “describable, testable, replicable techniques derived from scientific research, based o n knowledge that is testable, consensual, cumulative a n d convergent”, but then goes o n to argue that m u c h of what passes for scientific m a n a g e m e n t is irrelevant because business problems d o not c o m e well formed. Certainly, most marketing problems are messy a n d indeterminate a n d successful practitioners m a k e judgements using criteria which are difficult to define. M a n y - 17 - - - - academics would decry this as a lack of rigour, and in so doing exclude as non- rigorous much of what successful practitioners actually do. The following quotation from Schon neatly sums up the problems facing not only teachers and researchers of marketing, but, more importantly, the initiators of expert marketing systems : “In the varied topography of professional practice, there is a high, hard ground which overlooks a swamp. On the high ground, manageable problems lend themselves to solution through the use of research-based theory and technique. In the swampy lowlands, problems are messy and confused and incapable of technical solution. The irony of the situation is that the problems of the high ground tend to be relatively unimportant to society at large, however great their technical interest may be, while in the swamp lie the problems of greatest human concern.” Marketing Planning remains one of the last bastions of ignorance in the field of marketing. The benefits ‘of marketing planning are well documented and agreed,8 yet so complicated is the process of marketing planning, and so confusing are the interrelationships between the tools and techniques of marketing planning’, that very few British companies enjoy these benefits, as has been shown by a seminal paper by Greenley lo that reviewed all the major UK empirical research in this area. Indeed, there were as many dysfunctional results from the attempts of companies to initiate marketing planning procedures as there were benefits. The problem to be addressed by Expert Systems in the marketing domain centres around how to take account of the intuitive artistry displayed by experts in situations of complexity and uncertainty in a way that is describable and susceptible to a kind of rigour that falls outside the boundaries of technical rationality. An important - 18 - - /- - .- aspect of this is to identify where a computer system cannot hope to solve a problem on its own, and in such cases, how it can best assist the marketing manager. The question, then, is how an epistemology of practice can be captured and represented in an Expert System. 6. APPROPRIATE TECHNOLOGY IN THIS DOMAIN 6.1 Previous svstems To date, no one to our knowledge has seriously tackled the world of marketing with Expert Systems other than the MSI A D C A D ’ ’ system developed to advise on advertising design. After considering a variety of consumer and environmental factors, advertisers use a combination of empirical research, communication theory, and rules of thumb, to select communication objectives and select appropriate creative approaches. The authors themselves list a number of weaknesses in ADCM, but conclude: “As one advertising executive put it: “it helps us to think a little deeper about the issues we have to consider in developing ads that are both strategically and executionally kound”. Another interesting and relevant conclusion was that most managers, when asked, said they would like to make use of existing theoretical and empirical knowledge of marketing when making decisions. However, few actually did use such knowledge. Expert Systems can bridge this gap by structuring, validating and disseminating marketing knowledge, whilst at a theoretical level, they challenge their creators to understand and critically evaluate the elements of marketing knowledge and their interrelationships. - 19 - - - - 6.2 Problems suitable for exDert svstems In deciding whether marketing planning was a sensible domain for the application of Expert Systems technology, the M S 1 checklist” fits our experience. Four criteria are provided: 8 Are the key relationships in the domain logical rather than arithmetical ? Concepts such as the strength of a product in a market, the attractiveness of a market or product differentiation are clearly not arithmetical - though n u m b e r s m a y b e used to g o o d effect in clarifying thought. S o the answer is “yes” (though not in the sense of mathematical logic). 8 Is the problem domain semi structured rather than structured or unstructured? T h e marketing planning process has both structured elements - for example, the segmentation of a market, the relevant financial information - a n d related unstructured elements - for example, a mission statement or a list of forecasting assumptions. S o the answer is “yes”. 8 Is knowledge in the domain incomplete ? Since marketing planning a n d all its contextual problems remains o n e of the most under-researched areas of marketing, a n d since little has b e e n published about the interrelationships of all the techniques of marketing in systems design, the answer is “yes”. This is in fact the key to the whole project a n d why it was chosen in the first place by the club members. 8 Will problem solving in the domain require a direct interface between the m a n a g e r a n d the computer system ? - 2 0 - - T h e intention is to have operational marketing m a n a g e r s using the system for the production of marketing plans, a n d the system can solve few of the problems by itself, so the answer is “yes”. 6.3 Role of the comauter T h e potential benefits shown by the E X M A R demonstrator are d u e mainly to its assistance with the understanding a n d interpretation of the information entered. T h e e n d result m a y include a marketing plan, but it also includes a n e n h a n c e d a n d readily communicable understanding of the business gained by the marketing planner. These benefits are largely d u e to appropriate a n d varied display of the information. Apart from data presentation, a computer system in this domain can perform the tasks for which computers have traditionally b e e n used: m a n a g i n g data, maintaining constraints between data items like a speadsheet, a n d performing routine calculations. These free u p the user’s thoughts for higher-level problems. Finally, in s o m e cases the computer can b e m o r e pro-active, offering advice, pointing out decisions that g o against conventional wisdom, a n d so on. T h e most appropriate technology for this mix of roles will itself b e a mix. In the case of E X M A R , the software techniques included object-oriented programming, hypertext a n d use of windows-based programming environments, to enable swift development a n d a carefully tailored user interface. W e have not found rule-based representations so far to b e relevant, though they m a y b e in future developments. T o s o m e Expert Systems workers this emphasis o n data presentation a n d low-level data management, as o p p o s e d to sophisticated calculation or reasoning, would constitute s o m e sort of failure. W e consider, however, that the objective of computer systems is to m a k e the combination of user a n d system m o r e effective than the user alone, not to build ‘clever’ computer systems. Even in the classic scientific expert systems such as those quoted earlier, the user interface frequently constituted m o r e of the work, a n d m o r e importantly delivered m o r e of the benefits, than emphasis in the literature would suggest. W e suggest that this applies even m o r e in such ‘soft’ a n d ill-understood areas as marketing planning: the rapid recent progress in the power, a n d price, of the underlying software tools that enable graphical user interfaces to b e provided will enable m o r e such areas to b e tackled effectively in the future. - 2 1 - - - 6.4 Analvsis wroach T h e analysis approach u s e d for E X M A R w a s undogmatic a n d modest: to m o d e l the available expertise with whatever modelling techniques proved m o s t appropriate, starting with the m o s t well-established a n d documented, a n d verified, expertise. “D o n ’t run before you can walk” should not n e e d emphasising: but the early experience of the club shows that perhaps it still does. T h e very term “Expert Systems” has led s o m e to unjustified assumptions not just of the feasibility of building computer systems b a s e d o n expertise, but also of their utility, a n d of the m o s t appropriate modelling a n d system-building tool$. T h e alternative is classic software engineering, with a n e x p a n d e d tool kit of analysis a n d implementation techniques to d r a w u p o n as appropriate. This m a y lead to the question about h o w a n d to what extent the m o d e l a n d demonstrator m a y b e said to incorporate expertise. All aspects of the m o d e l a n d demonstrator can reasonably b e said to b e b a s e d o n expertise: the process, the data model, the m e a n s of presentation of information, the checklists provided, a n d the o n e case w h e r e data-dependent advice is given. T h e system thus takes the “low road” according to B r o w n ’s categorisation discussed earlier. T h e r e is certainly m u c h available (but not necessarily formalisable) expertise that has not b e e n captured: the critical design task has b e e n the effective definition of the boundary between the system a n d the user such that the user is encouraged to think about the issues that the system cannot of itself address. This conforms to the stated E X M A R system objective quoted earlier, of providing assistance for the marketing planning process in such a w a y as to spread knowledge a n d further understanding of the business a n d its markets. 7. C O N C L U S I O N S A n u m b e r of conclusions can b e d r a w n from the E X M A R experience: 6) T h e development of E X M A R shows that it is possible to use Expert Systems methodologies to build support systems in c o m p l e x areas of marketing m a n a g e m e n t , especially if the d o m a i n is well defined, has a large n u m b e r of factors to b e considered a n d relevant expert knowledge is available. (ii) T h e m o r e c o m p l e x a n d a m o r p h o u s the expertise to b e captured, the longer it takes both the expert a n d the knowledge engineer to reach a n acceptable - 22 - - - approximation. It is clear that to develop an Expert System that is of some practical use requires both time and resources of massive proportions. This is supported by the MS1 research paper 11 , which concludes: “There are no shortcuts to building a good Expert System. It takes a considerable amount of skill, patience, and years of effort to develop an Expert System in a new area and get it into the field”. (iii) Expert Systems provide a consistency to human decision making which is valuable, since people tend to forget or ignore knowledge. (iv) EXMAR has generated considerable interest and support among, the major multinational companies that form the club, because it forces them to think deeply and in a structured way about the issues that need to be considered in developing a strategic marketing plan. (VI Expert Systems are useful in helping both academics and practitioners to structure, validate, and use marketing knowledge and to better understand the interrelationships between the elements of marketing. (vi) Tight project control is vital. This view is supported by Mumford13. Many issues need to be considered, such as clear definition of subject matter, availability of inputs, and clear agreement with users on objectives, timescales and resourcing. The close involvement of the EXMAFt club members has been essential in this respect. It has been achieved through an active working party, through agreed quality assurance criteria for each stage of the work, and through the use of a demonstrator. (vii) The potential advantages of Expert Systems in marketing are: m consistent advice - 23 - H secure knowledge bases n making better use of experts H enhanced decision making n improved analysis. (viii) Since we live in an imperfect world, with imperfect problems and imperfect tools, it is unreasonable to expect a perfect Expert System until there are perfect experts and perfect technology. On the other hand, if an Expert System gives better advice than you would have had without it, it is probably worthwhile. In conclusion, it is unlikely that Expert Systems will ever be able to give the same value as real human experts, although clearly they can offer reasonable advice. Nor will they guarantee that you make the right decisions. But they can help you gain a proper perspective of the alternatives. In a sense, Expert Systems will always be a bit like Distance Learning programmes, which can replace a bad teacher, but never a good one. - 24 - / -. r 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. Moutinho L. and Paton R. “Expert Systems : a New Tool in Marketing” Quarterly Review of Marketing, Summer, 1988 Foster E. “Artificial Intelligence” Personal Computing, April, 1985 Cebrzyask “Artifical Intelligence : the goal is to store an expert’s real knowledge on a disk” Marketing New, Vol. 21, No. 5, 1987 Brown J.S. “The low road, the middle road and the high road” in The AI Business, P.H. Winston and K. Pendergast, Eds. MIT Press, Cambridge, Mass 1984 Buchan B. and Shortliffe E. “Rule-based Expert Programs : the MYCIN Experiments of the Stanford Heuristic Programming Project, Reading, MA., Addison- Wesley, 1984 Duda R., Gaschnig J. and Hart P. “Model Design in the Prospector Consultant System for Mineral Exploration” in “Expert Systems in the Micro- electronic Age” ed. Michie D., Edinburgh: Edinburgh University Press, 1979 Schon D. “The Crisis of Professional Knowledge and the Pursuit of an Epistemology of Practice”, Research Paper for the Harvard Business School 75th Anniversary Colloquim on Teaching by the Case Method, April, 1984 McDonald M. “The Theory and Practice of Marketing Planning for Industrial Goods in International Markets”, Cranfield Institute of Technology, PhD., 1984 Leppard J. “Marketing Planning and Corporate Culture - a Conceptual Framework which examines Attitudes in the Context of Marketing Planning” Cranfield Institute of Technology, M.Phil., 1987 Greenley G. “An Exposition into Empirical Research into Marketing Planning”, Journal of Marketing Management, 3.1., July, 1987 Rangaswamy A., Burke R., Wind J. and Eliashberg J. “Expert Systems for Marketing” Marketing Science Institute Working Paper Report, Nos. 87-107, 1988 Bobrow D.G., Mittall S. and Stefik M. “Expert Systems, Perils and Promise”, Commun. A.C.M. 29.9 (Sept. 1986) pp. 880-894 Mumford E. “Designing Computer Based Systems”, University of Wales Review Business and Economics, 3, 1988 -