7.2. Formulation Guidelines for Portraying Feedback Structure

It is sometimes said of our increasingly complex world that everything is connected to everything else, suggesting a huge tangle of interdependencies for people to manage and for the modeller to unravel. This is an overstatement. While it is true that more parts of business and society are connected now than in the past (think of mobile phones and the internet), the most important and useful connections are nevertheless sparse. Why? Anything that happens in organisations depends on the actions of ordinary people, who, no matter how talented, can handle only limited amounts of information as they communicate, collaborate and cajole to achieve their objectives. The complexity of society largely mirrors our own human capacity to process information and to coordinate action - and that capacity has not changed fundamentally since the industrial revolution or the Roman empire. Therefore, the task for the feedback system modeller is to portray, as convincingly as possible, the relatively sparse network of connections through which people enact plans and strategies. The majority of the network is made up of information flows.

Figure 7.5. Formulation guidelines

Source: Sterman, J.D., Business Dynamics: Systems Thinking and Modeling for a Complex World, © 2000, Irwin McGraw-Hill, Boston, MA. Reproduced with permission of the McGraw-Hill Companies.

There are five formulation guidelines to help modellers identify the most influential and realistic connections among the myriad that could in principle exist (Sterman, 2000, Chapter 13 on modelling decision making). They are summarised in Figure 7.5. The first and most general is the 'Baker Criterion', which requires that the inputs to all decision rules or policies must be restricted to the information actually available to real decision makers. Although this statement may seem obvious, it is vital for distinguishing, among the many potential influences on decision makers, those that really matter. Usually there are far fewer influences than one might at first suspect. In real life, managers use only a small fraction of the organisational information available to them to guide their decision making. Careful observation of managerial practice is therefore essential.

The term 'Baker Criterion' originates from Senator Howard Baker's famous question in the US Senate Hearings into the Watergate scandal in 1973: What did the President know and when did he know it?' The President (at that time Richard Nixon) could not be implicated in the scandal if he was unaware of the actions of his staff and subordinates. So figuring out just how much the President knew (and how much he should be expected to know) was of central importance. The same criterion can be usefully applied at any point of decision making in an organisation. What did the managers responsible for investment know and when did they know it? What did the people responsible for pricing know and when did they know it? What did customers know about the product or service and when did they know it. The point is that decision makers can't know everything and so, for the organisation as a whole, actions are only loosely coordinated and it can often appear that the left hand doesn't know what the right hand is doing. Such inconsistency is normal in management and is important for modellers to realistically portray.

The second guideline, a natural corollary of the Baker criterion, states that the decision rules or policies of a model should conform to managerial practice. All the influences on policies should be based on real-world counterparts and all variable names should be familiar to those working in the system. This advice leads to a descriptive model that represents the system as it currently operates, rather than an ideal model of how it should operate. Moreover, such a model is much more likely to engage managers and policymakers because it uses their own jargon and vocabulary.

The third guideline stresses the need for modellers to distinguish between actual and desired conditions in organisations. Actual conditions are what can be observed and measured such as capacity, workforce, price and so on. Desired conditions are in the minds of those with the power to act and reflect intentions, goals and aspirations. Nearly always, desired and actual conditions are different. Only by keeping the distinction clear is it possible to represent realistic disequilibrium pressures that drive corrective actions and contribute to dynamics.

The fourth guideline is a corollary of the third – equilibrium and optimality should not be assumed, they are convenient ideals but are not characteristic of the normal way organisations operate. Growth, fluctuation, stagnation, decline and overshoot are all disequilibrium phenomena and there is no guarantee that the various functions and sectors of an organisation that create such dynamics are optimally coordinated – far from it. A modeller who presumes optimal use of strategic resources or the existence of an efficient equilibrium between supply and demand is like a driving instructor who assumes that Michael Schumacher is representative of typical motorists. Perfection in timing, action and reaction is an interesting benchmark, but is not a realistic way to think about normally competent drivers or normally competent organisations. The ideals of optimality and equilibrium require lots of coordinating information, much more than decision makers typically use, and fast reaction times too.

The fifth formulation guideline states that decision rules should be robust under extreme conditions. This guideline can expose flaws in equation formulations and give useful hints about when and where to expect non-linearities in feedback loops. Consider, for example, a linear pricing formulation where price reduction is proportional to the gap between target demand and current demand. A sudden halving of current demand would be an extreme condition. If this thought experiment were to result in a negative price (something that would never happen in reality) then it would suggest a need to modify the pricing formulation, by, for example, adding a new non-linear pressure from margin on price change.

7.2.1. Review of Operating Policies and Information Flows in the Market Growth Model

Now we will use the formulation guidelines to interpret the operating policies and information flows of the market growth model, thereby shedding light on the model's feedback structure and its portrayal of growth dynamics. The main policies within the model boundary are customer ordering, sales force hiring, budgeting and production capacity expansion. We examine these policies one by one and then show how they fit together to yield the four feedback loops already outlined.

7.2.2. Customer Ordering

Figure 7.6 shows the customer ordering policy. To identify the influential information flows we need to get inside the world of the customer. Remember our model applies to the sale of a specialist technical product (advanced satellite navigation systems) to a manufacturer rather than to a consumer. Here we are thinking of a knowledgeable, technical buyer who needs to know about the functionality of the product, is seeking quality and reliability and wants the product delivered on time. The buyer is not particularly price sensitive since the navigation system is only a small proportion of the cost of a light aircraft or small boat in which it is installed.

In principle, all six factors shown in the diagram (availability, quality, relative price, sales effort, advertising and word-of-mouth) could influence customer ordering, and others besides. Which should be included? Here the formulation guidelines can help to tease out, through questioning of managers and business experts, the dominant influences. For this type of technical product and commercial customer, the main influences on ordering are sales effort, quality/reliability and availability, priorities that would become apparent from managerial practice, just as Forrester found in his original study. The other potential links (word-of-mouth, advertising and relative price) can be ignored in this case because they apply better to consumer products.

Figure 7.6. Customer ordering policy

A further simplification is possible. Quality and reliability can be eliminated entirely from the model if we assume they remain high and stable and therefore do not contribute to dynamics. This is a plausible assumption whose consistency with the facts can readily be checked. That leaves us with only two influences on customer ordering, sales effort and availability, and ultimately a much simpler feedback representation. The essence of the ordering formulation lies in the technical nature of the product, sold by a professional sales force to commercial buyers who want absolutely reliable delivery.

7.2.3. Sales Force Expansion

Figure 7.7 shows the policy for sales force expansion. In formulating this policy, the modeller's central question is how the company justifies the hiring of sales people. Again many influences may come to bear and the diagram shows seven possibilities. The amount of information in all these influences is far more than any management team would use and the Baker criterion suggests a need to be selective in order to identify those influences that conform best to managerial practice. What does the management team know and really care about in hiring, and when do they know it? Among the influences shown, it is possible to distinguish two distinct managerial approaches to sales force expansion. The influences on the left correspond to a budget-driven policy, while those on the right correspond to a forecast-driven policy. A budget-driven policy is one in which hiring depends on the sales budget and sales force operating costs (which in turn depend on the current sales force size and salary). If the budget exceeds operating costs then additional hiring is permitted and vice versa for firing. A forecast-driven policy is one in which hiring is linked to a market development plan that might include a business unit growth goal, sales objective or estimate of expected orders. Knowing the market plan and sales force productivity yields a target sales force. Either approach to hiring is perfectly valid, but in reality one or the other is likely to dominate. There is a need to read between the lines of management practice to discern which. In this case, we imagine the modeller has inferred from conversations that the sales department is budget driven. So the three factors on the left of the diagram are the active sources of information for the hiring policy. Once again a significant simplification is achieved in the information network and the resulting feedback structure.

Figure 7.7. Sales force hiring policy

7.2.4. Budgeting

Figure 7.8 is a glimpse of the myopic and political world of budgeting as it affects the sales function. In principle, there is an optimal way for firms to budget by allocating funds to those activities that generate the greatest marginal returns. However, formulation guideline number four tells us to beware of optimal decision rules. A truly optimal allocation of funds requires far too much coordinating information to be a credible description of what goes on in practice. Experience suggests that budgeting is normally myopic, often based on the pattern of past funding rather than a comprehensive trade-off among competing alternatives.

Figure 7.8. The myopic and political world of budgeting

This myopic logic is represented in two information flows. First, the total budget to cover operating expenses across all functions is assumed proportional to revenue. The more revenue generated, the bigger the total budget pie. Then comes the budget allocation, which is often based on precedent so that functions are normally entitled to last year's budget allocation. Sometimes strategic and political factors intervene (such as a growth target or desired size of 'empire') to cause departures from simple historical precedent. However, in this case precedent is assumed to be the dominant influence.

7.2.5. Capital Investment

Figure 7.9 shows three different approaches to capital investment, each relying on different sources of information. On the left is a finance-driven approach, while on the right is a planning-driven approach, and in the centre is an operations-driven approach. We begin on the left. A company that prides itself on rigorous financial appraisal might justify investment in terms of discounted cash flow, using information about expected revenue, machine price and a hurdle rate. Overlaid on this evaluation is the availability of funding that limits investment when the firm's pool of financial resources is low. By contrast, on the right is a forward-looking and planning-driven approach to capital investment. There are two information sources, expected demand and growth target, entirely different from the previous financial criteria and much simpler. They suggest a management team whose investment is driven by a belief in a market plan or vision for the future, rather than financial criteria.

Figure 7.9. The pressures and politics of capital investment

In the centre of the diagram is a reactive operations-driven approach. The information sources are different yet again. Here the top management team responsible for investment is assumed to pay particular attention to the capacity needs of the factory as reflected in current capacity, delivery delay and target delivery delay. Imagine the situation. Factory managers present their case for capacity expansion to an investment committee, arguing that current capacity is inadequate because delivery delay is too long relative to the target. The investment committee may or may not be sympathetic to the factory's case and has the discretion to fully or partially approve new investment. Such discretion is captured in the notion of executive bias and shown as a fourth influence on capital investment.

Of course in any firm all these sources of information are potential influences on investment. But that does not mean they should all be included in the modelled investment policy. Again the formulation guidelines come into play. What does the top management team responsible for investment know and really care about, and when do they know it? In this particular case, we adopt the same assumption as the original market growth model, that capital investment is operations-driven and reactive. Hence, the influential information sources entering the investment policy are reduced to only three: delivery delay, target delivery delay and current capacity, overlaid by executive bias. The implied style of investment is both practical and well-grounded in facts available from the factory as seen through the eyes of cautious executives who want solid evidence of demand before committing to capacity expansion.

The perceptive modeller looks for distinctive managerial attitudes toward investment. As a practical illustration consider the visionary (some would say reckless) capacity expansion plans of Sochiro Honda in the early days of the Japanese motorcycle industry.^[] He committed the fledgling Honda company to a massive increase of capacity, equal to twice industry capacity for the whole of Japan, to be installed in only two years. This investment programme was justified on the basis of his personal belief in the importance of the motorcycle to post-war Japanese society, and the manufacturing prowess of his own organisation. Moreover, he persuaded the rest of the management team to go along with his bold plan. There was no tangible business evidence to support his view, no burgeoning delivery delay, no solid cash flow projections nor even a demand forecast based on formal market research. The point is that influences on investment vary from one firm to another, sometimes dramatically. It is the job of modellers to recognise which influences dominate. Incidentally, Honda's blinkered optimism proved to be justified and his company grew swiftly at a time when less bold rivals held back their investment. Passionate belief in a future can drive successful capital investment, but not always.

^[] A description of Sochiro Honda's visionary and hugely confident approach to developing his fledgling motorcycle company can be found in a classic Harvard Business School case study entitled 'Honda (B)', Christiansen and Pascale, Case No. 9-384-050, HBS Case Services Cambridge MA. In the company's very early days he founded the Honda Technical Institute. It sounded like Bell Laboratories, but in reality it began as one man tinkering, who had a passion for engines and motorcycle racing. The point is that he believed his workshop would become an Institute. The first factory was an old sewing machine plant, but it had plenty of space for growth. A combination of supreme self-belief coupled with technical know-how and hard work led him to take risks in capacity expansion that no other rival would dare. In all likelihood he did not perceive the risk that others did, but instead saw a vast opportunity. This attitude underpinned the company's bold capital investment policy.

7.2.6. Goal Formation

At the heart of capital investment is corrective action – adding more capacity because it is deemed necessary. In this case, the trigger for investment is target delivery delay. But where does the target itself come from? It need not be a constant. In fact, it is often useful to view a management target as the output from a decision-making process of goal formation. In other words, targets and goals are a matter of policy. Recognising such fluidity naturally focuses attention on the information sources that drive goal formation. Figure 7.10 shows four potential information sources for target delivery delay. On the left are market-oriented factors such as competitors' delivery delay and the delivery delay expected by customers. On the right are operations-oriented factors such as the delivery delay recognised by the factory and a delivery delay management goal. Yet again the formulation criteria are important for recognising which information sources best coincide with management practice. It is highly unlikely that all four will be used. In some firms it will be possible to discern a culture of market awareness that gives priority to the factors on the left. In other firms, dominated by manufacturing, the factors on the right will be given priority. The modeller should not assume that one or other orientation is used just because prevailing wisdom or accepted theory says it should be. In this particular case, we choose the same operations-oriented formulation as found in the original market growth model.

Figure 7.10. Target delivery delay as the output from a goal formation policy