10.5. Model Structure Tests and the Soap Industry Model
Model structure tests build confidence in the model by demonstrating that its concepts and relationships are consistent with observations about structure and policies from the mental database of people who know the business well – in this case the senior managers from marketing, sales and manufacturing. These tests are particularly important in system dynamics because structure is central to understanding dynamic behaviour. Moreover, much of the data for identifying structure resides in the mental database as Figure 10.6 illustrates. The figure distinguishes three different types of information in the mental database: observations about structure and policies; expectations about system behaviour; and actual observed system behaviour.
Figure 10.6. Role of the Mental Data Base in Modelling and Confidence Building
Source: Jay Forrester, 1994, Policies, Decisions and Information Sources for Modelling in Modelling for Learning Organizations, Productivity Press. Reproduced by permission of Jay W. Forrester.
The mental database is particularly rich in structural detail about operations at a functional level. For example, at Old English the marketing manager could explain promotional price discounts and their impact on consumers. The factory manager knew the plant intimately and appreciated the practical difficulties in shifting capacity from bar soap to liquid soap. Collectively, the management team holds a wealth of information about causal links, operating procedures, decision rules and the relative priority of information flows – all of which underpin effective business modelling. The mental database also contains information about past behaviour (trends and patterns in key variables) that is useful for guiding model conceptualisation and for building confidence in simulations. In this case, the decline in bar soap volume was well known and easily cross-checked with time series data. However, not all information in the mental database is reliable. People's expectations about system behaviour are often misleading because they cannot infer cause and effect. Simulation is the only reliable way to relate the structure and dynamic behaviour of business and social systems.
Bearing in mind the contents of the mental, written and numerical databases, there are five main tests of model structure as listed below. The first two tests apply to conceptual models and the remaining tests apply to algebraic models.
Boundary adequacy – are the important concepts for addressing the problem endogenous to the model?
Structure verification – is the model structure consistent with descriptive knowledge of the system?
Dimensional consistency – are all equations dimensionally correct without the use of parameters that have no real world counterpart?
Parameter verification – are parameters consistent with descriptive and numerical knowledge?
Extreme conditions – does each equation make sense even when its inputs take on extreme values?
10.5.1. Boundary Adequacy and Structure Verification Tests Applied to a Simple Soap Model
The stock and flow network shown in Figure 10.7 was sketched during the first meeting with the management team. It is a simple conceptual model that was subsequently refined and improved, but it is worth careful consideration as a transitional object leading ultimately to better understanding of the market. Notice there are two conceptually separate market segments. At the top of the figure is the bar soap segment, containing bar soap volume for Old English on the left and for competitors on the right. These volumes are represented as stock accumulations to capture the typical inertia of consumer buying habits. Volume lost to other product types is shown as outflows. Volume exchanged through competition is shown as a net flow of bar soap volume between Old English and competitors. Notice also there are no inflows to the two stock accumulations, reflecting an important assumption that the market is mature. So far, so good – this part of the model is compact and captures several realistic features of the bar soap market.
Figure 10.7. Management team's first conceptual model of soap market
Source: From Supporting Strategy: Frameworks, Methods and Models, Edited by Frances O'Brien and Robert Dyson, 2007, © John Wiley & Sons Limited. Reproduced with permission.
At the bottom of the figure is the new market segment for liquid soap in which trials of the new product lead to an accumulation of trial users who then adopt either Old English's liquid soap or competitors' liquid soap. Adoption results in an increasing number of regular users represented by two stock accumulations. Note that managers expected to attract and retain loyal customers in the new premium liquid soap market as there is no flow from regular users of Old English's liquid soap to regular users of competitors' liquid soaps.
Figure 10.7 reveals three interesting weaknesses in the initial conceptualisation of the strategic initiative in liquid soap. First, the management team perceived the bar soap and liquid soap market segments as disconnected from each other, even though when pressed they would acknowledge the markets are linked. Second, users of bar soap were lost to 'somewhere' in the personal care market, through the outflow 'Loss of Bar Soap Volume'. In fact, much of this loss was to shower gel, but since Old English's management had neither a special interest nor the capabilities to compete strongly in shower gel, the slow draining of customers to gels was not clearly recognised or at least its cumulative effect was thought to be small. This blind spot may have influenced Old English's subsequent innovation. Third, the market for the new product was believed to be a one-off simple adoption process. Old English's management would convince bar soap consumers to trial the new product. These potential consumers, in the stock labelled 'Trial Users of Liquid Soap', would remain an uncertainty until they decided to adopt Old English's or a competitor's liquid soap. The strategic problem would be solved for Old English when trial users became regular users of liquid soap, protected by first-mover advantage – since management believed that Old English had a significant head-start in the technology required to produce liquid soap. In other words, the strategic intent was to contain competitors in the bar soap segment while the company built its leadership in the liquid soap segment. Then, liquid soap users would remain isolated from competitors' actions because of first-mover advantage.
Although this diagram was a useful start, and reflects the way the management team first talked about the product launch, it fails both the structure verification and boundary adequacy tests. The portrayal of rivals is inconsistent with management's own descriptive knowledge of the soap market. In fact, there are two main branded competitors (Old English with 26% market share and Global Personal Care with 31% market share). In addition there are supermarkets that collectively hold 16% market share with private label soaps. Together these rival groupings account for most of the market in value terms. However, Global Personal Care and supermarkets cannot simply be aggregated into a monolithic rival, because they are very different businesses with distinct goals and objectives.
Besides an overly-simplistic representation of rivals the conceptual model does not endogenously represent the flow rates that determine sales and market share of liquid soap. When these drivers are added to the model the interdependencies in the business become more apparent (such as cannibalism of bar soap by liquid soap, the effect of capacity on cost and price, the interplay between marketing and performance targets and the impact of rivals). An appreciation of these interdependencies sharpened management's understanding of the realistic growth prospects of the liquid soap business. All this factual information about rivals and internal operations was well known by the management team but was missing from the initial conceptual model.
10.5.2. A Refined View of the Market
After a presentation of the initial model and discussion with senior managers, their specific knowledge of competitors was incorporated into the modified model of the market shown in Figure 10.8. The three distinct rivals are now clearly represented, each with bar soap and liquid soap products. On the left is Old English with its established product 'Traditional English Bar Soap' and its new product 'Antibacterial Liquid Soap'. On the right is Global Personal Care with its established product 'Moisturising Bar Soap' and its new product 'Creamy Liquid Soap'. In the middle are supermarkets with their established product 'Me Too Bar Soap' and their new product 'Me Too Liquid Soap'.
Figure 10.8. Refined conceptual model of soap market
Source: From Supporting Strategy: Frameworks, Methods and Models, Edited by Frances O'Brien and Robert Dyson, 2007, © John Wiley & Sons Limited. Reproduced with permission.
This new picture illustrates structure verification and also the kind of learning that happens during model conceptualisation. It reveals more clearly than before the complex situation facing the main players in the bar soap market. Each player has to balance the attractiveness of their established products, taking account of three different forces simultaneously influencing their customers. The first force is traditional inter-firm rivalry in the bar soap market from consumer promotions and advertising, aimed at maintaining market share in bar soap. This rivalry is depicted in the flow rates between the three bar soap stocks in the bottom half of Figure 10.8. For example the bi-flow labelled 'Branded Bar Soap Users Switch Rate' represents competition in bar soaps between Old English and Global Personal Care. Similarly, the bi-flow labelled 'Price Sensitive Traditional English Bar Soap Users Switch Rate' represents competition in bar soaps between Old English and supermarkets.
The second force influencing customers is the development of the liquid soap market represented as the set of three stocks in the top half of Figure 10.8 ('Antibacterial Liquid Soap Volume', 'Me Too Liquid Soap Volume' and 'Creamy Liquid Soap Volume') and the corresponding adoption rates. The third force is the attractiveness of shower gels – a substitute product – represented as outflows from bar soap volume in Figure 10.8. For example, the outflow 'Traditional English Bar Soap to Shower Gel Substitution Rate' represents the loss of Old English bar soap volume to shower gel. Interestingly, the refined view of the market still implies a simple and incomplete perception by management of the shower gel market. Shower gel is effectively outside the model boundary and remained that way because the management team was not expecting to develop a presence in the market.
Since this is a mature market with a high level of penetration, there are no inflows to increase total volume. In other words, the development of the market is essentially a zero-sum game between brands and varieties – and in soaps this game is played against the backdrop of gradual volume loss to shower gels. While managers' intention at Old English was to move users from bar soaps into liquid soaps (and focus groups suggested that bar soap users would indeed adopt liquid soaps) they nevertheless faced a dynamically complex problem. The company needed to persuade bar soap users to adopt liquid soap without losing market share, while also improving profitability and avoiding costly price wars. Essentially they were managing a growth business (liquid soap) alongside a declining business (bar soap) against strong and diverse rivals.
10.5.3. Boundary Adequacy and Sector Map of the Complete Soap Industry Model
The refined market sector was an improvement on the first conceptual model, but it still does not pass the boundary adequacy test since it says nothing at all about the internal operations of the rival firms. Nevertheless, it is an important piece in the jigsaw because it tells the modeller a great deal about the shapes of the other missing pieces. In fact, the market sector became literally the centrepiece of a fully endogenous model whose boundary included the operational factors that control the movement of sales volume between the three main rivals and between bar soap, liquid soap and shower gel.
A sector map of the complete model is shown in Figure 10.9. The market sector is in the centre and includes formulations for consumer behaviour to represent both the adoption of liquid soap and response to promotions and advertising for liquid soap and for bar soap. On the left are two sectors representing Old English. The company is conceptually divided into bar soap and liquid soap operations each containing formulations for changes in production capacity, pricing, gross margin, promotions, advertising and sales targets. A similar approach is taken to representing Global Personal Care in the two sectors on the right. At the bottom of the diagram are two sectors representing supermarkets, again divided into bar soap and liquid soap, each sector containing formulations for pricing and share of display shelf.
Figure 10.9. Sector map of the soap industry model
I should stress that the conceptual separation of bar soap and liquid soap in the two branded manufacturers and in the supermarkets is a modelling convenience – a way of interpreting the complexity of the business that makes sense for the problem at hand. The separation does not imply that bar soap and liquid soap were organised as two distinct divisions; far from it. In Old English, a single factory produced both bar and liquid soaps, and product pricing was handled centrally as were promotions and advertising. These six sectors interact with each other and with the market sector in the centre to yield a rich feedback structure indicated by the arrows in Figure 10.9.
10.5.4. Managerial Decision-making Processes in the Old English Bar Soap Company
The majority of the model was devoted to representing managerial decision-making processes inside the three rival organisations that, directly or indirectly, influence consumer behaviour. These decision-making processes control changes in production capacity, pricing, promotions, advertising and the management of display shelf – the factors shown in the sector map. The corresponding formulations run to more than one hundred equations. They were constructed from concepts and facts gathered in many hours of meetings and sketched on a diagram occupying 12 A4 pages. There is not the space here to document each and every equation in detail. Instead, a brief verbal description of the formulations is presented below that gives a feel for the range of concepts included within the boundary of the model.
Production capacity is the responsibility of the manufacturing manager. The technology of liquid soap production is entirely different from bar soap and requires new equipment. Hence, the manufacturing manager faced a strategic dilemma about how quickly to build capacity for the new product and how quickly to retire capacity for the old product. The decision-making process for changes in production capacity is essentially driven by market size. The larger the expected sales volume the more capacity is needed and vice versa. Economies of scale are important too. The greater capacity, the lower unit cost and the lower price (at a given margin), leading to more sales and eventually to more capacity. The same process also works in reverse. When sales fall, cost per unit increases due to a combination of low capacity utilisation, high fixed cost and fewer scale economies. As a result, retail price increases too, unless the firm reduces gross margin to maintain sales. The model captures the interplay of manufacturing cost dynamics arising from the growth of liquid soap capacity and the simultaneous decline of bar soap capacity.
Promotions and advertising are adjusted to achieve a sales performance target. In Old English, the target is past sales. So the logic of the decision-making process is as follows. If current sales volume is much less than past sales volume then the company increases promotions (by offering bigger price discounts) or initiates new advertising campaigns to boost sales. On the other hand, if current sales volume is much greater than past sales volume then the company reduces promotions and advertising in order to dampen sales and also to improve operating cash flows. Small differences between current and past volumes tend to be ignored. The management response function is shown in Figure 10.10 and was calibrated by comparing observed volume changes with the historical behaviour of retail prices and the intensity of advertising campaigns. Paradoxically, this formulation of promotions and advertising implies that Old English's managers ignore competitors' actions when assessing which marketing actions to take. Instead, they focus myopically on their own sales volumes rather than benchmarking against Global Personal Care or supermarkets.
Figure 10.10. Function determining the strength of Old English's competitive response to market performance
Source: From Supporting Strategy: Frameworks, Methods and Models, Edited by Frances O'Brien and Robert Dyson, 2007, © John Wiley & Sons Limited. Reproduced with permission.
10.5.5. Managerial Decision-making Processes in Global Personal Care
The management team felt that Global Personal Care's decision-making processes were broadly similar to Old English. Therefore, Global Personal Care was modelled by replicating the formulations for Old English while modifying information flows or parameters to capture important differences of managerial emphasis. For example, it was assumed that Global Personal Care focuses its competitive actions on managing market share rather than sales volume. Hence, in Global Personal Care, promotions and advertising increase when market share falls below its historic value. Sales volume plays no significant role. Similarly, Global Personal Care's adjustment to mark-up or gross margin is formulated as a function of long-term market share instead of sales volume. It was also assumed that Global Personal Care offers a slightly higher trade margin than Old English in order to obtain adequate share of shelf space despite lower market share.
10.5.6. Managerial Decision-making Processes in Supermarkets
Supermarkets' pricing is much different than Old English and Global Personal Care for a number of reasons. First, supermarkets do not aspire to be market leaders. Rather they participate in the market enough to bargain effectively with existing branded manufacturers. Second, supermarkets do not manufacture or own capacity. Instead, they buy from manufacturers that specialise in private-label products. Third, supermarkets do not promote their product through advertising. They compete on price only.
Supermarkets' pricing is intended to boost income from display shelf. The decision rule for supermarkets' pricing is influenced by trade margin received and by product sales – the two main sources of retailers' income. The income received from branded products in the form of trade margin is compared with the historical trade margin. If income from branded products falls, either as result of a reduction in branded manufacturers' trade margin or market share, then supermarket managers reduce own-label retail price for two reasons. First, they want to expand supermarket sales to substitute for income lost from manufacturers. Second, they want to force an improvement in the trade margin. However, as supermarkets expand their market share, the income from branded products will decline even more (if manufacturers of branded products do not offer higher trade margins), and supermarkets will further reduce their prices.
An extreme outcome of this interaction between manufacturers and retailers is that supermarkets will dominate the market through continuous price reductions (as happened with Wal-Mart in some market segments of fast-moving consumer goods). Pricing decisions that respond to income from trade margin are embedded in a reinforcing feedback loop in which price spirals downwards. Although there is a lower limit to price, it depends on the sourcing cost of supermarket products and the actual trade margin obtained from branded manufacturers. However, supermarket managers usually prefer to set a target market share that is low enough to maintain bargaining power without pushing branded manufacturers out of the market. This policy introduces an additional balancing feedback loop that halts the spiral decline in price.
Display shelf is negotiated between retailers and branded manufacturers. Share of the display shelf is fiercely contested in fast-moving consumer goods, no matter how large or small the store. While big stores can offer lots of shelf space, it is easily filled by the huge proliferation of available products, thereby improving the bargaining position of retailers. The task of branded manufacturers' sales managers is to negotiate a significant share of display shelf at low cost in order to enhance daily sales and to increase the effectiveness of advertising campaigns. Conversely, retailers' management teams try to maximise the income received for allocated space by assigning the greatest share to the most profitable items. The decision-making process for changes in display shelf depends on trade margin and market share. The more market share or the greater trade margin then the larger the display shelf. This decision logic is embedded in a reinforcing feedback loop where the more shelf space, the greater sales volume, the higher market share and the more shelf space. If unchecked, this reinforcing loop enables branded manufacturers to dominate the market. However, retailers use private label products to retain some control.
That concludes the review of the model's main sectors and the underlying conceptual model. It should be clear that there were many opportunities to test the consistency of the model structure with descriptive knowledge of the business.