1. Analyze the Distribution Channel Objectives

Distribution channel objectives must be coordinated with the overall corporate mission, and objectives, and with the marketing plan.

2. Analyze Market and Channel Attributes

The appeal of a market can be evaluated looking at several dimensions that define the competitive environment. This is an area rich in modeling opportunities. Dimensions, which could be included in a model, include:

3. Specify Channel Goals

Channel goals must be specified to identify the specific activities that are to be performed by the prospective channel. A modeling effort that includes measured or managerial judgment of the effectiveness of various channels in meeting company goals could facilitate this step.

4. Specify Channel Structure Alternatives

Channel structure strategy must also be specified. Channel structure refers to the number of intermediaries that may operate within the distribution channel, moving goods from the manufacturer to the purchaser. Channel structure strategy involves specification of the number of different channels to use, and the number and type of intermediaries to use in each of the different channels. This is an area highly adaptable to an optimization modeling procedure.

5. Evaluate Channels and Make a Selection:

Finally, the selection of a channel structure strategy is made by carefully synthesizing thee components of the strategic planning process:

While firms undoubtedly use informal evaluation procedures for these areas, these structural aspects of a distribution strategy could readily be incorporated in a formal evaluation and determination model.

One key area in distribution modeling is distribution logistics -- that is, (1) the selection of the market, (2) the identification of the optimal number of outlets, and (3) the selection of site locations for the outlets. Logistical models for physical distribution provide managers with assistance in these three major areas. Because each area influences the efficiency of the distribution channel, they are discussed below.

Market selection is the process of choosing markets most appropriate for products, and is most often based on a simple estimate of market potential. Sometimes the market potential is determined for the total market and sometimes for geographic sub-markets (territories) within the total market. Many measures of market potential are found in the literature, including those of competitive structure, profit potential, net present value, and market opportunity.

Estimates of the total market potential are often calculated using "market breakdown" methods, in which the total market is broken into its component parts. Fundamentally, examining the average purchase quantity per buyer and the number of buyers does this. This relationship is usually expressed as:

where:

N = Number of buyers in the market

The difficulty in estimating total market potential lies in making an estimate of N -- the total number of buyers in the market. A different formulation sidesteps this problem, but decreases the accuracy of the estimate:

where:

This formulation is not sensitive to variations in the number of buyers or to variations in purchase quantities within purchaser/ non-purchaser segments, or other segments of the total market.

Instead of using these market breakdown methods, more precise estimates of market potential can be made by using the "market-buildup" method. This approach estimates sales potential for each targeted territory within the total market. The potential for targeted territories is then summed to produce an estimate of sales potential for the total market.

While the above method is most useful for consumer markets, for industrial markets market buildup methods may be applied by using geographic market information reported by Standard Industrial Classification (SIC) codes. One application of the market buildup method uses SIC codes to estimate market potential within a given geographic area.

IPR, a Boston based company, currently provides country specific estimates of worldwide computer and software sales for companies like IBM, Microsoft and Compaq. IPR's economists model the process using both break-down and build-up approaches to reach a common market estimate. The differences between the two approaches are reconciled to obtain what is believed to be a more accurate estimate.

Closely related to the market-buildup method is the market indexing method employed in Sales and Marketing Management magazine's "Survey of Buying Power". The survey of buying power estimates the relative buying power of a total area by weighting characteristics of consumer markets. While the "Survey of Buying Power" reports an expected buying power for each metropolitan region in the United States, it also encourages firms to develop customized indices for their particular products. Thus a firm marketing Digital Cameras might use multiple regression analysis to develop an index such as,

BP = .5Y₁ + .3R₁ + .2P₁

where:

For example, if San Diego has 0.86 percent of the population in the United States, 0.94 percent of the nation's income, and 1.20 percent of the nation's retail sales, San Diego would have a Digital Camera buying-power index of:

This indicates that 0.952 percent of Digital Camera sales in the United States might be expected to come from San Diego. The "Survey of Buying Power" employs a multiple-factor indexing system that is based on three independent variables. Multiple factor indexing systems are predictive formulas developed as additive linear model using multiple regression analysis. In this case, the weights used in the buying power index are simply the regression coefficients previously determined to predict buying power.

Once the potential of markets is determined, the number and location of wholesale and retail outlets to be used in the channel must be considered. But this consideration is only relevant when changing the number of outlets will make a difference to the firm. Hartung and Fisher and Lilien and Rao have shown that differential advantage of adding an outlet is not linear with the number of outlets. Like many relationships in marketing, the relationship between the number of outlets and market share is S-shaped.

Hartung and Fisher modeled purchases as though a single customer made them. Their model focused on the notion that a buyer could either "purchase our brand" or "purchase another brand". For example, using their model to assist in strategy development for Shinex (our spray-on floor wax), would result in the following market share prediction:

where

Hartung and Fisher have reported that the constants which estimate purchase probabilities in a linear model sense ( P(Sx|Not Sx) and P(Sx|Sx) ) are 4.44 and .64, respectively, across products they have studied. When these constants are multiplied by the outlet share and input to the formula (1 + C₂ - C₁ ), an index number is produced to show if more outlets should be added.

That the relationship between the number of outlets and market share is non-linear, this suggests that market efficiencies exist. But the main problem of determining an optimum number of outlets is still not solved. This problem is a classification issue that is solved by a statistical technique called cluster analysis.

The classification problem of determining how many outlets should serve a market has a variety of solutions. But at present we lack a way to optimize any of these solutions. That is, we can not be confident that the recommended number of outlets (or number of clusters of customers around a specified number of outlets) is in reality the best number.

The two conflicting objectives of classification analysis are (1) determine locations that are maximally distinct from one another, or (2) to determine locations which concentrate on customers that are maximally similar. Defining the problem in this way can result in the curious conclusion that different numbers of outlets may be equally satisfactory. However, we know that the most correct solution requires determining which solution is most appropriate for a specific problem. The best number of locations depends on a combination of market, economic, competitive, technical and legal-political factors.

Once the product's market potential and the appropriate number of outlets for it have been determined, the next step is selecting the locations for prospective outlets. This, too, can be modeled. Model builders find it a struggle, however, to keep good location models simple. Let us consider some of the simple classical models.

Huff created a model in 1962 that is regarded as the starting point of modern work. But this model was quite limited in its assumptions about patronage motives. It predicted patronage to be a function of two factors: the size of the outlet and the "economic distance" of competing outlets for the customer. This undoubtedly oversimplifies the issues. It ignores a range of patronage motives, including store attributes (image, reputation, product assortment, price, etc.), and customer characteristics (demographic, life style, and socioeconomic variables). However, we will first consider the Huff model and then a more current model explaining retail patronage.

The Huff model is based on the assumption that the probability of a consumer using a given outlet or site is equal to that site's share of the consumer's total utility (the sum of the utilities for using each of the possible sites). The model can be expressed in terms of the attractiveness of a retail outlet at location j for a consumer located in market area i. The attractiveness of a site is directly proportional to the size of the retail center and inversely proportional to the customer's distance from the center. That is,

where

The Huff model was developed to represent geographic areas within the Midwestern United States, where cities are geographically separate, with distance between them. More modern applications of this model have been directed at refining customer motivation variables and incorporating them into the model.

One such model suggests that most site location solutions are inadequate because they focus on cost minimization rather than long run profit maximization. This model, called the Beaumont, is summarized in the technical appendix at the end of this chapter. Its solution to the location problem has two important benefits. First, it uses Euclidean distance measures to represent cost structures.

Second, it parameterizes some qualitative consumer evaluations that influence demand. The most important limitation of this model is the coarseness of the consumer evaluation variable. As noted by Gautchi, there is a variety of consumer motivators which must be included to improve the predictive validity of models.

A second major area for distribution modeling is inventory management, naturally also important in production modeling. The management of physical inventory revolves around meeting some management-approved level of customer service, while minimizing inventory cost. Once again, conflicts occur between economic criteria and other criteria.

Economic order quantity (EOQ) formulas exist to determine ordering practices that minimize the overall cost of inventory. The problem in balancing inventory costs and service levels focuses on the difficulty of predicting demand and avoiding an out-of-stock condition. A stockout results in immediate lost sales and possible long-term decline of the customer base.

You will find two EQQ models at the web site for this text, a basic (and traditional) EOQ model, and a discrete time period EOQ model.

The traditional EOQ model requires the balancing of ordering costs and carrying costs. Ordering costs include all expenses involved in placing an order. Total order costs increase with the total number of orders. Carrying costs increase with the average level of inventory. Lower levels of inventory result in lower carrying costs, but in higher order costs due to the larger number of orders that must be placed.

A number of formulations for EOQ may be used to minimize the total inventory carrying/ordering cost function. The most basic model is,

Where

The basic EOQ model may be expanded to include levels of safety stock and base stock. Safety stock is the additional units of inventory that must be stocked to serve peak demand fluctuations, and base stock is the minimum amount of inventory necessary to serve typical demand levels. Let us look at a simple EOQ model that also considers demand over time.

The discrete-time-period EOQ model is the second EOQ model found at the web site

http://marketing.byu.edu. This model treats the time periods as discrete opportunities for placing orders. (For a discussion of the Excel macro that drives this model, see the more detailed discussion in the technical appendix at the end of this chapter).

This model adapts the standard EOQ model (for constant demand) to a variable demand situation. It requires the calculation of an optimum EOQ, which is based on an assumption of constant demand, and tracks demand as it accumulates to the point that optimal EOQ is reached. The model then adjusts order size to be as close as possible to the calculated EOQ amount, which is exactly equal to the total amount demanded for several time periods. The discrete time period EOQ model gives a good approximation of the optimal order schedule as long as demand is relatively constant over time.

Figure 6-1 shows a screen appropriate to the above explanation, from the discrete period EOQ 1-2-3 model on the web site. Because the model requires that the quantity ordered be as close to the EOQ amount as possible, 170 units are purchased at the start of month 1, thus providing the beginning inventory that meets the optimal EOQ. As a result of monthly demand of 95, 50, and then 25 units, the 170 units of beginning inventory is depleted in month 3. This process continues as further orders are made in months 4, 6, 7, and 10.

Inventory control is a key element in physical distribution. The elements of physical distribution models considered in this chapter bring into consideration the geographic and time related questions relevant in responding to the market. Market evaluation and physical facility evaluation models, as well as inventory models are planning models are planning models critical in adapting to changes in market structure.

Load Excel, with a backup copy of your template on the hard drive of your computer. When the blank worksheet appears on your computer screen, use the File Open command to open the EOQ file.

After a moment, the EOQ analysis template will load into your computer, and will display the first of its two command menus on the screen:

SIMPLE INPUT EXECUTE PRINT QUIT

These EOQ worksheet commands are invoked by selecting the desired option and clicking on OK.  If you wanted to go into the "SIMPLE" EOQ model, you would select "SIMPLE" and that EOQ model sequence would be invoked, with the alternate command menu being displayed.  Each command in the command menu sequence invokes a predefined macro sequence that performs a specific set of functions. 

The simple EOQ model command menu is:

DISCRETE INPUT EXECUTE PRINT QUIT

Each of the commands included in the primary command menu is discussed below.

While the default model when the worksheet is first loaded is the discrete time period model, invoking "SIMPLE" switches to the simple EOQ model.

Invoking "INPUT" accesses a secondary command menu:

DISCRETE SIMPLE QUIT

Invoking this command causes the model to process the input data and calculate new output data.

For either model, this command will print the output data, as shown in Figures 6-1. (Before running this command, be sure your printer is loaded with paper and is on line.)

Choosing "QUIT" from the main command menu exits the main menu of the EOQ model and allows you to explore the worksheet. Type <ALT>M to re-enter the EOQ menu, or /Q to exit Excel. If you have entered data into the model which you wish to save, be certain to save the model first, and be sure to use a file name other than "EOQ" to avoid overwriting the original "EOQ" file).

The nature of distribution and EOQ models makes them particularly appealing for quantitative model development. Compared with other marketing areas, they are quite easily parameterized and their output, while suffering from some of the same validity questions that plague other models, is easy to interpret.

In this chapter we have reviewed a few popular approaches to modeling economic order quantity. In particular, we have examined both a simple EOQ model and a model over time. While both are useful as conceptual illustrations of determining economic order quantity, they are also potentially useful as freestanding models to do just what they contend to do: to model economic order quantity.

With respect to distribution decisions, we need more complete models, incorporating all aspects of distribution functions, including interactions with other marketing functions. We need models with emphasis on channels decisions (focused on the economic and power criteria discussed early in this chapter), on multi-product distribution decisions, on risk considerations.

Much work has been done in these areas, but much yet remains to be done.

Cluster Analysis. Cluster analysis is concerned with the classification, or grouping, of sets of similar items. In the case of our sales force example, our interest is in finding similar groups of accounts. The objective is to group these similar accounts, based on a set of descriptive variables, in such a way that we can understand the accounts better while still maintaining most of the distinctions between the accounts.

While discussion of the specific algorithms is beyond the scope of this text, a brief discussion will be found in most advanced marketing research or analysis texts. 9

 Gravity Models

The Huff Model. The gravity model developed by Huff is useful for estimating the size of trading areas. The basic Huff model was developed to estimate probability of purchase, but has been modified to be equally applicable in estimating the number of customers, the available purchase potential (expenditure levels), and shopping frequency.

Thus we see that there is a 40 percent probability that customers from market area 1 will travel to Hillsdale Shopping Center to buy French pastries.

Basic EOQ Models. The basic EOQ model attempts to minimize the total inventory carrying/ordering cost relationship by the formula,