GEOGRAPHY  AND  STRATEGIC  PLANNING

To place retail location into the context of business practice, let’s start with the broadest possible view on strategic planning.

Strategy is the process of

• identifying organization-specific assets,
• understanding the organization’s environment, and then
• deciding how to deploy or augment assets to earn the greatest return.

Corporate strategy

Question 1:  What business(es) should we be operate?  In what business environments are our core assets most valuable?

Question 2:  What should we be doing internally, versus hiring out or not being involved with at all?

This is often implemented through decisions to enter and exit industries, by acquisition/sale or growth/closure.

This is often implemented through vertical or horizontal integration or disintegration.

How do we compete in this line of business?

This is often answered in terms of

• product breadth,
• target market, and
• quality/price

What are the best ways for us to serve our target markets with our target products?

Geography enters all three levels of strategy:

CORPORATE: Are we a US company, a N.Am. company, or a global company?

BUSINESS: Do we have the assets to compete on the basis of worldwide low costs?

FUNCTIONAL: Where should we obtain financing, source inputs, locate production, and locate distribution or outlets? Given a production location, what technology and personnel policies work best here? Given a retail location, what product mix and price points work best here?
 

However, location decisions are most often made explicitly at the level of functional strategy. (So if you want to become CEO, you need to gain experience and expertise in additional contexts; geography is generally a means to an end in an organization).

Retail location is generally an element of functional strategy, but it must be congruent to the overall competitive strategy..
 
 

RETAIL  LOCATION  PROBLEMS

There are several distinct (but related) retail location problems, each of which must be approached through a series of steps.

1.  What regions are our markets?

This is the market-selection scale of retail location decision making [Thrall et al., April 1998 ("Steps One and Two");  Ghosh & McLafferty Ch. 3].

At the international scale, this generally reflects explicit corporate strategy.

Within a national or world-regional market, it’s a question of logistics and opportunity.
 

2.  How many retail outlets do we need to serve our chosen or desired market?

The first answer to this depends on the basic economic-geographic tradeoff between economies of scale and friction of distance.  These concepts can be related to the way that central-place theory uses the concepts range and threshold. [Be able to describe this tradeoff, and the conditions under which each concern gains or loses importance].

An additional answer depends on our relative concern for cannibalization versus eliminating competition.  [Be able to describe these terms and to think about this tradeoff].
 

3.  Where to locate each retail outlet?

This question is typically and reasonably asked at two distinct scales:

1. in what part(s) of the market?  (Ghosh & McLafferty refer to this scale as areal analysis)
2. on what site:  in what parcel or shopping center?  (Ghosh & McLafferty refer to this scale as site evaluation)
Analogous to this might be a 3.b. Where to close retail outlets? Locations are not permanent: marketing strategies change, as do geographic environments.

4.  What product mix is optimal for a given retail outlet?

A. Non-systematic processes (see Birkin et al., pp.63-64)

1. "gut feeling" or "environmental observation"
2. locational imitation of competitors (or near-competitors)

First question: how would you assume the size of the market area?

• conventional wisdom in that type of retailing; general rules of thumb (see the rules of thumb in Thrall et al. ("Retail Analysis, Step Four," June 1999), p. 39)
• survey of current customers: what radius or drive time would encompass a certain percentage of all customers? (But would this be the same in a denser, less-dense, or less auto-prone region?)

B.1.  Look for generally desirable attributes of market areas
Ghosh and McLafferty mention generally published data on the household income, consumer expenditures, business growth, or population growth by metropolitan area or broadcast media market.  Regardless of your product, high-income and growing regions are generally better prospects for new retail operations.

B.2. Analog technique (see Birkin et al., pp. 65-66)

Look for market-area characteristics that are similar to the market areas of successful, analogous stores (yours or competitors’). (Note that while Jones & Simmons Table 11.1 lists the Analog technique under procedures that assume penetrable (shared) market areas, their "market profiling" procedure is closest to the Analog technique as we describe it in this class).

How would you identify what characteristics to use?

Guesswork:  what market-area characteristics should be important for your product?

Discriminant analysis:  analyze what characteristics (of market areas and of stores -- their product mix, management, size) discriminate between your most-successful and least-successful locations.

Multiple regression:  model the level of profitability (or revenues) across all sites (perhaps sales per square foot) as a function of a set of characteristics, and use the location-specific variables that are most significant. If your model includes non-geographic variables (age or training or experience of managers, age of the store, years since last renovation, etc.), then you have begun to control for some of the things that make stores not analogous to one another.  Note that you could use your resultant regression coefficients to predict the level of profitability or revenues from each alternative format and location proposed for a new store.  (See Birkin et al., pp. 66-67)

In any case, you have to have data on sales by existing outlets. What if you have no existing outlets?

B.3. Market-area analysis (see Birkin et al., pp. 68-73)

Look for market areas (the size of which you guess as noted above) that have generally desirable characteristics, building these market areas up from small-area data.  What’s "desirable" depends on your product and your marketing strategy.

This requires:

geodemographic data (data on the median or average economic and demographic characteristics of inhabitants within small geographic areas), or

lifestyle data (data on the location and buying habits of individuals), or

geo-lifestyle data (data that draws inference about the buying habits of the inhabitants of small geographic areas).

(Note that ESRI, for example, offers a "Business Analyst" extension for ArcView that comes with private value-added geodemographic and geo-lifestyle data; this extension is quite expensive, because of the expense of these data).

1. Actual market areas are not "yes/no" delineations.  (In Jones & Simmons' terms, they are not spatial monopolies). There are always some customers from outside the primary market area, and customers near the "edge" of the market area are less likely to use your location than more proximate customers.
2. Actual market areas depend not only on your characteristics, but on your competition’s locations and characteristics. This clearly will not be the same for each of your locations.

B.4.  Consider the degree of competition within different market areas:

To begin to address the second caveat above, we might at least characterize market areas with respect to the degree of competition present.

See Birkin et al., p. 74; Jones & Simmons, pp. 354-361.  Jones & Simmons refer to these types of analysis as assuming market penetration or dispersed markets -- what's the difference?

How much competition exists already within a market area?  Here's an approach to estimating the saturation index (SI) for market area i.

where P is a measure of population, E is a measure of per capita expenditure (on consumer items in general, or on your particular product category), and R is a measure of the amount of retail space (or space devoted to your particular product category), each within the market area
where max [R / (P E )] is the maximum value of R / (P E) that can be achieved in any market area.

Note:  what simple indicator of market penetration do Thrall, del Valle, and Hinzmann use in the September 1998 article ("Retail Analysis, Step Five")?

C. Modeling the market area based on competitors’ locations

In this section, we'll try a different approach to dealing with the second caveat above.  If a retail outlet exists in an isolated context (as opposed to an agglomeration of similar retailers), it may be assumed to have a primary market area that extends to the primary market areas of its major competitors. How large is this market area likely to be?

Reilly’s law of retail gravitation is an old formulation, based on the analogy of gravity, that suggests that

• the attraction of a consumer to a given retailer is
• directly proportional to [what I will call] the "quality" of the retailer, and
• inversely proportional to the distance to the retailer.

(Compare Jones & Simmons Eq. 9.2, page 308;  Thrall & del Valle's article "The Reilly Model").
The most common measure of "quality" is the size, in square feet of retail space. [What alternative or additional measures could there be?]

Distance, of course, can use any number of metrics. [Such as?]

We can compare the attractiveness of competing retailers (or shopping centers, or central places) in this way. The breaking point in the primary market areas between two competing (and spatially separate) retailers (or centers), expressed as a distance from center A, is

The square root is a simplification, assuming that the distance decay function (a ) is a 2.

We can make use of Reilly’s law to generate polygons whose sides are the locus of breaking points between our outlet (or center) and competing outlets (or centers) -- see Chapter 11 of Jones & Simmons;  see Thrall and del Valle "The Reilly Model".  A GIS can be programmed to do this very quickly, given the locations and sizes of all the outlets or centers.

One can then analyze and compare the profitability of alternative, empirically derived market areas, through any of the market-area analysis approaches outlined above.
 

D. Estimating the number of customers from each possible location

Geography is more than discrete "market areas" that we can distinguish by characteristics.  Geography is also continuous.  Some of your customers will come from any and everywhere.  (Jones & Simmons refer to this as the market penetration approach).  How can we model this? How can we affect this?
 

D.1.  Customer spotting
We can simply define customer spotting as observing and mapping the origins of actual customers.  (See Jones & Simmons, p. 355).  What are potential data sources?

• in-store surveys
• point-of-sale query (e.g., ZIP Code)
• bank check information
• redeemed coupons that had been mailed with the customer's address
• license-plate survey

1. Divide the entire region into zones (Census tracts, block groups, etc.).
2. Compute the ratio of the number of observed customers to the number of potential customers (population?  number of households?) in each zone;  this is your market penetration by zone.
3. Identify and map the zones with market penetrations that meet particular thresholds (60%, 25%, 10%, for example).
4. Relate the market penetration in each zone to characteristics of the zone, such as distance from the store, per capita income, etc., using multiple regression (see Jones & Simmons Table 11.3).  Now you have some idea of what influences people from far away to come to your store.  Should you encourage this by increasing your store's attractiveness to them?  Should you open a store where these zones are concentrated?

where
I = interaction
k = scaling factor that related the amount of I proportionately to P and d,
P = measures of the amount of demand at the origin and supply at the destinations
D = a measure of the cost of getting between i and j
a = a measure of impedance (i.e, does the importance of transport cost vary as the cost is greater or less?)

where
S = interaction
A = the reciprocal of the combined attractiveness of all destinations j, each weighted by its distance from i
O = demand at i
W = attractiveness of j
f(c) = cost of getting from i to j, expressed as some function of the distance

A_i = 1 / S_j W_j f(c_ij) is just a way of controlling the equation so that the S_ij bears some proportional relationship to the total amount of interaction (i.e., consumer demand) in the system. Thus, it’s a particular form of k.

Let’s use the terms of the original gravity model, if we may:

The amount of sales you expect at j from any i

• increases with the attractiveness (e.g., square footage) of j,
• increased with the amount of demand at i,
• decreases with the distance (or cost of distance) between them,
• decreases with the amount of other destinations j, each competing destination weighted negatively by its distance from i.

The Huff model expresses the same thing in terms of probabilities:

where R is a measure of retail attractiveness (directly analogous to W in equations above), such as the retail floor space in each retail store or shopping center.

We can attempt to increase P(C_ij) by increasing R_j relative to S_j R_j -- by reducing price, or by increasing sales area, selection, or service.
 

D.3.  Geodemographic marketing
This approach to marketing recognizes that some part of the retailer's market may be highly fragmented geographically (what Jones and Simmons refer to as dispersed markets);  customers respond strongly, across various distances, to some specialized attribute of the retailer.

While the location of the retailer may not be affected, geographically targeted promotions (e.g., direct mail) may be arranged on the basis of geodemographic data:  targeting individuals by targeting neighborhoods (see relevant notes).

We can use geographic analog reasoning for this:  (1) customer spotting to determine the sources of current customers;  (2) identifying the generalized characteristics of the neighborhoods or zones housing concentrations of current customers;  (3) identifying other neighborhoods or zones, with similar generalized characteristics, to target.

REFERENCES

Birkin, M., G.Clarke, M.Clarke, and A.Wilson.  1996. Intelligent GIS: Location Decisions and Strategic Planning.  New York:  John Wiley & Sons.

Ghosh, A. and S.McLafferty.  1987.  Location Strategies for Retail and Service Firms.  Lexington, Mass.:  Lexington Books.

Jones, K. and J.Simmons.  1990.  The Retail Environment.  London:  Routledge.

Thrall, G.I. and J.C.del Valle.  1997.  The Calculation of Retail Market Areas: The Reilly Model.  GeoInfoSystems 7(4): 46-49.

Thrall, G.I., J.C.del Valle, and G. Hinzmann.  1998.  Retail location analysis, step four: identify situation targets.  GeoInfoSystems 8(6): 38-43.

Thrall, G.I., J.C.del Valle, and G. Hinzmann.  1998.  Retail location analysis, step five:  assessmarket penetration.  GeoInfoSystems 8(9): 46-50.