Register or Login To Download This Patent As A PDF
|United States Patent Application
Kiefer, Nicholas M.
May 22, 2003
Method for site selection for retail and restaurant chains
A method of selecting a site for one or more retail units uses a database
of information on existing units in a chain, including store sales or
profits, site characteristics available from various databases, and
information on marketing expenditure in the relevant market. Sales of
potential new units are forecast abstracting from advertising efforts.
Thus, potential profit characteristics of a site can be evaluated without
regard to existing advertising programs. This corrects a major problem
with existing approaches; site decisions can be made simultaneously with
marketing allocation decisions.
Kiefer, Nicholas M.; (Ithaca, NY)
CLARK & BRODY
1750 K Street, NW
November 13, 2002|
|Current U.S. Class:
|Class at Publication:
What is claimed is:
1. A method of selecting a site for one or more a retail units comprising:
a) providing a database of information, the information comprising a
number of variables for a number of proposed sites;, the variables
including a binary advertising allocation variable and a non-linear
advertising allocation variable; b) for each proposed site, regressing
sales over a selected time of period on the site variables, the
regressing step further comprising at least one of: (i) holding constant
the effect of the advertising across the proposed sites during the
regression step to show the relationship between sales and the
non-advertising variables; and (ii) accounting for the effect of the
advertising across the proposed sited during the regression step to show
the relationship between sales and advertising.
2. The method of claim 1, wherein the relationship is displayed in a map
for each proposed site.
3. The method of claim 1, wherein profits are calculated from the sales,
the relationship between profits and advertising or non-advertising
variables being shown.
4. The method of claim 1, wherein the regression step generates
coefficients for each variable, and the coefficients for the advertising
variables are adjusted to not exceed a maximum value.
5. The method of claim 1, wherein a number of retail units exist in one or
more of the proposed sites, and an average sales figure per existing
retail unit for each site is used as another variable in the regression.
6. In a method of site selection for one or more retail units wherein site
characteristics are analyzed to determine where one or more retail units
should be located, the improvement comprising controlling for advertising
allocation as part of the analysis.
7. The method of claim 6, wherein advertising is held constant for the
8. The method of claim 7, wherein advertising is varied for the analysis.
9. The method of claim 8, wherein the advertising effect is compared to
sales or profitability.
10. The method of claim 7, wherein site characteristics without the effect
of advertising are compared to sales or profitability.
 This application claims priority under 35 USC 119(e) based on
provisional patent application No. 60/331,215 filed on Nov. 13, 2001.
FIELD OF THE INVENTION
 The present invention is directed to a method for selecting a site
for a store such as a retail unit or restaurant, and in particular, to a
method that parcels out the effect of advertising so that pure
characteristics of the site can be used to determine whether the site
should be selected, and one that allows for site selection based on the
level of advertising.
 In the prior art, retail industry decision-making is becoming more
sophisticated in nature, relying upon scientific methodologies and
quantitative measures. Site selection is one of the most crucial
decisions made in the chain retail (including especially restaurants)
environment as it is capital intensive, has serious long term marketing
implications, and is critical to corporate--franchisee relations. While
there are a number of site selection tools
in the prior art, the
presently available tools
forecast sales based on the sales of existing
sites, and do not explicitly incorporate the effect of advertising into
the site selection process. The problem with this approach is locating a
new site in an area populated with other successful stores may not be
solely related to site characteristics. A site selection finding that a
geographic site has high sales may be a result of high advertising
expenditures at that site. Consequently, putting a store at that site may
only be successful if the same level of advertising is used. This flaw in
prior art site selection
tools is a result of the inability of the prior
tools to isolate the effects of advertising so that this effect can
be removed from the site selection process.
 Accordingly, a need exists to provide site selection
tools that are
capable making site selection analyses without the effects of advertising
skewing the site selection results. The present invention solves this
need by providing a method which allows for site selection without the
effects of advertising so that business owners have a more accurate
picture of site characteristics without the effects of advertising.
SUMMARY OF THE INVENTION
 It is a first object of the present invention to provide a method
of site selection for retail units.
 Another object of the invention is a method of site selection which
parcels out advertising effects to better focus on other site
characteristics for site selection purposes.
 Still another object of the invention is a method that is suitable
for sites that include existing stores as well as sites that do not have
 One other objective is a model that "controls" statistically for
advertising, allowing comparison of sites with the effects of advertising
held constant. The model also allows "what if" sales forecasts with
different hypothetical advertising allocations.
 Other objects and advantages of the present invention will become
apparent as a description thereof proceeds.
 The primary objective of the inventive method described is to
identify, with a high degree of certainty, whether or not a site is
suitable for development, the volume of business that can be expected,
and the general degree of profitability to be expected at the site. The
associated site-selection method compiles and analyzes data and presents
findings in a user-friendly manner, allowing the user to adjust for "what
BRIEF DESCRIPTION OF THE DRAWINGS
 Reference is now made to the drawings of the invention wherein:
 FIG. 1 is a graphical representation comparing target rating points
to incremental sales; and
 FIG. 2 is a map showing profitability of different geographic areas
when advertising is held constant at a national average; and
 FIG. 3 is a map showing profitability of different geographic areas
at actual advertising levels.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
 The invention involves a number of steps in order to arrive at an
output that informs the business owner whether a particular site is
desirable for development. A first mode of the invention addresses
situations where other stores exist in the site being investigated for
further development. A second mode addresses the situation where there
are no other stores in the site.
 A first step in either mode is assembling of a database. The
business owner or client supplies a sample of stores, or if possible the
population of its stores, together with location information (latitude
and longitude; these can be calculated from the address if necessary) and
whatever store-specific information the system has. Often this includes
store characteristics (size, model), and perhaps some local demographics.
Many if not all of the variables are fixed items which are out of the
control of the business owner.
 When using the first mode where stores exist in the site being
investigated, store sales (or profit, or an alternative performance
measure) over some period, preferably but not necessarily a year, must be
included. A measure of advertising should also be included, typically
target rating points or TRPs. Often these are available only for "market
areas," (DMA's). This is appropriate, as the advertising in a DMA affects
sales for all stores in the DMA. These variables are augmented by census
information available routinely. Data at the block group, tract and zip
level are typically used. This data is available, for example in the
commercial product MapPoint (MS), which is used, though many other
mapping companies offer mapping capabilities augmented with census data.
As the client desires, this data can be augmented with local business
statistics available from Claritas or other sources. More detail leads to
higher explanatory power in the regressions and hence better forecasts
and better site selection.
 With this data at hand, a DMA average sales per store for each DMA
containing 4 or more stores is constructed. While four stores are used,
this number can be adjusted up or down if need be. Of course, for many
chains there will be many DMA's without stores and this situation is
addressed in the second mode of the invention.
 The next step involves a regression analysis and calculations using
typical regression programs that are widely available. More specifically,
the sales for each store are regressed on the assembled data including
the DMA average for the sample of stores in DMA's with a reliable DMA
 As part of the analysis, insignificant variables are dropped. This
step may require some experimentation, but in general those variables
showing the lowest t-statistic after regression is completed are dropped.
This leaves a grouping of remaining variables, which all have
t-statistics generally greater than 1.5 in absolute value works. Of
course, other values could be used as the cutoff to define which
variables should be considered as part of the site selection.
 Once the insignificant variables are dropped, an equation for
forecasting sales for sites in DMA's having stores already there is
produced. In other words, sales could be predicted using the coefficients
generated by the regression for each variable left in the equation.
 In the second mode of the invention, sales data is not available
since the business does not have stores in the site being investigated.
Therefore, all stores in the sample are used, and sales are regressed on
the variable list, excluding the DMA average variable. After again
selecting significant variables and dropping insignificant variables as
done for the first mode, a forecasting equation is generated that is
relevant to sites in DMA's where the chain does not have accumulated
 The interpretation of the forecasting is done in two ways. One way
involves predictions for a site on the basis of site characteristics
alone. A second way involves the situation where there are already stores
in the market, and their average sales are also used to predict sales at
the new store.
 Forecasts for either mode are reported along with the regression
standard error and probabilities associated with sales intervals e.g.,
Sales will be 1.0 to 1.25 million$/year with a probability 0.33). The
interval of a year can vary, and be selected according to a client's
 Another important aspect of the invention is accounting for or
controlling for advertising. As noted above, if advertising is not
accounted for, site selection may be based, at least in part on
advertising allocation and give a misleading prediction as to store
 Referring again to the second mode and the second regression
described above wherein DMA average data is excluded, the regression
includes variables controlling for the TRP or advertising allocation
within the DMA. The regression uses three variables for the control; a
dummy variable indicating whether the TRP allocation is zero, a variable
indicating the level of TRPs, and a variable TRP-squared reflecting
hypothesized TRPs. The point of the specification is use of a non-linear
specification for the hypothesized TRPs so that its effect is allowed to
be nonlinear. If the effect was linear then the logical conclusion is
that all advertising expenditure should be concentrated at the same
place, and such would not a useful recommendation for any business.
 The purpose of the suggested specification is to estimate the
profitability of a site while holding constant the effects of
advertising. Using the formula described above, the effect of advertising
is zero up to some critical level, then positive. The formula for
adjusting sales forecasts for incremental TRP level is: if the current
TRP level at a site is zero, subtract the coefficient of the dummy
variable indicating TRP zero, then add the estimated quadratic effect of
the hypothesized level of TRPs. If the current level is nonzero,
calculate the effect of TRPs at the current level, then subtract this
amount from the calculated effect of the hypothesized TRP level. In
effect, this formula is parceling out the effects of existing advertising
in the sales forecast, either in situations where there is no advertising
or a level of advertising at some defined level.
 With the formula, one can pick hypothetical TRPs and engage in
"what if" analysis with respect to the effect on sales when the TRP is
 The formula is adjusted so that "way out of sample" values are not
exaggerated. In particular, negative effects at low, unobserved levels
are excluded (set to zero) and effects at values greater than 1.25 times
the maximum value observed in the sample are set constant at the
estimated effect at 1.25 times the maximum value. The 1.25 level can be
adjusted according to the client and the client's statistical
 The specification described above is illustrated in the FIG. 1
below. Here, the observed range of TRP in the sample is a through b. The
graph shows that up to a certain point, TRP has no effect on incremental
sales (the coefficients are negative or low) and is therefore zero. Then,
incremental sales increase dramatically until a certain point wherein
additional TRP does not increase sales. It should be noted that the
maximum value adjustment described above is reflected in the leveling off
of the curve.
 The specification or function is advantageous since it will allow a
business owner to plug in TRPs for a site and forecast sales,
independently of existing advertising effects. 1
 This technique offers advantages over current site selection
analysis. That is, it is often the case that, if a chain is not
interested in varying its advertising allocation, it decides that it is
best to locate in areas where the advertising level is already high.
However, this does not take into account the situation where it may be
better to change the advertising allocation. Using the inventive method,
a better decision can be made by considering the advertising allocation
jointly with the site selection process. The inventive model allows
management to consider the potential profit from opening additional
stores in an area and adjusting the advertising allocation appropriately.
 In practice, the presentation of the results of practicing the
inventive method is preferably driven by a Visual Basic front-end program
linking with mapping software. One preferred example is MS MapPoint
because it is convenient, and it has Active-X links. However, other
mapping software could also be used. The front-end program could also be
linked with financial performance display screen. The financial
performance screen takes the sales forecasts given by the model and
user-inputted site cost variables to provide multiple proforma profit and
loss projections, break-even analyses, and net present value calculations
(at user-chosen capitalization rates). A final observation relative to
probable unit profitability will be made in this module. Default values
are provided for key financial ratios using industry average values (for
the restaurant industry these are assembled by Delolitte and Touche for
the National Restaurant Association). It should be understood that once
the sales forecast is made using the specification, any number of
techniques can be used to show profitability based on the other data
related to the business. Since generating a profitability in terms of a
forecast of sales per year and a probability is within the skill on the
art, a further description thereof is not deemed necessary for
understanding of the invention.
 To illustrate, we show two schematic maps, Maps 1 and 2, for a
hypothetical chain in FIGS. 2 and 3. Here, a number of geographical
regions A-G are displayed in terms of zip codes. The regions are rated in
terms of profitability by number, with those rated 1 being the most
profitable, those regions rated 2 being less profitable, and so on with
the least profitable site being rated 5. The first map shows
profitability when advertising is held constant at national average
levels. In terms of the regression, Map 1 of FIG. 2 takes into account
the site characteristics but holds the advertising constant. Using this
model, any sites anywhere in the country that show the effect due to site
characteristics will exhibit similar levels of profitability. This map
isolates the effects of site characteristics on profitability, the site
characteristics in regions A, E, and G contribute to better
 The second map in FIG. 3 shows profitability at actual advertising
levels. In this example, the regression is done using actual advertising
allocations rather than the average allocation as used to generate Map 1
of FIG. 2.
 Comparing the two maps of FIGS. 2 and 3, it is apparent that
profitability is higher when holding advertising constant at national
levels. This information would be generated using the specification and a
set level of advertising for each store. Looking at Map 1 alone, the
analysis identifies the high profitability areas for site selection
purposes. Map 1 says that regions numbered A, E, and G are the best
candidates for site selection based on national advertising levels.
 However, the invention also has the ability to look at the effects
of actual advertising, which can be much more informative when making
site selection. Only actual advertising is used to determine
profitability as shown in Map 2. The levels of profitability in Map 2 as
compared to Map 1 drops significantly, with just region E showing a "1"
rating. This indicates that the actual advertising is below the average.
Map 2 also enables the identification of sites where the profitability is
still high, i.e., region E, but that not many locations will be
profitable at the actual levels.
 To relate Map 1 to Map 2, one could perform a series of "what ifs"
in terms of actual advertising and ultimately arrive at the profitability
shown in Map 1 by continuing to increase the advertising until it would
match the national average. Likewise, starting at an advertising
allocation equivalent to the average and continually reducing the amount
would result in a profitability of Map 2.
 Again, Map 2 is beneficial in indicating that there are areas that
need advertising input to improve profitability such as C, D, and F. Map
2 allows you to make decisions about site selection in instances where
advertising dollars can be allocated. If advertising dollars are
available, then sites showing higher profitability when average
advertising is used may be good candidates for increased advertising.
 The two maps in comparison demonstrate that actual advertising
levels are low compared to the national average and this is reflected in
the profitability calculations. In addition, taking into account just
actual advertising as shown in Map 2, development in the area is less
attractive with the current levels of advertising. Development would be
worthwhile is additional advertising investment is made; the area looks
quite attractive if additional advertising dollars can be allocated, see
Map 1. This hypothetical example illustrates the importance of
controlling for advertising in site evaluation.
 As noted above, another advantage of the invention is that the TRP
can be forecast and the regression run to determine what effects may
occur, the "what if" approach. Referring again to the graph above,
plugging in estimates for TRP can generate an estimate as to incremental
sales, and give one an idea of where they are on the curve for purposes
 The mapping software is widely available and is only used for
presenting the results of the patented method and any type available and
known in the art are suitable for the invention.
 Another advantage in using TRPs in the evaluation of site selection
is that TRPs are in the control of the business owner whereas site
characteristics are typically fixed variables. Thus, while site
characteristics can be isolated to determine their relative effects on
profitability, advertising can also be controlled for site selection.
While restaurants have been exemplified, any type of a retail unit can be
sited with the invention.
 As such, an invention has been disclosed in terms of preferred
embodiments thereof which fulfills each and every one of the objects of
the present invention as set forth above and provides a new and improved
method for site selection of retail units.
 Of course, various changes, modifications and alterations from the
teachings of the present invention may be contemplated by those skilled
in the art without departing from the intended spirit and scope thereof.
It is intended that the present invention only be limited by the terms of
the appended claims.
* * * * *