energing approaches to retail outlet management … approaches to retail outlet management l. lilien...
TRANSCRIPT
ENE RGING APPROACHES TO
RETAI" OUT"ET MANAGEMENT
Ga ry L . Eilien
Ambar G . Rao
New York Univers ity
"P 882- 76 Novemb er 19 76
EMERGING APPROACHES TO
RETAIL OUTLET MANAGEMENT
L . Lil ien
M O I Q T .
Ambar G . R30
New York Univers ity
"P 882- 76 November 19 76
Abstract
Ma j or problems faced by the mul ti- outlet retailer include those of
strategic planning , tacti cal deci sions and operational control . The
strategic planning que stions include : How many outlets should be bui lt
in the next Y years" In what cities should they be bui lt" When"
Tactical decis ion s concern the sele ction of a particul ar si te : What is
the potential of a particular s ite" What are the characteristi cs that
lead to the succes s of a particular s ite " What will be the impact of an
outlet on ne ighboring outlets" Operational control procedure s include
exception reporting , dete ction of new trends , planning and monitoring
of advertis ing and promotional expenditure s .
Thi s paper reviews emerging quantitative approaches to retai l out
let management and shows how a few key concepts can be and have been used
in a variety of product areas .
0731 908
Introduction
In many industries , products are offered to consumers through company
control led retail outlets . Some example s are supermarket cha ins , gaso
l ine , banks , and fast foods , where the outlets are supermarkets , servic e
stations , branch banks and franchised restaurants respective ly . As in other
industries , the problems of retai l outlet management can be c las sified
into three groups strateg ic planning , tactical dec isions and operational
control .
The strategic planning questions are
How many outlets should be built in the next Y year s"
In what c ities (markets ) should they be built"
When"
The tactical dec is ions conc ern the se lection of specific s ite s
What i s the potential of a spec if ic s ite"
What is the impac t of changes in the environment on s ite potentia l , e . g .
construction of a motel on ga s station sales"
What is the impact of an outl et on a neighboring outlet"
Operational control procedures inc lude
exception reporting
detection b‘
f"
he;"
fifen'
a‘s
planning and monitoring of advertis ing and promotional expenditures .
In .the past much of the quantitative work has focused on tactical
problems , particularly on s ite evaluation ( see Green and Appelbaum [5 ] for a brie f
review ) .
Strategic is sues have large ly been handled in an informa l way . Recently ,
some re searchers have recognized that s ite potential is re lated to the
product or company image ( see Stanley and Sewal l [ 1 7 ] but they do not
indi cate how ima ge can be changed in a way that increase s potential .
The purposeof this paper i s to des cribe some emerging quantitative
approache s to retai l outlet management , and to show how a few key concepts
can be and have been suc ces s ful ly used in a variety of product areas . The
paper does not exhaustively review the literature in re tai l outlet manage
ment , but rather , highlights the important ideas that underl ie problems in
thi s area .
S trategic Planning
Conventional market share analys i s suggests that , afte r a bui lding plan
is implemented , marke t share wi l l equal outlet share . I f this were the case ,
the plann ing de cis ion rule s would be s imple : build where the next uni t of
outlet ( equal to marke t ) share brings the highest profit (perhaps in Ne t
Present Value te rms ) , continue bui lding unti l e ither budge t is exhausted ,
or additional share doe s not produce acceptable re turns .
Li fe i s rarely thi s s imple as seve ral authors have pointed out ( see
Kotler pp . 7 86 Several reasons have been suggested for why
market share does not seem to increase l inearly with outlet share , including
inequal ities in promotional and outle t" qual ity effe ctivene s s among companies ,
some vague economie s- of - s cale argument , e tc . , but a simple explanation for
this behavior is stil l the sub j e ct of debate ( see Lilien [1 2 ] for one analys is ) .
The important observation , however , i s that a ngnlinea r , generally S- shaped
relationship exists between outlet share and marke t share , making the question
of how many outlets to bui ld , and in what markets , cons iderably more complex
(Figure 1 illustrate s this re lationship ) .
Market Share
Outlet Share
FIGURE 1
et Share " Outlet Share Re lationship
A procedure has been developed and implemented to he lp management
make both the se dec is ions . It can help make better deci sions even if those
dec is ions are not always Optimal "
( in the narrow , mathematical sense ) .
And these " vaguely- right (Davidson [ 131) answers are , in fact , accurate
enough for a practi cal , planning s ituation .
The implications of the S - shaped relationship are important . From
the point of view of maximiz ing incrementa l market share , a firm would
only build in markets where the firm already had an establ ished pos ition .
A ma j or o il company took explic it account of this re lationship in its
entry strategy into a new state dur ing the late sixties . A dec is ion was
made not to enter unles s outlet share could be immediately built up to a
reasonable ( S% leve l ; i . e . , where marginal market share returns for
adding additional stations were reasonable . Severa l local d istributors
were acquired and an instant 5 % outlet share was developed on which a
building program was then based . Bui lding up from scratch was recognized
a s an unprof itable s trategy . Thus when a company expands geographically ,
the strategy should be to build up share quickly in each new market
before proceeding to the next . S imultaneous expans ion into a large number
of new markets is l ikely to be unsucces sful , unles s f inanc ial and managerial
constraints are non " existent .
Even after understanding this market share " outlet share relationship ,
however , the development of plans are not straightforward .
Using the S -Curve as Part of a Planning System
Suppose that a firm has empirical ly developed an S- curve for its
markets and now wants to know how many outlets to build in each of a large
number of markets during a several ( say , five ) year planning period .
Generally the first year re sults become budget items building funds
are al located in accordance with plan " year 1 The following year
results are used to prepare profit plan proj ec tions and to help allocate
outlet- s ite procurement funds ( in anticipation of building ) .
Observe the nature of the managerial dec is ion process : all planned
outlets may not always be built due to changing local building
codes , construction difficulties , lack of sites , etc . I f an extra , choice
s ite becomes available in a des irable area , an outlet may be constructed
on it immediately , even if no money was originally allocated . What
management needs to know is whether to build five outlets or twenty outlets
in a given area . The different between ( say ) five and six outlets is
un important as it often washes out during implementation .
Before looking for a good , or best plan , let us examine how to
determine the value of a particular , known , plan . XYZ Company ' s plan for
ma rket A may be to build three outlets this year and two in each of the
rest of the years of the planning horizon . Let us assume that the firm
ha s a forecast of competitive activity (building plans for all firms other
than XYZ ) as well as market volume and volume growth rate , current and
forecast prices and margins , cost of land , etc . Thus XYZ has ava ilable
all the information needed to make an obj ective financ ial evaluation of
this situation as fol lows
"tep_ l : For each year of the planning period calculate XYZ s\ pla ns
and industry plans .
Step 2 : Use the S - curve relationship to find the outlet share and
thus the assoc iated market shar e (Mst) for each year t .
Step 3: Multiply Mstby the proj ected volume in the market to get
XYZ's volume , (Volt ) .
Step 4 : Multiply Volt by the proj ected gross margin (Mart ) to get
Gros s Revenue (GRevt) .
Step 5 : Adj ust Gros s Revenue by investment , debit factors to get
a cash flow as soc iated with each year : CFt
.
Final Step : Determine the net present value (NPV ) assoc iated with the
plan as
T
Zcpt
t=1 ( 1 -R)t " l
where T is the planning horizon and R is the discount rate .
Now , i f every outlet , ( in NPV terms ) were worth less than the one
built be fore it , the best building plan would be the one which first built
P"AN TOTA"
Tota l Build ing Constra int A
TAB"E 1 : ExamplegAlloc a tion Proc edure Results
Note that here , i f an incremental ana lys i s were used ( for Total Bui lding
Constraint 2 ) outlet numbe r 2 would be bui lt in Marke t B . This would yield
a total NPV of 25 instead of the bes t case , 30.
The procedure above looks not only at the value of s ingle outlets but
of groups of outlets to assess the ir profitabi lity . The actual me chanics
of the procedure , especial ly with bui lding plans over time , are somewhat
te chni cal and are treated e l sewhere ( see Li lien and Rao The important
point is that in practice the procedure has proved e ffic ient , and easy to
use , producing re sul ts that are optimal or close to optimal . The procedure
was run on a 1 70-market , five- year planning problem cons idering outlets .
The allocation procedure took le s s than a minute on an IBM 360- 7 5 ; including
input " output and NPV calculations the entire procedure ran in we ll under
five minute s . This e ffi ciency is important for allowing update runs and
sens itivity analyses at low cost .
This system has been used as an aid in outlet building planning at a
ma j or U . S . Corporation s ince 1 9 6 9 . The procedure was used in the following
wa y
- 8
( 1 ) To generate allocation plans , given a set o f assumptions and
inputs ;
( 2 ) To test the allocation against change s in those as sumptions .
Step 2 is cr itical in implementation and use . Suppose the model
al location doesn ' t change with proj ected pos s ible var iation in land costs
but is very sens itive to variations in profit margin proj ections . Then
have your analysts spend most or all of the ir time firming up the margin
proj ection and forget about the land cost figure s .
(3) To aid in determining the overal l bui lding allocation .
Outlets are a n investment and firms have alternative use s for the ir
building capital . Then as many outlets should be buil t as still return
a pos itive NPV when discounted at the firm ' s cut- off or internal rate of
return . Thus , no overall building constraint i s needed rather , the
procedure should shut of f when the incrementa l NPV for the last outlet
planned becomes zero or negative . The discount factor is another important
impact on the over " al l building l evel as we ll as the allocation .
The model did not replace or transcend the manager ; rather it helped
provide more meaningful inputs and thus more useful outputs . The managers
are involved at each step of shaping the final results and by be ing able
to control the proces s they grow to trust it , Managers learn a great deal
about the ir own dec is ion s ituation and how it interacts with the firm ' s
problems . They become more secure in the ir own j udgments as well .
To summa rize ,the key idea here i s that deve lopment of e ffi cient
bui lding plans requires explicit re cognition of the S~curve relationship ,
i .e . the impact of the new bui lding upon the entire structure of the mar
ke t . Use of this re lationship leads to significant improvements in retai l
outlet building policie s .
Site Se le ction and Site Potential
A second key problem faced by the multi- outlet retai ler is , given the
de cision to locate in a market area or city , where the outle t should be
placed . Approaches vary . Applebaum [ 2 ] reports that 10% of a sample of
1 70 large retail chains did no systemati c analys is for location of re tail
outlets . That same study had re search expenditure s varying wide ly , with
the average research per new location be ing about 1 % of the s ite- inve stment
cost .
When Eastern Shopping Centers appraise s a s ite , that location is sub
jected to a searching analys is covering current populations population
trends , current and per capita income of the area , competing centers or re
tailers , road patterns ,
" etc . In contrast , a drug chain appraises a site
" j ust by taking a ride around a particular area , talking to some of the people
living there and getting a fee l for the s ite ' s expans ion capabil ities .
"
(Duncan a n d Phil lips
What seems called for is a structure for analyz ing outle t s i te potential .
A variety of di fferent mode ls have been suggested to aid in the evaluation
and measurement of this potential , several of which are reviewed be low .
The se model s share a common underlying structure , modi fied or customized
for the parti cular bus ines s or purchase s ituation . This structure can be
summarized as
( l ) Site Potential Local Sales Component
Transient Sales Component
Re lationship ( 1 ) says that the sale s potential at a parti cular s ite
has two separate components : those sales derived from people who live
Obtain importance we ights of these attributes from consumers
wj average importance weight of a ttrubute j .
One key attr ibute to include is brand- image or market presence , l inking this
model . with the S " curve model in the previous section .
Now estimate
x Area Gasoline x Fraction of
Potential Sales bought
local ly .
The other component , f is cal culated s imilarly , with a roa eg2:
replac ing the local trading area , and traffic count data replacing Area
Gasol ine Potential
The development for fast foods is es sential ly equivalent , with con
sumption habit data replac ing gasoline usage . Funct ional forms other than
have been used as wel l , with no s igni ficant improvement in predictive
power .
Much has been written on the art a n d science of superma rket s ite poten
tial . Green a n d . ApPleb a um [ 55 ] review the re levant literature . We highlight
here some important model- deve lopments .
Applebaum [ 11 ] des cribes a procedure cal led an analogue approach ,
based on the as sumption that the drawing power of a site wil l be close to
that of other stores of the chain under s imilar conditions . A maj or diff
culty is the identification of " s imi lar conditions .
Huff and e limi nate s thi s problem by using the following mode l
- 1 2 _
1 1n
ZSkSTJk
t
k=l
Probabil ity of a customer in area i shopping at
reta il location j
Square feet of retai l sell ing area of location j
Driving time from area i to reta il location j
Parameters as soc iated with se ll ing area and driving
time respectively .
This model postulate s that patronage i s positively re lated to the
s ize (and , hence , merchandise assortment ) of the out t and inverse ly
related to the di stance from the store to home . Huff shows that At
varie s with merchandise be ing sought , and , hence reasonable search
effort .
Stanley and Sewel l [ I7 ] modi fy the Huff model by replacing s ize
with an image variable , developed as the distance from the chain pos ition
to an " ideal " chain pos ition us ing multi- dimens ional scal ing . They
report s igni ficant improvements in the predictive abil ity of the model .
The Stanley and Sewall procedure seems to be an important improvement
but does not suggest how to improve image .
Returning to the S - curve idea , it appears that image might we ll be
re lated to outlet share in a nonlinear way . Further work in thi s area
would clearly be he lpful in making this concept operational . In addition ,
a transient component could c learly be introduced to e stimate potential of
a s ite in a shopping mall . In thi s case , the tran sient componentn could be
related to the volume of foot traffic attracted to the mal l for purposes
other than grocery shopping . Again , further work seems j ustified .
Hlavac and Little [6 ] des cribe a procedure for estimating s ite poten
tial for automobile dealerships . The structure of the model is as follows
Hlavac and Little as sume that consumer selection of a dealer i s related to
dealer location and customer pre ference for a particular auto-make , and i s
relative to the attraction of al l other dealers .
Cons ider one particular dealer , and spl it the region into a set o f
areas , i 1 , I . Then de fine
9 1pul l of the dealer on a buyer in geographic segment or area i ;
hi
intrins ic pul l of a dealer , independent of make (related
primari ly to ease of acces sabi lity ) ;
qi
make pre ference for the brand for buyer s in segment i .
Then the pul l of the de aler for buyers in geographi c segment i is de fined as
9 1hi
qi
and letting pi
the purchase probabil ity for a buyer in area i then
(where j re fers to the dealer , o to the dealer o f interest )
Pul l of dealer x make pre ference
Pul l of dealer x make preference
al ldealers
The authors po stulate an exponential drop off in dealer pul l with distance
and equate make pre ference with brand share . A procedure i s deve loped to
estimate the parameters of the dealer pull function and the mode l i s shown
to fit data quite wel l .
- 1 5_
high turnover area
turnover area
Time
FIGURE 3
Branch Bank Growth to S ite Potential
- l6
There are many multi - outlet retai ling industries not touched here
inc luding giant retailers , hote ls , and specialty item chains but the results
de scribed are representative of the current state of knowledge an d are indica
tive of what is be ing used . The main point of the review in this section
is to suggest that structuring the si te evaluation problem as Potential
local transient sales seems we l l adapted to problems in the retai l outlet
industry , and should be exploited .
A key opportunity in thi s area is the integration of S - curve e ffects
into site evaluation . The re sults from the previous section suggest the
power of evaluating the tota l market impact of a building plan . That total
impact wi l l not be the sum of the e f fe ct of individual sites , by themse lves ,
due to S- curve synergy . Thus , there is a maj or opportunity to improve these
mode ls by including this synergi stic e ffect of current market pos ition in
evaluating building and divestment plans . Then a total market impact would
be estimated and the e ffect on marke t profitabil ity more accurate ly gauged .
-1 7
Operational Control
Most multi - outlet retailers and , in fact , most commercial organ i zations
have sales and profit reporting systems . These systems report on various
measures of performance such as sales , cost , profits , etc . and compare what
happened in the current period with past performance . Such a comparison
impl icitly assume s that thi s period should be like the last period or should
show some amount of growth gene rally naive and uninformative as sumptions .
To be meaningful and actionable , however , such comparisons must be
made to what should have happened . In order to say what should have happened
a mode l is required , linking control procedure s to spe cific , quantitative
predictions of what performance should be or is expected to be like .
Modeling approaches to strategi c planning and s ite location evaluation
have been described in the previous se ctions o f this paper . An operationa l
control procedure can be bui lt around these mode ls . Cons ider a marke t area
in which a site expans ion p lan has been implemented . Us ing the mode l ,
predictions o f sales e ither in total or by product type in each period can
be deve loped . Thes e predictions ne ce ssari ly use planned values for company
activities and e stimate s of competitive activity in the market . The predic
tions can be disp layed in tabular form ( Figure 4 ) and in graphical form
(Figure The solid line in Figure 5 shows the his torical sale s in the
market , the dotted line the mode l fore casts .
In addition to the predictions , an estimate of fore cast error is also
deve loped . This e stimate ca n either be obtained dire ctly from the mode l or
observed from the performance of the mode l against actual sale s in the past .
Now suppose a spec i fic ma rketing plan is implemented . Actual sales
for a period are reported , as are the actual values of company and competi
tive activitie s . Actual sales are compared to sales predi cted by the model .
20
I f the di fference between actual and predicted exceeds the previous ly determined
forecast error- l imi t , an exception report is issued . This report triggers
a " cause analys is " to determine why the exception occurred . (This i s often
referred to as a marketing audit
There are several pos s ible circums tance s that could have led to an
exception
( a ) company and competitive activities differed from as sumptions
used in the development of predictions . Management can investigate the
reasons for such di fferences , and deve lop administrative procedure s to
ensure greater conformity to plans .
(b ) a new type of activity ( ei ther company or competitive ) not con
si d ered in the mode l structured oc curred in the market . For example , a
nove l type of promotion may have been introduced . In such cases , the mode l
needs to be enriche d by including a factor representing the new activity .
( c ) actual activities were in line with planned and e stimated activities .
In such case s the mode l itsel f might be at fault . It needs to be re- examined .
In this case , the control procedure reports on itse l f and suggests that it be
improved .
Figure 6 shows some important types of exceptions that can occur . The
shaded area around the fore cast indi cate s the range of variability to be
expected . As forecasts of periods further away in time are made , this range
increase s . Exceptions 4 , 5 , and 6 in the figure are of particular interest .
Exception 4 shows a si tuation where an increase in sale s in one period
Unsucce ss ful promotions often pre sent such sale s profiles . Exceptions 5 and
6 are early indi cators of deve loping trends .
Figure 7 shows a convenient way to display the exception information ,
and re late period sales , and cumulative sales to date , to predictions .
DATA BEHAVIOR
- 2 1_
FIGURE 6
UPDATE STATUS TREND ANALYS IS
STATUS REPORT
IN CONTROL
UPDATE EXCEPTION
LAST POINT EXCEPTION ,
RETURN TO NORMAL TREND
UPDATE AND LAST POINT
BOTH EXCEPTIONS
UPDATE AND LAST POINT
BOTH EXCEPTIONS ,
POSS IBLE NEW TREND
CUMULATIVE EXCEPTION
NE" TREND
- 23
Systems such as that outlined above have been implemented by the authors
in severa l organizations . Yorke [ 18 ] gives details o f one system that has
proven to be managerially use ful . Rao and Shapiro [ 13] deve lop some new ,
sophi sticated forecasting methods that have proven to be use ful in several
forecasting systems . Ra o and Lilien [ 1 4 ] show how the e ffe cts of promotion
can be incorporated in a forecasting system to improve fore cas ting accuracy
a n d to be tter as ses s the re lative impact of promotional programs . Rao and
Mi l ler [ 1 5 ] de s cribe some powerful methods of assess ing the impact of product
adverti s ing on sales and indicate how these methods , too ; can be incorporated
into a forecasting and control system .
These deve lopments are indi cative o f the state of the art . Systems
and procedure s are currently available for ope rational reporting and control
of retai l outlet performance which inte l ligently integrate mode ls and data
into managerially use ful information . They invariably reduce the volume o f
reports , highlight important information , and indicate potential problems
and opportunitie s in a timely , ye t routine fashion . They integrate mode l
building research into day to day operational control , and spec ify the
parameters of a use ful , usable management information system , and associated
d ata base .
-2 4
Conc lusions
Our aim in this paper has been se le ctive rather than exhaus tive . We
po int out that quantitative approaches to problems of retail outlet manage
ment are emerging which are of use in problems of strategic and tactical
planning as we l l as operational control . underlying these approache s are
a few key ideas
1 . Outlet share is generally related to share o f market in a nonlinear
way . This re lationship bears important impact on the deve lopment
of strategic bui lding plans as wel l as in the evaluation of retail
s ite Ope rations .
Structuring the problem of evaluating s ite potential as
potential local component trans ient component
has been quite succe s s ful and should be exploited .
The heart of a good operational control procedure i s a se t of
accurate fore casting model s l inked to exception reporting capa
b ilit ies . Succe s s ful mode l s have been deve loped which can great
ly improve the operation o f many such systems .
There i s much need for new research in th is field . However , there is
much more need for integration and implementation of existing methodology .
Many new tools a nd concepts are now avai lable ; the chal lenge to the multi
outlet retai ler is to make use of them .
- 2 5_
REFERENCES
Applebaum , Wil liam . Me thods for Determining S tore Trade Areas ,
Research , Volume 3 (May
Applebaum , Wi lliam .
" Survey of Store Location by Retai l Chains ,
in Guide to Store Location Re search , Curt Kornb la u (ed ) (Reading ,
MA : Addi son- Wes ley ,
Davidson , Sidney , As I See It , Forbe s (April 1 ,
and Me thods (Homewood , IL : Richard D . Irwin ,
Green , Howard L . and Wil l iam Applebaum .
" The Status o f Compute r
Applications to Store Location Research ,
" paper presented at the 7 lst
Annual Meeting o f the As sociation o f American Geographers , Mi lwaukee
Wis consin (April 2 3,
Hlavac , Theodore E . J r . , and John D . C . l ittle . A Geographic Mode l
Hall ,
Huff , David L . Determination of Inter " Urban Retail Trade Areas
(Los Ange le s : Univers ity of Cal i fornia , Real Estate Research Program ,
1 9 6 2 )
Huf f , David L .
" A Probabil istic Analys i s of Consumer Spatial Behavior ,
in Emerging Concepts in Marketing , Will iam S . Decker , (e d . ) ( Chicago
American Marketing Association ,
Huff , David L .
" Defining and Estimating a Trade Area , Journal of
Marketing , Vol . 2 8 (July
Kotler , Phi lip . Marketing Management : Analysi s , Planning a n d Contro l ,
2nd edition (Englewood Cl iffs : Prentice Hal l ,
Li l ien , Gary L . and Ambar G . Rao .
" A Mode l for Allocating Retai l
Outlet Building Re source s Acros s Market Areas ,
" Operations Research ,
volume 2 4 , Number 1 (January- February
"i l ien ,Gary L . A Market Share- Outlet Share Mode l ,
"
Working draft ,
Septembe r 1 9 7 6 .
Ra o ,Ambar G . and Arthur Shapiro ,
" Adaptive Smoothing Us ing Evolutionary
Spe ctra,
" Management Science , Volume 1 7 , Number 3 (November
