A NEW ENTRANT S DECISION ON VIRTUAL VERSUS BRICKS-AND-MORTAR RETAILING

Transcription

1 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar NEW ENTRNT S DECISION ON VIRTUL VERSUS BRICKS-ND-MORTR RETILING Glles Renhardt Department o Management DePaul Unversty grenhar@ondor.depaul.edu Moren Lévesue Department o Marketng Poly Studes Case Western Reserve Unversty mxl0@wru.edu BSTRCT Frms retal an nreasng proporton o ther produts on-lne. How muh to sell through the Internet s a deson drven y market onsderaton supply han eeny. In ths paper we use mro-eonoms to derve a rm s on-lne o-lne uantty that est trade-os osts revenue ompettve ehavor. We study the ase o a new rm whh s a monopolst. We show that t may not e optmal or the rm to retal on oth hannels. Indeed the rm must onsder all osts n retalng ts produts. The model s then rened to study the ase o a new entrant ang an already populated market that operates on one or oth hannels. To maxmze prots t may e temptng or the new entrant to retal on-lne the numents are retalng on a rks--mortar network or ve-versa. We show that ths deson depends almost solely on a produt s typology the rm s supply han eeny. We also oserve a new ompettve strategy n the proporton o output one rm sells onlne: as that proporton nreases or the exstng rms t wll derease or the new entrant ve versa. Keywords: E-ommere retal hannels supply han management optmzaton mro-eonoms. Introduton Motvated y the potental o the Internet rms retal an nreasng proporton o ther produts on-lne. The deson o how muh any to sell through the Internet s essentally drven y market onsderaton supply han eeny. In ths paper we use a smple mro-eonom model to derve under a set o regularty assumptons the rm s on-lne o-lne uantty that est trade-os osts revenue ompettve ehavor. We egn y studyng the ase o a new rm entrant whh enters a market as a monopolst wth a radal nnovaton; e.g. a value proposton not yet avalale on the market. The rm sees potental revenues rom a presenstve market dem on oth a rks--mortar an on-lne hannel. We show that ths n tsel s nsuent to ndue a rm to retal on oth hannels. Indeed the rm must onsder supply han osts n retalng ts produts namely manuaturng produton nventory shppng. Many entrants on the market have had serous dulty turnng a prot mostly due to the ndued omplexty o managng produton nventory dstruton when operatng on oth hannels Chopra Van Meghem 000. Falng to dstngush properly manage ts osts on eah hannel has proved at least hallengng not atal to many rms. The model s then rened to study the ase o an entrant wth an nremental nnovaton ang an already populated market that operates on one or oth hannels. Puttng asde ompetton amongst the exstng rms n the market we model the entrant as a new rm ompetng aganst a group o exstng rms whh we ollapse nto a sngle monopoly. Dong so permts the dervaton o useul sometmes ounter-ntutve results. For example to maxmze prots t may e temptng or the entrant to retal on-lne the numents are retalng on a rks-mortar network or ve-versa. We show that ths deson depends almost solely on a produt s typology the rm s supply han eeny. We also oserve a new ompettve strategy n the proporton o output one rm sells on-lne: as that proporton nreases or the exstng rms t wll derease or the new entrant ve versa. We antpate our ontruton rom ths analyss to e manold. Frst a dstntve produt typology dgtalzale vs sold whh s exlusvely dened y the supply han osts eah type nurs when eng retaled on-lne or o-lne. Seond a ontruton to the e-ommere lterature y ouhng the prolem n an analytal ramework that expltly models the dynams o sales revenues aganst supply han osts or a rm that operates Page 36

2 Journal o Eletron Commere Researh VOL. 5 NO on two dstnt retalng hannels on- o-lne. Thrd the solutons derved rom our model unearth prot maxmzng desons ompettve strateges that more tradtonal approahes may overlook agan hely due to the mpat on the new entrant s ottom lne o enterng or nreasng ts market va the Internet. Fourth we target our analyss towards two dstnt lasses o entrants: those wth a radal nnovaton they enter the market as monopolsts those wth an nremental one they enter an already populated market. The remer o ths paper s organzed as ollows. Seton rely revews related lterature whereas seton 3 oers a produt typology. Our results are derved under a set o osts market model ased assumptons whh we dsussed n seton 4. Seton 5 presents the monopoly results seton 6 studes analogous uestons n a duopoly settng. Seton 7 onludes the paper spees dretons or urther researh. Tehnal detals appear n an appendx.. Related lterature When a new entrant s makng the vrtual/rks--mortar deson t must onsder a numer o ssues nludng onsumer preerenes hannel oetves onstrants supply han struture overall value reaton or the rm. New aadem researh on the e-ommere phenomenon has emerged rom most elds o management suh as marketng strategy normaton operatons Homan 000; nupnd Bassok Zemel 00. Numerous studes have nvestgated the role o a vrtual exstene on onsumer shoppng ehavor. Messnger Narasmhan 997 nvestgate why how some market segments preer one retal hannel over others whle Keeney 999 dentes the value that Internet ommere rngs to the ustomer. Kauman- Sarorough Lndust 00 analyze how onsumers use ntegrate the mult-hannel optons n ther shoppng experene y testng how the reueny o on-lne rowsng or shoppng relates to varous lexlty attrutes provded y ths hannel. Shoenahler Gordon 00 propose a model o mult-hannel uyer ehavor to takle the hallenge rought aout y the multplty o hannels the rm aes suggest a onsumer-entr vew rather than a hannel oused vew to work through the hallenges unue to the multhannel marketer. These are onsumer oused papers that analyze the mpat o dgtalzaton on the shoppng experene. Researh has also shown to new entrants the mportane o hannel oetves on makng the hannel hoe. Balasuramanan 998 oers Salop s 979 rular spatal market to apture the ompettve trade-os etween a dret marketer N retalers. The o o that paper are the role o store loaton as a ompettve advantage the mpat on exstng rks--mortar loatons on a dret marketer entrant the role o produt normaton as a ompettve lever. Tsay grawal 00 nvestgate hannel onlts etween a manuaturer a reseller va a stylzed supply han wth ndependent deson makng at oth the manuaturer reseller levels. Vshwanath Mulvn 00 provde evdene that pure-players rms that exlusvely retal on-lne are gradually eng outperormed y retalers who explot more than one hannel ut who do so as omplements to one another y denng dstnt roles or eah hannel y leveragng on- vs o-lne synerges. They state that less than 5% o pure-players are protale today many are n dre strats. Levn Levn Heath 003 also study how a retaler an est utlze multple hannels. Ther study measures the preerenes o onsumers or retalng hannels as they progress through the shoppng experene searhng omparng then purhasng. They also test onsumer s reatons to strateg allanes amongst retalers o on-lne o-lne rs. Whle some have gone eyond a onsumer ous modeled an overall value reaton potental o e-usness e.g. Bakos 00; Devnney Latukeu Mdgley 00; Lukng-Reley Spuler 00 or the eets o dret sales on rms onsumers welare altogether e.g. Hendershott Zhang 00 other sholars have taken the rm s perspetve rather than the onsumer s. These nlude Htt Fre 00 who ompare the value to the rm o ankng ustomers who use on-lne serves rather than the rks--mortar loatons. Leamer Storper 00 have looked at the role o the Internet n nreasng the tness o the dvson o laor automatng ntermedaton oordnaton tasks whereas Dewan Jng Sedmann 000 have examned sellers osts reduton or olletng uyer preerene normaton managng multple pres. Others have looked at rks-lks ntegraton e.g. Gulat Garno 000 ompettve mplatons o undlng normaton goods Bakos Brynolsson 000. Our analyss addresses how an nomng rm s hoe o retal hannel speally o- vs on-lne mpats ts operatonal protalty ompettve advantages. Our model studes these mpats y rst parttonng the rm s ost unton nto supply han omponents produton nventory & dstruton retalng n order to dretly norporate nto ts supply han the ost mpat o a produt s dgtal ontent seond y ntegratng total output proporton dstruted on- o-lne as deson varales. We thus model a new eonomy deson that most retalers must now ae. Page 37

3 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar 3. Produt typology Fgure desres a gener supply han that llustrates the low o goods through a rks--mortar network. Fgure desres the dstruton o the same produt ut through an on-lne retaler s supply han. Frms satsy ther ustomer s orders dretly rom nventory n eet dsntermedate ther retalng atvty whh they ondut ether through ther own network or y sellng to exstng retalng hans. We term dgtalzale a produt or whh the nventory dstruton osts derease as the proporton o output retaled on-lne nreases. nalogously we term sold a produt or whh the nventory & dstruton ost s nreasng n that proporton. Sotware data newspapers mus arlne tkets nanal transatons an all take dgtal orm. We lassy suh produts as dgtalzale as they an e stored dstruted retaled at very low varale osts le transers & redt ard transatons. sold produt suh as a omputer a ook a toy lothes groeres nurs hgher nventory dstruton osts sne t reures storage hlng delvery rrespetve o ts retalng hannel. Ths dgtalzale-sold typology ders rom the hgh touh low touh dstnton rom the lterature The Eonomst 000 Levn Levn Heath 003 n that t ouses on supply han ssues not onsumer nteratons wth the produt eng oered. Supply Chan Produton Inventory Dstruton Retalng Consumer Cost Components p r h Fgure : Supply han or a rks--mortar network Supply Chan Produton Inventory Dstruton Retalng Consumer Cost Components p h r =0 Fgure : Supply han or a vrtual store network E-talers have thus had to mplement nnovatve supply han management tehnues suh as talorng a new nventory dstruton system speally or on-lne retalng estalshng large sale ontrats wth thrd party shppers or launh strateg allanes wth rks--mortar outlets. Conversely hgh touh goods suh as spealzed applaton sotware or omputer games are ute expedent or on-lne retalng sne they an seamlessly e delvered to a ustomer s desktop an e test drven eore eng purhased or example y releasng evaluaton versons or lmtng untonalty untl the ustomer uys the produt. Crpple-ware tme-lmted demos are now ommonly avalale as they help onsumers to ondut ther own researh on omparsons apaltes relaltes eore uyng suh sotware. lthough dgtalzale produts approprate themselves ute well to on-lne retalng we show that sold produts may not. For a monopolst we demonstrate that nteror solutons are rare where y nteror we mean solutons where the rm dstrutes oth on-lne o-lne. orner soluton s one where the rm dstrutes on produt s deemed hgh touh t s easer to sell when the ustomer s gven the opportunty to losely nterat wth t eore makng the purhase deson whereas a low touh good s more ustomary reures less nteraton wth the ustomer eore the purhasng deson. The low touh attrute tends to make on-lne retalng approprate sne the ustomer reures lttle nteraton wth the good eore makng the purhase. Yet ooks oten seen as low touh goods are not neessarly expedent or protale on-lne retalng sne the untonalty ost eetveness o the Internet annot e harnessed to redue ther nventory & dstruton osts. Page 38

4 Journal o Eletron Commere Researh VOL. 5 NO only one hannel. The exstene o orner or nteror solutons n our model depends solely on the ost struture o the supply han partularly on the dreton shape o the nventory & dstruton retalng ost omponents. Sne our denton o dgtalzale sold goods s entrely ased on the supply han ost they nur our model does not take nto aount the anllary enets o operatng on oth hannels suh as oerng to Internet ustomers the opportunty to rowse ompare attrutes or provdng Brks--Mortar ustomers the opportunty to go on-lne nsde the loaton ustomze ther orders. In a speal duopoly ase the leader-ollower model we show that oth rms should retal a sgnant proporton o output on-lne or dgtalzale produts. We derve ondtons or orner nteror solutons note that nteror eulra n whh ompettors smultaneously sell on- o-lne are ute restrted. Ther exstene depends agan on eah rm s ost struture. 4. Cost struture 4.. Cost desrpton We assume a three-tered ost unton p h r = where s the rm s total output s the rm s on-lne strategy.e. the proporton o that output suppled onlne - s the proporton suppled o-lne. We let x e the ost omponents gven output where x = p h or r are as dened n Fgures Euaton. Note that Euaton exludes marketng osts the ost o estalshng mantanng a produt s market share through an on- or o-lne hannel. We wegh our ous towards supply han osts as those oten dtate a rm s protalty when more than one hannel s avalale Chopra Van Meghem 000. Marketng osts o ourse play a key role n retalng hannel suess. Shoenahler Gordon 00 ndate that We only retalers spend nearly doule $4 vs $ the amount to aure a new ustomer than mult-hannel marketers spend on ustomer auston. Mult-hannel players though spend over ve tmes what Internet only marketers spend on ustomer retenton Jae 000. They also state that mult-hannel players spend sgnantly less or marketng advertsng than pure-players n spte o the szeale ost o reatng a We presene whh ranges rom $.5 mllon to $3 mllon. Our model however s only mnmally aeted y ths egregous dsrepany n marketng ost etween pure mult-hannel players. Its man prmtve as prevously speed s a new entrant who reates a market monopoly model y settng a prot-maxmzng pre or who enters a populated one ompettve model sets an eulrum pre. In oth settngs they ae a market haraterzed y a dem urve whh we assume ndependent o the rm s atons to nrease reognton or spur sales ether are onseuenes o nreasng marketng osts. The results derved rom our model ould possly arry over to a settng where marketng osts are nluded ut the solutons would then exht suh added omplexty that although more prese would sheld the nsghts our results add to the perspetve o supply han eeny. p the produton osts nlude ator nput other aptal laor. Moreover we assume p to e the purhasng ost o a produt the rm does not manuature n-house. h the nventory & dstruton osts over the pakagng storage o nshed goods the mantenane o sad storage the retreval o goods n nventory ther delvery. We expltly model the delvery harge as a ost to the rm even t s oten transerred to the ustomer. r the retalng osts nlude the ost to own rent manage or aess shel spae n shoppng areas. We set r = 0 or rms that retal exlusvely on an on-lne hannel. 4.. Cost rate o returns: ormal denton or produt typology t no loss o generalty n our results we assume eah ost omponent derentale n oth output onlne strategy. In Tale we make addtonal rst order assumptons on the ost unton that allow us to ormally dene the noton o a dgtalzale produt as ollows: DEFINITION: We term dgtalzale a produt or whh oth the nventory & dstruton h retalng r ost omponents are nreasng n dereasng n gven. nalogously we term sold a produt or whh the nventory & dstruton ost s nreasng n n gven or whh the retalng ost s nreasng n ut dereasng n gven. In our model the derene etween sold dgtalzale les solely n the rst order ehavor o the nventory & dstruton ost n gven. The ost o produton s assumed nreasng n ut onstant n gven sne the rm s hoe o retalng hannel has no eet on ths ost. We note that or dgtalzale produts produton osts or addtonal unts eyond the rst one are stll nreasng ut at a slower pae than or sold goods. For sold produts nventory & dstruton osts n gven do suer when the numer o delvery ponts nreases or when orders nrease n Page 39

5 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar numer ut derease n sze. The nventory systems are more omplex sne on-lne orders or a spe tem rarely exeed one unt e.g. a ustomer hardly ever orders more than one opy o a ook. n on-lne retaler o sold goods must also ensure that the purhases reah the ndvdual ustomers generally through a mal or ourer serve. Even orne y the ustomers the dstruton osts entaled oten prevent the on-lne retaler rom prng ompettvely aganst a rks--mortar retaler. For dgtalzale produts however that ost dereases sne only one unt needs to e stored opes an e made shpped eletronally on reuest payment. In at h eng nreasng or sold goods dereasng or dgtalzale goods s the only ost omponent that does not ehave the same or sold versus dgtalzale goods. The ost o retalng s assumed to derease n gven sne the rm does not need to aess as many rks--mortar outlets to aheve the same sales level. Tale : ssumptons on rst order ehavors rate o returns o ost omponents Sold Produt Dgtalzale Produt Cost Component n n gven n n gven Produton: p nreasng Constant nreasng onstant Inventory & Dstruton: h nreasng Inreasng nreasng dereasng Retalng: r nreasng Dereasng nreasng dereasng Knowledge o the seond order ehavor o the ost unton s key to haraterzng a rm s on-lne strategy. For example onsder an exlusve o-lne retaler o a dgtalzale good. By our denton o dgtalzale < 0.e. the dervatve rate o return o the total ost unton wth respet to gven s negatve. Now suppose the rm egns retalng ths good on-lne. Its total osts ether derease at an nreasng rate onave or dereasng rate onvex. In the latter ase the rm aptures the ulk o the savngs rom movng on-lne mmedately: savngs rom shuttng down rks--mortar outlets n the r porton outwegh the nreased osts ndued y the new omplexty o managng nventory & dstruton at the ustomer level the h porton ths gap dereases as the on-lne presene nreases. 5. Monopoly Results We solve the prolem or the on-lne strategy gven rm output or a new rm wth a radal nnovaton enterng a market as a monopolst. Fndngs are or an eonomy wth one two markets. 5.. One market The monopolst s prolem s to nd the value o that maxmzes total prots.e. p h r Max where - s the lnear dem urve that the monopolst aes. s the postve nterept on the pre axs - s the total revenue that the rm generates. Tale presents the on-lne strateges under a xed output level.e. gven that solve Euaton. Sne we assume lnear dem the optmal on-lne strategy only p h r depends on the ehavor o the total ost unton. Formally PRESCRIPTION : I ts produt s dgtalzale then the sum o nventory & dstruton osts retalng osts h r s dereasng the monopolst only retals on-lne Tale ases 5. I ts produt s sold a prot maxmzng monopolst only retals on one hannel whenever the sum o ts nventory & dstruton osts retalng osts s ether.a non-monotone has non-nreasng returns.e. weakly onave: Tale ase 3 or. monotone Tale ases 4 or 5. Note that ases 5 apply to oth sold dgtalzale produts eause the sum o a rm nventory & dstruton retalng osts ould also e dereasng or sold produts. Page 40

6 Journal o Eletron Commere Researh VOL. 5 NO I ts produt s sold a prot maxmzng monopolst should retal oth on- o-lne whenever the sum o ts nventory & dstruton osts retalng osts s non-monotone has nreasng returns.e. onvex: Tale ase 6. Tale : Optmal on-lne strategy gven Inventory/Dstruton Retalng Costs n gven : h r Case : Conave nreasng 0 Case : Conave dereasng Case 3: Conave non-monotone 0 h r h 0 r 0 h r < h 0 r 0 Case 4: Convex nreasng 0 Case 5: Convex dereasng Case 6: Convex non-monotone h r 0 at d d = 0 d d In o-lne retalng s not attratve eause oth nventory & dstruton retalng osts derease monotonally n the on-lne strategy. Ths result explans the many usnesses that although operatng n ompettve envronments have shted retalng o ther dgtalzale goods on-lne nanal nsttutons transportaton & aommodaton supplers sotware data retalers. In the rm only retals o-lne = 0 when the savngs n retalng osts do not outwegh the addtonal osts n nventory & dstruton.e. h r s nreasng or all [0]. However the rm only retals on-lne = h r s dereasng or all [0]. Ths presrpton proposes a ustaton to the soarng suesses o onlne retalers o dgtalzale goods ut also to the hallenge at turnng a prot or many on-lne retalers o sold goods. To turn protale the atvty a monopolst should exht a lower h nventory & dstruton ost than ts opportunty savngs o not operatng rks--mortar. These results may not e oservale when oetves other than prot maxmzaton.e. stok valuaton market penetraton produt or rm poston dtate the on-lne strategy. Oserve rom Fgure 3 that although not operatng n monopoly settngs ndustres retalng a sold hgh touh good exht oreasted on-lne sales o no more than 0% o total. Nonetheless the popular e-ommere stes are typally retalers o sold goods The Eonomst 000 most are operatng at losses. Ths has ndued on-lne retalers to adopt state-o-the-art nventory dstruton systems loatons stokng/pkng mehansms strateg allanes eent ontrats wth thrd party shppers. In the rm sets ts on-lne strategy exatly where the rate o savngs rom shuttng down one retal outlet mathes the nreased expenses rom the addtonal delvery ponts. Many suessul retalers o sold produts have ndeed opened an on-lne retalng hannel ut have also kept ther rks--mortar presene. Thus one possle avenue or an entrant monopolst who sells say a new home eletron deve would e to retal t on oth on- o-lne hannels ut keep ther xed osts low perhaps y sellng through ntermedates suh as mazon Crut Cty. The proporton o on-lne supply s gven y h r 0 at d d = 0. d d Page 4

7 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar 5.. Two markets The open-to-the-world nature o the Internet suggests that dem or a produt avalale on-lne domnates o-lne dem eause the potental reah o the on-lne market s gloal that many ustomers would not e reahed wthout an on-lne market. Ths ele has argualy ontruted to the dot om eet soarng values o ntal pul oerngs o Internet retalers. Many e-talers exp ther on-lne oerngs do so ased on the tenet that on-lne potental dem domnates o-lne. Our model attempts to predt on- o-lne protalty o produts ased on ther dgtal ontent level. We thus nlude ths tenet n our model to te our results wth ths ele. Speally we assume that the rm smultaneously aes two markets haraterzed y the ollowng dem urves: P = I - I or on-lne retalng P = B - B or o-lne retalng where I > B > 0 0 I / B. Smlarly to the desrpton n Seton 5. I - I s the lnear on-lne dem urve wth I as the postve onlne nterept on the pre axs the slope o the dem urve. B - B s the o-lne dem urve wth a stardzed slope eual to B s the postve o-lne nterept on the pre axs. The two ondtons I > B > 0 0 I / B ensure that the on-lne dem urve domnates the o-lne one. We plot a representaton o the dem urves n Fgure 3. P= I - I as nreases rom 0 to I / B P I B P= B - B B Fgure 3: Dem urves aed y the rm The monopolst s prolem s now Max 3 I B where I s the total revenue the rm generates rom the on-lne market whereas B s that rom the o-lne market. Total ost s gven y Euaton. Exstene o nteror solutons depends on the onvexty or asene thereo n the ost unton on the on-lne o-lne dem proles I - B o the value proposton. Let e the partal dervatve o wth respet to. The rst order suent ondton or a maxmum s I B =. 4 We dstngush two ases: lnear n onvex nreasng returns or onave dereasng returns n. Formally or the ormer ase Page 4

8 Journal o Eletron Commere Researh VOL. 5 NO PRESCRIPTION : Let total ost e lnear n the on-lne strategy wth a onstant slope = α. For I dgtalzale goods α < 0 thus / or euvalently I B. 3 More generally α the rm should exlusvely retal on-lne I B = whenever ; α the rm should retal o-lne I B = 0 whenever ; 4 the rm should retal on oth hannels I B α = whenever α I B. The rst part o Presrpton alls or postve supply to the on-lne hannel the rm retals a dgtalzale good ut not neessarly exlusve supply sne n ths model there s dstnt prot opportunty on the o-lne hannel. More generally the monopolst s on-lne strategy gven a xed output s represented n Fgure 4. Ths strategy prole s remnsent o the monopolsts optmal output level set where margnal revenue euals margnal ost. In ths development s xed α s the savngs rate emoded n the rm s total ost unton wth respet to ts on-lne strategy. We oserve that I - B s the nremental revenue to the rm o supplyng on-lne. I the derene s too postve negatve then the rm should exlusvely retal o-lne respetvely on-lne. = - =0 α I B Fgure 4: Monopolst s on-lne strategy gven a xed output level = α When s strtly onave or onvex n we perorm our analyss usng Fgures 5 6. The optmal strategy prole s gven y whh solves Euaton 4. The optmal on-lne strategy thus depends on the onvexty or onavty o the nventory & dstruton retalng portons o the total ost unton. For a dgtalzale produt as Fgure 5 shows a rm may enet rom retalng oth on a rks--mortar network on a vrtual store ts ost unton despte eng dereasng does not apture enough savngs rom movng more output onlne.e. s not onvex enough. Indeed as the unton s onvexty dereases the let-h sde o Euaton 4 lattens or a xed we note that the set o possltes or a rossng that would yeld < oarsens. For dgtalzale goods the let-h sde urves are oth negatve at =0. They nrease derease the ost unton s onvex onave ut not neessarly lnearly. 3 Ths result dretly ollows rom Euatons 4 [ 0 ] I B. 4 lthough the rght-h sde o Euaton 4 annot exeed ts let-h sde an. Page 43

9 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar Rght-h sde o Euaton 4 as nreases onvex Let-h sde o Euaton 4 or derent orms o onave Fgure 5: Charaterzaton o or dgtalzale goods gven y the -axs oordnate o the lak dots We also note that dgtalzale goods wth more elast on-lne dem small all agan or a strong not exlusve on-lne presene sne the set o possltes or a rossng that would yeld < oarsens agan. Moreover or a gven ost struture we oserve that dereases as nreases. Fgure 6 shows that eters parus the on-lne strategy wll e less or a sold produt than a dgtalzale one. We rst reall that a sold produt need not exht < 0 see Tale. To llustrate the eet on we smply redraw the -axs at a lower poston oserve that the nterseton ponts all sht let on the horzontal axs. For sold goods the let-h sde urves sht up ompared to Fgure 5. For a xed we oserve that sht let or tangle goods.e. a larger proporton o output s retaled o-lne. Rght-h sde o Euaton 4 as nreases onvex Let-h sde o Euaton 4 or derent orms o onave Fgure 6: Charaterzaton o or sold goods gven y the -axs oordnate o the lak dots Page 44

10 Journal o Eletron Commere Researh VOL. 5 NO Duopoly Model We now rene our model to nlude the senaro where the new rm enters an already populated market. duopoly approah mples that the market only has an nument monopolst when the new rm enters. Hene the results derved do not mmedately apply to a general settng o multple numents. Yet the general presrptons approprate themselves to ths general settng one puts asde the ompettve ehavor aross the exstng numents. We derve results under a leader-ollower senaro: gven an nument monopolst operatng exlusvely on- or o-lne where should a potental entrant to the ndustry poston tsel.e. how muh o ts output should e retaled on- o-lne? For example a szeale porton o the market share or omputers has shted to Dell other on-lne retalers. Compa the leader n o-lne retalng or omputers reated y openng ts own on-lne store. Smlar dynams an e oserved n ook retalng ankng tradng travel serves the lke. Eulrum ondtons under whh oth rms aheve maxmum separaton one on-lne one o-lne ompete exlusvely on one market or on oth are also derved or the leader-ollower senaro. We now talor the model assumptons notaton or a duopoly ramework. In a deoupled market settng dem s gven y pb = B B = B 5 pi = I I = I = ; the ost untons or eah rm are gven y p h r d = d = d d = 6 where we add the supersrpt euals to or to dstngush etween the two rms. In ths analyss always reers to rm s osts not the suare o. We also make a dstnton etween market uanttes rm output wth a aptal or the ormer. Frm s assumed to e the leader an nument monopolst operatng on only one hannel o-lne or B I on-lne. Hene we x = 0 or = or = respetvely where s rm s ost unton dervatve wth respet to. Frm s an entrant to the ndustry maxmzes ts prots gven y Π ; = 7 I B where s rm s ost unton gven a xed output level. ssumng that has nreasng returns onvex n we derve that rm s optmal on-lne strategy s the soluton to I B B = 8 4 rm operates exlusvely o-lne = 0 the soluton to I B I = 9 4 rm operates exlusvely on-lne =. gan we perorm our analyss usng the shemat llustratons n Fgures 5 6 7a 7. Comparng Euatons 8 9 to Euaton 4 we note that the let-h sde s larger smaller rm retals exlusvely o- respetvely on- lne. 5 The representatve let-h sde urves n Fgure 7a respetvely wll hene sht down respetvely up ompared to the analogous urves n Fgures 5 6. Hene there ndeed s a ompettve advantage or an entrant to separate rom the nument hoose a dstntvely lower or hgher on-lne strategy. Fgure 7a shows that the larger the leader s output s the lower the let-h sde urve shts. For a onvex ost unton ths means that Euaton 8 s only satsed or values o that are strtly postve. Moreover we oserve that the range o solutons where = nreases.e. extensons o the let- rght-h sde urves are 5 We also note that ths hange s losely related to the leader s output sne t nreases y / B / a multplatve onstant away rom the leader s output the stard monopoly prot maxmzng uantty. Smlarly the leader retals exlusvely on-lne the hange dereases y / I /. Page 45

11 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar more lkely to nterset eyond =. The opposte eet ours n Fgure 7. The let-h sde o Euaton 9 s now hgher as a unton o the nument s output level. Keepng the rght-h sde o Euaton 9 xed we oserve that moves let.e. dereases that the entrant should agan derentate tsel y retalng a szeale proporton not all o ts output o-lne whenever the nument only retals on-lne. These oservatons are summarzed n Presrpton 3. Rght-h sde o Euaton 8 or a xed onvex Let-h sde o Euaton 8 or derent orms o. onave Fgure 7a: Charaterzaton o gven = 0 Rght-h sde o Euaton 9 or a xed onvex Let-h sde o Euaton 9 or derent orms o. onave Fgure 7: Charaterzaton o gven = Page 46

12 Journal o Eletron Commere Researh VOL. 5 NO Page 47 Tale 3 presents the ondtons under whh two rms operate exlusvely on one hannel = 0 or = or orner solutons. Tale 3 also reports on the ondtons where one rm only retals on one hannel whle the other retals on two or oth rms ompete on oth hannels.e. the nteror solutons. We ormally state our ndngs as Presrpton 3-v. Tale 3: Condtons or orner nteror solutons = I - B {} {} 0 > 0 {} > 0 {} > > 0 {} 0 0 > 0 {} {} 0 PRESCRIPTION 3: I the nument operates exlusvely on one hannel produes a large enough output volume t wll ndue the new entrant to retal exlusvely on the other hannel. I the produt retaled y oth rms s dgtalzale.e. 0. h 0. h then oth rms always retal a postve proporton o ther output on-lne. Eah rm must use dssmlar tehnologes or one rm to retal on two hannels when ts ompettor operates exlusvely on one. v Rarely should oth rms hoose to dstrute on oth retal hannels.

13 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar Result s a onseuene o our deoupled markets as one market gets saturated the nument aptalzes on a larger prot opportunty rom the other market. Result ollows rom the at that the only possle orner soluton or a dgtalzale produt s or oth rms to exlusvely retal on-lne. Ths result s ongruent wth ntuton. The sad ondtons pn the nremental revenues rom movng on-lne always postve sne the on-lne market s y assumpton larger than the o-lne aganst the orrespondng ost nreases whh are negatve y denton. In suh outomes an our whenever eah rm uses dssmlar tehnologes sne ther ost unton s onavty s ounded elow rm aove rm. In v the ondtons o Tale 3 only mpose a lower ound on the onavty n o eah rm s ost unton. Ths entals uk savngs rom movng on-lne that derease as the on-lne poston nreases. Only rms that exht ths somewhat restrtve unlkely ost struture would enet rom ompetng on oth hannels o dstruton. We also nvestgated a smultaneous-game senaro where eah rm s ost struture was lnear n output gven the on-lne strategy where these rms ompeted on a sngle market. Fndngs were onsstent wth that o the aove models. The eulrum or dgtalzale produts alled or exlusve on-lne retalng or oth rms or sold produts exlusve o-lne retalng when ther ost untons nrease n the on-lne proporton. Tehnal detals explanatons or these eulra are oered n the appendx. 7. Conluson In ths paper we showed that the deson to retal on-lne nvolves more than the lure o addtonal revenues rom sellng to an exped market that the Internet provdes. The supply han tehnologal attrutes emoded n the ost unton along wth the produt s key haratersts dgtalzale or sold should e the drvers ehnd how muh produt a new entrant should sell on-lne or o-lne. We demonstrated that there are ndeed protalty monopoly models or ompettve duopoly models advantages n a rm s strateg hoe o dstruton hannels. For a monopolst new rm dgtalzale produt should e solely retaled on-lne as seen wth nanal nsttutons transportaton & aommodaton supplers sotware data retalers that have suessully shted the retalng o ther dgtalzale goods to an on-lne hannel even when operatng n ompettve envronments. Sold produts on the other h should e retaled on one or two hannels ased on the ehavor o the monopolst s nventory & dstruton osts retalng osts as retalng a sold good on-lne may ause an nrease n supply han osts that on-lne revenues wll not outwegh. When the monopolst new rm aes an eonomy wth two markets o- on-lne t may enet rom retalng a dgtalzale produt on oth hannels whenever ts ost unton despte eng dereasng does not apture enough savngs rom movng more output on-lne. For an entrant ang an already populated market we analyzed a leader-ollower senaro where gven an nument monopolst operatng exlusvely on one hannel a potental entrant to the ndustry must dede how muh o ts output should e retaled on- o-lne. We presre that whenever the nument operates exlusvely on one hannel produes a large enough output volume the entrant should retal exlusvely on the other hannel thereore revealng a new ompettve strategy n the proporton o output one rm sells on-lne where as that proporton nreases or the nument t wll derease or the entrant ve versa. lso the produt retaled y oth rms s dgtalzale then oth rms should always retal a postve proporton o ther output on-lne. Moreover eah rm must use dssmlar tehnologes ost strutures or one rm to retal on two hannels when ts ompettor operates exlusvely on one rarely should oth rms hoose to dstrute on oth retal hannels. There are many ways n whh a new entrant an harness the Internet to enet ts usness. In reent years we have oserved the evoluton o on-lne retalng rom new rms oerng an exlusve on-lne presene to tradtonal rks--mortar rms openng on-lne stores. Suessul on-lne retalers today are those that have een ale to keep ther operatng osts low wth sophstated nventory management systems eent ontrats wth thrd party shppers strateg allanes wth exstng rks--mortar outlets. Presrng alternatve strateges suh as restrtng Internet presene y only keepng an on-lne lstng or advertsng produts ut only sellng through tollree numers or rks--mortar outlets may all outsde our model sope. We have assumed the hoe set o the rm to e restrted to the proporton o output sold on-lne that all the rm s perormane tehnology ould e emoded n the three-tered ost unton Euaton. There are a numer o useul dretons or extendng ths researh to apture more o the omplexty n retalng desons. One s to add marketng osts to the model whh would nrease the preson o our results. However our prelmnary analyss shows that suh osts render the dervaton o useul results rom our urrent models ute tedous. Ths suggests that numeral or empral analyses would e etter suted to apture the nsghtul trade-os etween marketng supply han osts when a rm has the opton o operatng on more than one hannel. There s muh stll to e learned on the vrtual/rk--mortar deson ths deson s Page 48

14 Journal o Eletron Commere Researh VOL. 5 NO entral to a rm s perormane. Stll ths new researh oupled wth our ndngs wll help new entrants make more normed desons aout whh hannels to utlze. REFERENCES nupnd R. Y. Bassok E. Zemel general ramework or the study o deentralzed dstruton systems Manuaturng & Serve Operatons Management Vol. 3 No. 4: Fall 00. Bakos Y. The emergng lsape or retal E-ommere Journal o Eonom Perspetves Vol. 5 No. : Bakos Y. E. Brynolsson Bundlng ompetton on the Internet Marketng Sene Vol. 9 No. :63-8 Wnter 000. Balasuramanan S. Mal versus mall: strateg analyss o ompetton etween dret marketers onventonal retalers Marketng Sene Vol. 7 No. 3:8-95 Fall 998. Chopra S. J.. Van Meghem Whh e-usness s rght or your supply han? Supply Chan Management Revew July-ugust 000. Devnney T.M.. Latukeu D. F. Mdgley E-usness: revoluton evoluton or hype? Calorna Management Revew Vol. 44 No. : Dewan R. B. Jng. Sedmann dopton o Internet-ased produt ustomzaton prng strateges Journal o Management Inormaton System Vol. 7 No. : Fudenerg D. J. Trole Game Theory. The MIT Press Camrdge M 993. Gulat R. J. Garno Get the rght mx o rks lks Harvard Busness Revew Vol. 78:07-4 May- June 000. Hendershott T. J. Zhang model o dret ntermedated sales Workng paper Unversty o Calorna at Berkeley Unversty o Rohester 00. Htt L.M. F.X. Fre Do etter ustomers utlze eletron dstruton hannels? The ase o PC ankng Management Sene Vol. 48 No. 6: June 00. Homan D.L. The revoluton wll not e televsed: Introduton to the speal ssue on Marketng Sene the Internet Marketng Sene Vol. 9 No. :-3 Wnter 000. Jae R. Mult-hannel vs pureplay phlosophes appear nght day PaneWeer Researh Note Vol. 4: ugust 000. Kauman-Sarorough C. J. D. Lndust E-shoppng n a multple hannel envronment Journal o Consumer Marketng Vol. 9 No. 4: Fall 00. Keeney R.L. The value o Internet ommere to the ustomer Management Sene Vol. 45 No. 4: prl 999. Kellee H. Vendors sellng dret onlne Sportng Goods Busness Vol. 34 No. :8 ugust 00. Leamer E.E. M. Storper The eonom geography o the Internet age Journal o Internatonal Busness Studes Vol. 3 No. 4: Levn.M. I.P. Levn I.P. C.E. Heath Produt ategory dependent onsumer preerenes or onlne olne shoppng eatures ther nluene on mult-hannel retal allanes Journal o Eletron Commere Researh Vol. 4 No. 3:85-93 ugust 003. Lukng-Reley D. D. F. Spuler Busness-to-usness eletron ommere Journal o Eonom Perspetves Vol. 5 No. :55-68 Wnter 00. Messnger P.R. C. Narasmhan model o retal ormats ased on onsumers eonomzng on shoppng tme Marketng Sene Vol. 6 No. :-3 Marh 997. Salop. S. Monopolst ompetton wth outsde goods Bell Journal o Eonoms Vol. 0: Shoenahler D.D. G. L. Gordon Mult-hannel shoppng: understng what drves hannel hoe Journal o Consumer Marketng Vo. 9 No.: The Eonomst Shoppng around the we. survey o e-ommere Feruary Tsay.. N. grawal Manuaturer reseller perspetves on hannel onlt oordnaton n multple-hannel dstruton Workng paper Santa Clara Unversty 00. Vshwanath V. G. Mulvn Mult-Channels: The Real Wnners n the BC Internet Wars Busness Strategy Revew Vol. No. : Page 49

15 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar PPENDIX Tehnal detals or seton 5.. One market solvng or rm output The monopolst now solves or total output M must thus Max wth = elted n Presrpton. The prot maxmzng uantty M s the soluton to d =. d The general nature o Euaton hnders the dervaton o an ntutve losed orm soluton. The prot maxmzng uantty M that ndues 0 = = 0 s the soluton to = respetvely =. From as mro-eonoms we oserve that these two uanttes are exatly the prot maxmzng levels or a monopolst gven a lnear dem urve a general margnal ost unton. Tehnal detals or seton 5.. Two markets solvng or rm output The monopolst solves or total output M must thus Max 3 I B wth = elted n Euaton 4. Here agan the general nature o Euaton 3 hnders the dervaton o an ntutve losed orm soluton. We nonetheless note that the prot maxmzng uantty M that ndues = s the soluton to B 0 I = the analogous uantty that ndues = 0 s the soluton to =. These two uanttes orrespond to the stard prot maxmzng output level or a monopolst gven the approprate dem urve margnal ost unton. We oserve that or dgtalzale produts wth a more elast on-lne dem the ormer uantty domnates the latter one. sde rom stard eonom ntuton optmal output level depends on oth market strutures on the rm s ost unton dervng a general orm o M reveals lttle addtonal nsght over the results already stated. Tehnal detals or seton 6 Duopoly Model The rst seond order dervatves o Euaton 7 are respetvely Π = I B [ ] Π = 4. 5 ssumng that s onvex n then settng Euaton 4 to zero s suent to maxmze Euaton 7. Ths yelds [ ] I B =. 6 Euaton 6 together wth the at that rm s the leader a monopolst operatng only on one hannel B I thus we an x = 0 or = or = respetvely lead to Euatons 8 9 hene Presrpton 3. Condtons or nteror orner solutons. Consder Euatons We note that = 0 Π ; = 0 whenever rom Euaton 5 whenever maxmzes Page 50

16 Journal o Eletron Commere Researh VOL. 5 NO = 0 ; = 0 Π or euvalently where = I B. I = 0 we see that settng = 0 maxmzes Π = 0 ; = 0 = ; = 0 Π 0 ; = 0 = ; = 0 Π 0 Π or euvalently. Ths result s reported n Tale 3. In addton Tale 3 presents all ondtons or the other leader-ollower orner nteror solutons. ddtonal tehnal detals or dervng these ondtons are omtted as they parallel the aove development. Presrpton 3 s derved as ollows. By denton we know that h h = or sold produts. Thereore the ondtons or = = 0 = 0 = or = = 0 o Tale 3 an never e aheved. Moreover the only possle orner soluton or a dgtalzale produt s = =. For Presrpton 3 onsder solutons where rm sets = 0 or = rm preers a strategy n 0 we an ondut a smlar analyss or = 0 or = rm preers a strategy n 0. n nteror soluton otans whenever rm s ost unton s n onvex or not too onave.e. > whereas rm s s strtly onave.e.. For suh outomes to our eah rm must use dssmlar tehnologes sne ther ost unton s onavty s ounded elow rm aove rm. Fnally or Presrpton 3v we ous on ases where oth rms hoose to dstrute on two retal hannels 0 =. Here the ondtons only mpose a lower ound on the onavty n o eah rm s ost unton. Nash eulra or two rms wth lnear ost untons In ths somewhat restrtve senaro rm s ost unton s lnear gven y = δ where we assume δ = to e non-negatve derentale on the nterval [0]. Oserve rom Tale that rm retals a dgtalzale produt then ts ost s nreasng n ts on-lne strategy dδ / d > 0 ut that the reverse s not neessarly true.e. retalng a sold produt does not mean that the rm s ost s dereasng n ts onlne strategy dδ / d < 0. In addton we assume that oth rms ompete on a sngle market that has lnear dem gven y -. Furthermore we make a strutural assumpton or easlty where δ δ 0 or any n [0] =. Ths assumpton ensures that the eulrum n output s non-negatve or eah rm. Frm maxmzes ts prots gven y Π = =. PRESCRIPTION : Dgtalzale produts should never e exlusvely retaled o-lne ut sold produts ould whenever rms osts are nreasng n ther on-lne strategy. Sold produts should never e exlusvely retaled on-lne unless rms osts are dereasng n ther on-lne strategy ut a Nash eulrum exsts where dgtalzale produts should e exlusvely retaled on-lne. separaton strategy where one rm exlusvely retal on-lne whereas the other rm has an exlusve presene o-lne s never optmal or dgtalzale produts. v It s never optmal or one rm to retal a dgtalzale produt on two hannels when ts ompettor operates exlusvely on one. PROOF: We note that Π thus the rst-order ondton s suent or a maxmum. Moreover / Π / = > = Π / sne there exsts a rossng etween the rms reaton unton Fudenerg Trole 993 hene an eulrum n output where = [ 3 δ δ ] Π = 3[ δ δ ] or =. We now detal ondtons under whh orner nteror solutons are otaned. Frst reall that = = 0 means that oth rms operate exlusvely o-lne. To aheve ths the ollowng ondtons must hold Page 5

17 Renhardt & Lévesue: New Entrant s Deson on Vrtual vs. Brks--Mortar Π 0 = 3[ δ 0 δ ] 3[ δ 0 δ 0] = Π 00 Π 0 = 3[ 0 δ ] 3[ 0 δ 0] = Π 00 or arg mnα = 0 arg mnα = 0 should never e exlusvely retaled o-lne rom Tale t ollows that dδ / d < 0 δ δ. From these we readly onlude that dgtalzale produts ut sold produts ould whenever dδ / d > 0 Presrpton. nalogously or oth rms to retal exlusvely on-lne = = one must have argmnδ = argmnδ =. Hene a Nash eulrum exsts where dgtalzale produts should e exlusvely retaled on-lne ut sold produts should not unless / d < 0 = 0 = 0 one must have dδ Presrpton. For = = arg mn δ = 0 argmnδ = argmnδ = argmnδ = 0 respetvely. Thereore or dgtalzale produts we do not oserve the separaton eet as we dd n the prevous seton Presrpton. Fnally or dgtalzale produts a soluton orm suh as = 0 0 exsts as long as a argmnδ = 0 [ 0 δ ] [ dδ / d] = 0 gven that holds / d < 0 δ st order ondton or nteror soluton dδ nd order ondton or nteror soluton whh s mpossle. Smlar ondtons must hold or nteror solutons 0 & 0 & 0 0 & 0 & 0 = = =. Thus t s never optmal n ths model or one rm to retal a dgtalzale produt on two hannels when ts ompettor operates exlusvely on one Presrpton v. ollow rom ntuton. Indeed prot maxmzaton mposes on oth rms that they low an optmal volume on eah hannel. We otan those results sne the on-lne hannel s more expedent or dgtalzale produt as oth the nventory & dstruton the retalng osts thereore total ost derease wth the onlne proporton. The o-lne hannel s more expedent or sold produts whenever the nventory & dstruton ost s over aove that o retalng thus makng total ost nreasng n the on-lne proporton. However s somewhat ounter-ntutve sne we have demonstrated eore that a separaton strategy an e alled or even or dgtalzale produts. Stll under our sngle market assumpton there s no guarantee that exlusvely usng one hannel when the other s ouped s warranted sne the other hannel does not provde aess to addtonal ustomers. Fnally v ollows rom the model s premses namely the lnear ost unton whh makes suoptmal solutons where a rm retals a dgtalzale produt on oth hannels. Page 5