From Web Vis­its to Firm Orders

“Our project demon­strates that, at a min­i­mum, firms should keep track of their click data, start­ing with who is on the web­site and for how long,” says Jan Van Mieghem, a pro­fes­sor of man­age­r­i­al eco­nom­ics and deci­sion sci­ences at the Kel­logg School, who under­took the study with his Ph.D. stu­dent Tingliang Huang. The research also indi­cates which types of clicks are the key fac­tors in site vis­i­tors’ deci­sion-mak­ing. That infor­ma­tion, the pair con­cludes, will allow ven­dors to pre­dict the prob­a­bil­i­ties that spe­cif­ic vis­i­tors will order prod­ucts fea­tured on the web­sites, as well as the like­ly amounts and tim­ing of those orders.

The extrac­tion of oper­a­tional val­ue stems from the time between poten­tial cus­tomers’ vis­its to non-trans­ac­tion­al web­sites and their result­ing place­ments of orders, or lead time. By gath­er­ing data on those lead times — which great­ly exceed the neg­li­gi­bly small click-order inter­vals in e-com­merce sites — ven­dors can use the vis­its to pre­dict orders and there­by make plans and adjust­ments before the actu­al orders arrive.

Such a capa­bil­i­ty presents obvi­ous busi­ness ben­e­fits — in con­trol­ling inven­to­ry or pro­duc­tion plan­ning, for exam­ple. ​“Match­ing sup­ply to uncer­tain demand is chal­leng­ing for many firms, and mis­match­es are cost­ly,” Huang explains. ​“If ven­dors can pre­dict future sales bet­ter, they can sig­nif­i­cant­ly reduce the costs of the mismatches.”

Cor­re­la­tions in Non-trans­ac­tion­al Web­sites

Researchers have exten­sive­ly stud­ied the cor­re­la­tions between e-com­merce con­sumers’ online behav­ior and their pur­chas­ing propen­si­ties. But lit­tle has been done to under­stand those cor­re­la­tions in non-trans­ac­tion­al web­sites, such as those com­mon­ly used by busi­ness-to-busi­ness (B2B) firms. ​“What we believe is nov­el is link­ing the clicks to the oper­a­tional per­spec­tive,” Van Mieghem says. ​“This is prob­a­bly the first empir­i­cal study of this kind.”

The research is sig­nif­i­cant because of the pop­u­lar­i­ty of non-trans­ac­tion­al web­sites. ​“Their use is sur­pris­ing­ly large in B2B sales,” Van Mieghem says. The approach makes sense for many com­pa­nies sell­ing rel­a­tive­ly small num­bers of prod­ucts that must be cus­tomized, that require exten­sive test­ing by buy­ers before they decide to make a pur­chase, or that involve detailed price nego­ti­a­tions before a sale is com­plet­ed. Many web­sites of this type con­tain lit­tle more than online ver­sions of their print­ed catalogues.

The Kel­logg researchers set out to dis­cov­er whether track­ing vis­i­tors’ clicks on the web­sites has any val­ue in fore­cast­ing future trans­ac­tions — and if it does, in quan­ti­fy­ing the rela­tion­ship between spe­cif­ic forms of click behav­ior and sales. ​“The rea­son for our analy­sis was not just to con­firm the rela­tion between click behav­ior and pur­chas­ing behav­ior but also to iden­ti­fy and quan­ti­fy the key fac­tors,” Van Mieghem explains.

To do so, Van Mieghem and Huang used infor­ma­tion on the North Amer­i­can mar­ket of a sin­gle, anony­mous com­pa­ny that sells indus­tri­al prod­ucts glob­al­ly through its non-trans­ac­tion­al web­site. The company’s CEO — a Kel­logg alum­nus — gave the pair access not only to click data but also to sales infor­ma­tion about cus­tomers’ accounts. ​“That’s a sen­si­tive issue for com­pa­nies,” Van Mieghem says. ​“We were for­tu­nate in hav­ing such a close col­lab­o­ra­tion with the company.”

A Vari­ety of Vari­ables

The research involved trolling through the entire­ty of what Van Mieghem calls ​“noisy data” to dis­cov­er the key fac­tors relat­ed to future orders. ​“There’s a vari­ety of click vari­ables you can look at, but only a few seem to be dri­ving most of the infor­ma­tion,” Van Mieghem explains.

Based on their analy­sis, he and Huang report­ed in their paper that ​“vis­i­tor online click behav­ior is indeed pro­vid­ing the firm use­ful infor­ma­tion to pre­dict future order­ing prob­a­bil­i­ties.” Specif­i­cal­ly, they found that both the fre­quen­cy of site vis­its and the num­ber of vis­its to impor­tant web pages cor­re­lat­ed in com­plex ways with the propen­si­ty to order prod­ucts. For exam­ple, the longer vis­i­tors stayed on the site, the more like­ly they were to order from it. That propen­si­ty starts to decline, how­ev­er, after a cer­tain length of stay. In addi­tion, the click behav­ior of new cus­tomers dif­fered from that of exist­ing ones.

Detailed analy­sis of the cor­re­la­tions pro­duced a regres­sion equa­tion quan­ti­fy­ing the like­li­hood that a par­tic­u­lar click will lead to a pur­chase. ​“The com­pa­ny had been using click strik­ing to detect the ten most impor­tant sales prospects they would cold call,” Van Mieghem says. ​“We are pri­or­i­tiz­ing each of these cus­tomers in terms of the like­li­hood that they will buy. We can even pre­dict how large a pur­chase will be and when the pur­chase is like­ly to occur.” The analy­sis also pro­vid­ed insights into web­site vis­i­tors’ strate­gies — whether, for exam­ple, they will decide to buy imme­di­ate­ly or to wait in hopes of get­ting a bet­ter deal but at the same time delay­ing their con­sump­tion of the product.

Broad Appli­ca­tion Does the result of research on a sin­gle non-trans­ac­tion­al web­site have broad appli­ca­tion to all web­sites of that type? ​“Our find­ings must be inter­pret­ed cau­tious­ly, giv­en the lim­i­ta­tions of our study,” the researchers note in their paper. Nev­er­the­less, Van Mieghem says, ​“I’m con­fi­dent that a sim­i­lar type of analy­sis can be done for any site, though the coef­fi­cients will dif­fer. I expect the over­all pic­ture — our lon­gi­tu­di­nal analy­sis — will work.” Indeed, he and Huang believe that ven­dors with non-trans­ac­tion­al sites that have a rea­son­able lead time between vis­its and order place­ments can also ben­e­fit from the same analy­sis of vis­i­tors’ click behavior.

Huang sum­ma­rizes the over­all con­clu­sion of the study: ​“It turns out,” he says, ​“that click track­ing typ­i­cal­ly brings win-win out­comes for the firm and its cus­tomers, espe­cial­ly com­pared to tra­di­tion­al oper­a­tions and mar­ket­ing strate­gies stud­ied in the literature.”

Relat­ed read­ing on Kel­logg Insight

Beat­ing the Bot­tle­necks in E-Com­merce: Effec­tive­ly allo­cat­ing Web sys­tem capacity

Peo­ple are Talk­ing: Antic­i­pate prod­uct per­for­mance through online discussions

Learn­ing from Zil­low and Zoots: Improved per­for­mance through ser­vice inven­to­ry management