It was a split-second decision.”

How many times have you heard those words or used them your­self with­out giv­ing them another thought? Today, let’s con­sider the sig­nif­i­cance of this famil­iar phrase. It sug­gests a men­tal sum­ma­tion so com­pelling and yet so quick that we’re unable to iden­tify all the con­tribut­ing fac­tors. Like a “hunch,” the phrase describes a deci­sion process that takes place beneath the level of our con­scious aware­ness, as if from a black box.

In this accel­er­ated and impa­tient age, when mar­keters have a brief win­dow of oppor­tu­nity to cap­ture the inter­est of online cus­tomers, some­times it seems like every sit­u­a­tion requires a split-second deci­sion. Pre­sent­ing a prod­uct or ser­vice unsuited to the online vis­i­tor, or tak­ing an exas­per­at­ing few moments to present the right one, will lead to high bounce rates and lost con­ver­sion oppor­tu­ni­ties. Con­sumers are rest­less, and if the right infor­ma­tion is not there the first time, they are quick to go else­where. They will rapidly sur­mise, from very few cues, how likely a site will be to sat­isfy their needs.

Recently, Brad Rencher, Adobe’s Senior Vice Pres­i­dent and Gen­eral Man­ager of Dig­i­tal Mar­ket­ing, called for mar­keters to adjust to how “the mar­ket­ing world has become more com­plex, sophis­ti­cated, and fast-moving,” result­ing in the need to address the chal­lenge of “last mil­lisec­ond” mar­ket­ing. To cre­ate the kinds of dig­i­tal expe­ri­ences that engage con­sumers and drive them to act, brands must deliver a per­son­al­ized expe­ri­ence in that instant between an action—a click, a tap, a swipe—and the next step in the consumer’s jour­ney. The dif­fi­culty of achiev­ing this is cer­tainly famil­iar to those of us who think about how our brand presents to dif­fer­ent psy­cho­graphic seg­ments, or about which offers would be most com­pelling to indi­vid­u­als based on their unique pur­chase his­tory, or about any other num­ber of expe­ri­ences we might tar­get to indi­vid­u­als based on the like­li­hood of their dri­ving con­sumer behav­ior. As clear as the oppor­tu­nity is if mar­keters can get it right, it’s equally evi­dent that the chal­lenges to achiev­ing this suc­cess are significant.

In fact, it’s only in the last few years that mar­keters have gained access to tech­nol­ogy pow­er­ful enough to take on this chal­lenge and aspire to make the most of each moment a vis­i­tor spends onsite. Com­put­ers and net­works are now fast enough to eval­u­ate the avail­able data and make the best deci­sion on what to present to a con­sumer, in a times­pan that is barely per­cep­ti­ble. More­over, sophis­ti­cated machine learn­ing algo­rithms can ensure that over time, mar­keters are able to antic­i­pate, and suc­cess­fully respond to, visitor’s inter­ests with ever-increasing accuracy.

The recent release of Adobe Tar­get Pre­mium pro­vides mar­keters with the lat­est data sci­ence tech­nol­ogy and algo­rithms to deliver auto­mated, per­son­al­ized results at the last mil­lisec­ond to a mul­ti­tude of devices, over a vari­ety of band­widths and chan­nels, wher­ever the cus­tomer may be. Specif­i­cally, Target’s auto­mated per­son­al­iza­tion capa­bil­ity addresses the chal­lenge of how to deter­mine and deliver the right, tar­geted expe­ri­ence to each con­sumer, based on every­thing known about that indi­vid­ual at the point of interaction.

The fig­ure below illus­trates the basic steps of the auto­mated per­son­al­iza­tion algo­rithm. First, Tar­get gath­ers all rel­e­vant data about a con­sumer for pro­cess­ing and analy­sis. Each time con­sumers visit a social media page, dis­play ad, or web­site, they gen­er­ate poten­tially use­ful data. Auto­mated per­son­al­iza­tion will sift through the data and learn over time which vari­ables are most pre­dic­tive of how a con­sumer will respond to dif­fer­ent experiences.

SHEINBERG-image 1

Using on a wide vari­ety of these pre­dic­tive fac­tors, it gen­er­ates a “user-based score” that rep­re­sents the like­li­hood of a response to an expe­ri­ence. Then, a “gen­eral score,” reflec­tive of how the over­all pop­u­la­tion is respond­ing to dif­fer­ent expe­ri­ences, is com­bined with the individual’s user-based score to cap­ture the right weight­ing between the individual’s past behav­ior, and the “wis­dom of the crowd” in pre­dict­ing how that per­son might respond to dif­fer­ent expe­ri­ences in the future. Finally, a machine learn­ing approach known as the “multi-armed ban­dit” ensures that auto­mated per­son­al­iza­tion con­tin­ues to test and learn on an ongo­ing basis.

SHEINBERG-image 2

Those learn­ings are reg­u­larly pushed out across Target’s global server net­work, where the rules for the user-based score are updated every six hours and the rules for the gen­eral score are updated every 10 min­utes. This ensures an ongo­ing cycle through the fun­da­men­tals of last mil­lisec­ond mar­ket­ing: 1) lis­ten (gather data), 2) ana­lyze and pre­dict, 3) make deci­sions, and 4) deliver a per­son­al­ized mes­sage, at scale and with a split-second speed that can only be achieved through the most cutting-edge technology.

Of course, in the end, all the data sci­ence and com­put­ing power in the world won’t make a dif­fer­ence if the prod­uct sits idly on your shelf. That’s why auto­mated per­son­al­iza­tion was designed with usabil­ity and marketer-friendliness as defin­ing fea­tures. Ulti­mately, the tech­nol­ogy is just an enabler. The com­pa­nies that get the most out of it will be those with savvy mar­keters who eval­u­ate the results, refine the strat­egy, and adjust the imple­men­ta­tion to bet­ter con­nect with con­sumers and drive busi­ness objec­tives. To that end, Tar­get offers a work­flow that makes it even eas­ier to set up and run one of these activ­i­ties than it is to cre­ate an A/B test, and includes report­ing capa­bil­i­ties that high­light the insights gen­er­ated by the under­ly­ing math­e­mat­i­cal models.

To learn more about auto­mated per­son­al­iza­tion in Adobe Tar­get, or to con­tact us, visit our site.

0 comments