In the con­stantly chang­ing world of dig­i­tal mar­ket­ing, seg­ment­ing your audi­ence using rules based on broad assump­tions or unsub­stan­ti­ated insights can work against your suc­cess. Unfor­tu­nately, many mar­keters rely only on lim­ited rules-based tar­get­ing sys­tems, with­out real­iz­ing that bet­ter options exist.

The real­ity is that given today’s com­plex rela­tion­ships between busi­nesses and their cus­tomers, tra­di­tional rules-based tar­get­ing sys­tems are often not pow­er­ful enough on their own to give cus­tomers richer, more rel­e­vant expe­ri­ences when mov­ing through a dig­i­tal universe.

We all know that even cus­tomers who might appear sim­i­lar on the sur­face can ulti­mately have wildly dif­fer­ent inter­ests when it comes to what they buy and what prompts them to make a pur­chase. Instead of rely­ing only on rules-based sys­tems, mar­keters also need to tap into more auto­mated, behavior-based tar­get­ing solu­tions built on pow­er­ful algorithms.

So what’s the dif­fer­ence? And why is at least know­ing a lit­tle some­thing about the under­ly­ing algo­rithms so important?

Adobe Tar­get incor­po­rates machine learn­ing to give mar­keters a clearer, nuanced pic­ture of site vis­i­tors. In real time, algo­rithms in Adobe Tar­get look for as much infor­ma­tion as pos­si­ble about a vis­i­tor so the sys­tem can instantly dis­play rel­e­vant, mean­ing­ful con­tent. It takes into account a mul­ti­tude of infor­ma­tion to deliver just the right con­tent, whether the sys­tem is con­sid­er­ing an individual’s past ses­sion behav­ior, basic demo­graphic infor­ma­tion, par­tic­i­pa­tion in loy­alty pro­grams, spec­i­fied inter­ests, or other pos­si­ble influencers.

A lot of dig­i­tal mar­ket­ing tech­nol­ogy ven­dors offer rules-based tar­get­ing sys­tems that dis­play cer­tain con­tent to vis­i­tors based on assump­tions. For instance, a web­site vis­i­tor from Cal­i­for­nia might be greeted with an ad for sun­screen or shorts, while some­one from a colder cli­mate might only see jack­ets and other gear to keep them warm. Unfor­tu­nately, this approach treats large cus­tomer seg­ments the same, shut­ting out poten­tial oppor­tu­ni­ties. There’s no real intel­li­gence behind the automa­tion, even if ven­dors claim oth­er­wise. To make rules-based tar­get­ing work bet­ter, mar­keters have to spend a lot of time defin­ing rules and com­ing up with hypotheses—which is impos­si­ble when deal­ing with a vir­tu­ally lim­it­less num­ber of cus­tomer back­grounds, inter­ests, etc.

It’s impor­tant to note that rules-based tar­get­ing isn’t a bad thing—we employ the tech­nol­ogy in many of our activities—but dig­i­tal mar­keters should know it’s only one of the things they should be doing as part of their site opti­miza­tion pro­gram. They need more auto­mated mod­el­ing capa­bil­i­ties as well.

With auto­mated tar­get­ing built on machine learn­ing, mar­keters don’t need to man­u­ally man­age and build mod­els, and instead can spend their time on more valu­able activ­i­ties. Even the staunchest pro­po­nent of man­ual inter­ven­tion and con­trol would find it hard to argue the fact that for many mar­ket­ing activ­i­ties, machines are much bet­ter equipped to quickly move through thou­sands of vari­ables and make cor­re­la­tions between things that—at least to us—could seem com­pletely unre­lated. Machine learn­ing also has the advan­tage of get­ting bet­ter at learn­ing over time; tar­get­ing accu­racy improves as data is col­lected and analyzed.

We know empow­er­ing our cus­tomers to make smart deci­sions relies on our suc­cess at giv­ing them smart tools built on pow­er­ful algo­rithms. To help us stay on the lead­ing edge, our devel­op­ment teams reg­u­larly par­tic­i­pate in rig­or­ous aca­d­e­mic mod­el­ing com­pe­ti­tions, as well as con­tin­u­ally refine our tar­get­ing solu­tions in our in-house labs.

A good dose of healthy com­pe­ti­tion with some of the world’s lead­ing devel­op­ers is a good thing.  As is our focus on doing right by our customers—and your customers—which leads us to want­ing to help you under­stand what and when rules-based ver­sus more machine-based tar­get­ing is best. We’ll con­tinue to explore this topic fur­ther in another upcom­ing blog.