In the age of Big Data, there’s no short­age of his­tor­i­cal infor­ma­tion to crunch. You can spend count­less hours ana­lyz­ing vis­i­tor behav­ior, past pur­chases, cus­tomer rat­ings, and more—but smart mar­keters don’t stop there. They lever­age all that data to make pre­dic­tions about the future, and then opti­mize accordingly.

The most widely known exam­ple of pre­dic­tive ana­lyt­ics is Ama­zon. Using past cus­tomer pur­chases, page views, reviews, and demo­graphic info, the site offers tar­geted rec­om­men­da­tions for cross-sells and up-sells. It also adds site-wide ana­lyt­ics to the mix, with its “What Other Items Do Cus­tomers Buy after View­ing this Item?” Accord­ing to Ven­ture Beat, these auto­mated rec­om­men­da­tions gen­er­ate 35 per­cent of the site’s revenue.

Ama­zon isn’t the only site that can suc­ceed with this strat­egy. Any busi­ness can mine his­tor­i­cal data—from trans­ac­tions to sur­veys to social media posts—to antic­i­pate what prospects and buy­ers are likely to do next. And prod­ucts are just the spring­board: auto­mated rec­om­men­da­tions can also be used to serve up arti­cles, trend­ing news sto­ries, videos, and other media. Of course, it’s eas­ier when you have the right tools.

Maybe you already have prod­uct rec­om­men­da­tions on your site—but are they work­ing as well as they could be? Below are some of the essen­tial capa­bil­i­ties to look for when choos­ing a rec­om­men­da­tion tool:

Full automa­tion: Prod­uct sug­ges­tions should be auto­mat­i­cally gen­er­ated based on a blend of his­tor­i­cal and real-time vis­i­tor data, as well as global site data. Tools like Adobe Tar­get offer auto-optimization deliv­ery, which auto­mat­i­cally serves up the top-performing prod­ucts to the seg­ments most likely to pur­chase them.

Mar­keter con­trols: You don’t want to have to rely on your tech sup­port team every time you need to tweak your prod­uct rec­om­men­da­tions. Built-in mar­keter con­trols allow you to change algo­rithm set­tings or prod­uct lay­outs on the fly, all in a sin­gle user-friendly inter­face. No pro­gram­ming or train­ing required.

Effort­less A/B test­ing: Inher­ent test­ing is essen­tial to ensur­ing that you’re using the right algo­rithms and most effec­tive place­ments for your rec­om­men­da­tions. With built-in A/B test­ing, you can com­pare results and answer impor­tant ques­tions, such as whether it helps or hurts to dis­play cross-sells in the shop­ping cart.

Mul­ti­chan­nel ver­sa­til­ity: Cross-sells and up-sells should per­sist across all of your mar­ket­ing chan­nels, from your web­site to email cam­paigns to mobile apps.

Easy inte­gra­tion: You want a rec­om­men­da­tions engine that inte­grates seam­lessly with your other tools. For instance, if you’re using Adobe Ana­lyt­ics for data report­ing, you can bring it into Tar­get for seg­men­ta­tion and tar­get­ing. Tar­get even plays nicely with third-party tools.

In today’s data-driven mar­ket­place, pre­dic­tive ana­lyt­ics is the future of mar­ket­ing. Using auto­mated, real-time rec­om­men­da­tions, you can turn yesterday’s insights into today’s conversions.