An Overview of Predictive Targeting

Read the full transcript

Today we’ll review Dynamic Yield’s new Predictive Targeting solution. A machine learning engine that helps marketers maximize performance and revenue from their personalization campaigns while saving valuable time and effort. Getting started with Predictive Targeting is really easy. Let’s take a quick look. In this example, we’re testing the impact of a mobile notification on revenue. We’ve set up three tests, each targeting a different area of the mobile site. And each set up just like any ordinary A/B test. We’ve already set up the tests for the product and category pages, each consisting of one message variation and a control group that doesn’t receive the message. We’re now finalizing the home page test. We set up the targeting rules, confirm that our variations are in place with the desired traffic distribution, select our KPI and launch. The Predictive Targeting algorithms kick in the moment any test is launched. They run behind the scenes, continuously analyzing the performance of all traffic segments for each test variation, identifying the optimal targeting configuration for the variations in the test. Now that our test is live, Predictive Targeting can analyze our data in real time. Once opportunities are identified, Predictive Targeting notifies us by email, as well as in the test reports, giving us the control to take quick action. But before we move forward, let’s take a look at the test results. In the test report, we see the full breakdown of the results with an indication that the message variation is the overall winner, projected to generate a 1.8% uplift if rolled out to all visitors where Predictive Targeting identified a personalization opportunity that would achieve an even greater uplift. According to Predictive Targeting, if the message were rolled out to all traffic, excluding the Sale Lovers segment, we would see a 13.6% uplift. Predictive Targeting’s results indicate that the Sale Lovers segment is an exception to the rule and performs better without being exposed to our message. We can go ahead and apply the recommendation immediately but since personalizations shouldn’t be a black box, we can access a breakdown report that shows us Predictive Targeting’s full analysis. After that, we can hit the Review Before Publishing option to make sure that all targeting rules are applied correctly. As seen, the new targeting conditions are the same as before with the addition of the Sale Lovers audience exclusion. Within the variation allocation, the message is now targeted at 100% of the users who meet the new Targeting conditions. Now we launch to see the uplift from personalization.

Walk through real use cases for Predictive Targeting, a new AI-Powered solution by Dynamic Yield, which allows businesses to automatically convert A/B test results into effective personalization opportunities.