This is the second post in a 2-part series that examines look-alike modeling in Adobe AudienceManager.

Introduction

My previous post on look-alike (or algorithmic) modeling provided a conceptual overview of how modeling worked and the potential benefits it provides to our buy and sell-side customers.  In this post, I’m going to change our focus and walk through a buy-side use case. I want to talk about how to use look-alike modeling to solve a basic remarketing problem and tie this back to the concept of data/modeling transparency, which is an important difference between AudienceManager modeling and the somewhat more opaque solutions provided by other products.

Retargeting is limited in scope:  A buy-side use case

Just by its very nature, re-targeting is only reaching visitors who have previously been to a site; which means, it’s limited in reach.  Re-targeting is proven to be quite effective, by delivering relevant offers to a set of users who are interested in a specific product or type of product.  However, the critical element in re-targeting is delivering the right creative to the right audience.

Take, for example, visitors to the Adobe.com site.  We have a mix of customers who have bought products or browse the site for information or the community interaction.  Since Adobe offers a wide range of products, and it’s very likely that the same users who purchased or browsed content for the Creative Cloud or products like Photoshop, Illustrator, or InDesign are not the same type of buyers for SiteCatalyst, Test&Target, or Search&Promote.  Looking for creative users (art, design, content, etc.) among analytics customers reduces the effectiveness of your retargeting efforts along with wasting time and money.

Therefore, it is critical to build meaningful segments that can be targeted accurately with the right content – which is the point of this blog.  And, it is also desirable to extend the reach of a re-targeting campaign to become an acquisition campaign.  To do this, we begin with that highly valuable set of anonymous data captured from our site.  It is a great starting point to create look–alike models which will help our marketing team build a retargeting campaign that is accurate with respect to finding the right type of customer and has enough reach to help meet campaign goals.

Build a look-alike model in the AudienceManager UI: Create a baseline

To build a look-alike model that helps find new users, our marketing team would log in to AudienceManager, select Models from the Manage Data section, and click Create New Model.  After providing basic information (model name and an optional description), our team selects an existing trait or segment that already contains the type of users they want to find more of.  Next, they would choose a 30, 60, or 90-day look-back period to set a time range for the model.  Together, the selected trait, segment, and time interval form a baseline for the TraitWeight algorithm.  The baseline is the basic group TraitWeight looks for when it searches for new users in other data sources.

In the illustration above, our marketing team has selected an existing trait, “Creative Cloud Purchaser.”  This contains the type of users our marketers want to find more of in their retargeting campaign.

Select data sources to find new users

In this next step, the marketing team selects the first and third party data sources they want the algorithm to model.  In this case, we want to find users that we haven’t seen on an Adobe property before, so we’re also including data licensed from a third-party data provider.

Once we select a trait or segment, a time interval, and an available data source, the model is ready to start searching for new users similar to those in the selected trait or segment.  When finished, you can create and target new traits and segments with this data in AudienceManager.

Data transparency

Providing our customers with an unmatched level of control and transparency over the model results is central to our look-alike modeling feature.  This makes itself evident in the data users get after the model runs.  In the results, our UI exposes the specific underlying data points we’ve collected about new audiences.  Additionally, it orders the results according to how closely they model the type of customer you want to reach.  These results appear in Trait Builder in a table of influential traits as shown below.

 In the results, AudienceManager gives you access to a new set of users discovered in your selected data sources.  These look-alike modeling results provide you with real feedback in terms of how important, accurate, or valuable various traits are to a campaign.

Compared to AudienceManager, other products just generate results and say, “here’s a new set of users, good luck targeting them.”  Our modeling process goes beyond that and gives marketers the ability to determine how specific they want to be with their targeting.  Is their goal to run an effective direct response campaign?  Well, algorithmic modeling lets them see the most accurate traits they have access to and build new segments with those users.  When the goal is brand awareness, as it is with our Adobe example (we want to reach a lot of new users), marketers can build new segments that contain a desired audience size (up to 25 million).  Note, however, it’s important to understand that accuracy declines as you increase reach.

Data control

With a given result set, a marketing team can build their own segments with this data.  Also, they can take this process to the next level and have AudienceManager build the segments for them, but use radio buttons or a slider control on the results graph (shown below) to retain control over the reach and accuracy of the model.

Final thoughts

Once you choose how many users you want in a new segment (and you can create multiple segments with different size thresholds from the same model), those users are now targetable in real-time.  With look-alike modeling, marketers have a new, powerful tool they can use as part of their acquisition strategies and brand objectives.  Also, our model provides data transparency and choice by giving you access to all the results from each model run.

For more information about look-alike modeling in Adobe AudienceManager, talk to your partner solutions representative.

0 comments