This is the first in a series of posts about algorithmic audience modeling offered by AudienceManager.
Look-alike modeling arrives
I’d like to take a moment to announce the release of look-alike (algorithmic) modeling in Adobe AudienceManager. Look-alike modeling has long been a buzzword, and a set of desired features, within the online advertising community. Indeed, industry professionals have been talking about automated audience discovery since the late 1990s. However, no one really has been able to deliver on the promise of algorithmic data analysis, discovery, and making this information easy to use. True, some products in the ecosystem say they offer look-alike modeling, but most simply offer black-box solutions and don’t provide transparency into their systems. As a result, you may never be sure that data was subjected to a true, algorithmic analysis. However, for those of us working on AudienceManager, we really wanted to take the mystery out of the modeling process, describe how our model works, and deliver results that matter to our customers. AudienceManager modeling is the result. This is a standard feature available to all AudienceManager users.
Let’s review how our look-alike model works and the potential benefits you might get from this new feature
Find new users with TraitWeight
AudienceManager uses a proprietary algorithm called TraitWeight to discover new, unique audience members. This process starts after you select a trait or segment, a time interval, and first or third-party data sources for analysis. As it runs, TraitWeight looks for users in the data sources that are identical to qualified users in your selected trait or segment. Next, TraitWeight weighs the results. Weight numerically ranks newly discovered traits in order of influence or desirability. The scale runs from 0 to 1. Traits weighted closer to 1 means they’re more like the audience in your baseline population. Also, heavily weighted traits are valuable because they represent new, unique users who may behave similarly to your established audience. In the final step, AudienceManager displays the weighted results in Trait Builder, where you can create traits based on the weighted score generated by the algorithm. You can use these results to build accurate traits, or trade some accuracy for reach to help expand audience size.
Advantages of look-alike modeling
Some benefits to working with algorithmic modeling include:
- Data accuracy: The modeling processes run at regular intervals, which keeps results current and relevant. The AudienceManager modeling system is always working for you to find and extract new value from all your data.
- Automation: The model will find new users for you. As a result, you don’t have to manage a large set of static rules for audience discover and creation.
- Save time and reduce effort: With our modeling process, you don’t have to guess at the traits or segments that may work or spend scarce resources (time, effort) on campaigns to discover new audiences. The model does that for you.
- Reliability: Modeling works with server-side discovery and qualification processes that evaluate your own data and the selected third-party data that you have access to. This means you don’t have to see visitors on your site to qualify them for a trait.
Buy-side use case: Extending the utility of retargeting
Now that you have a general understanding of how the model works and its benefits, let’s look at how look-alike modeling might be applied to help an advertiser with audience retargeting.
Let’s not mince words here: advertisers love retargeting. What isn’t to like? It delivers on digital advertising’s promise of finding a very specific set of users. And, the performance for these users is usually off the charts. However, the problem with retargeting is that it mainly reaches existing customers rather than new customers. Also, retargeting typically reaches a relatively small pool of users. So, basically, retargeting does not address the need to find new customers that have never interacted with a particular brand online. AudienceManager’s look-alike modeling helps solve these problems by finding new users who may be interested in a product (accuracy) or by helping expand your potential qualified customer audience (reach).
Finally, let me point out that the real value in retargeting is more about giving marketers insight into how their customers behave rather than the targeting itself. Previously, we’ve had to guess at the answer to the question “What is special about these users that make them my customers and how do I find more of them?” Look-alike modeling can help answer that question for us. In this case, behavior is key.
Sell-side use case: Delivering unique value to the buyer
As much as the buy-side loves retargeting, it is safe to say the sell-side loathes it. Frequently, the first phrase I hear from many publishers (when the topic comes up, especially in relation to RTB), is ‘cherry-picking.’ Historically, advertisers were forced to come to publishers and buy pages that contextually matched the audience they were looking for. However, more recently, many large publishers have moved away from selling content and towards offering audience ad products based on their users’ online behavior. With both the buy-side and sell-side looking to identify a specific audience, it is often difficult to match the users that an advertiser is looking for with the audiences the publishers have built using their own first part data sources. This typically causes the buy-side to revert to basic re-targeting, so they can feel comfortable they are buying the right audience. Look-alike modeling offers a solution to this problem that can work for both sides.
In the end, both the sell-side and the buy-side benefit from look-alike modeling. Publishers can build models and create new audiences from their own data. They no longer have to sell context specific pages or pre-packaged audiences to buyers. This helps advertisers find net new users on a publisher’s site and allows business partners to identify the audiences that perform well. Basically, publishers benefit by expanding reach vs traditional retargeting campaigns and by creating strategic advertising programs for the buyer, a rare situation in the traditionally (somewhat) antagonistic relationship between the buy and sell side.
Follow up post: Implementation steps
Now that we’ve reviewed how look-alike modeling works and how it can help buyers and sellers, we’ll follow up with this post with a step-by-step walkthrough of the model creation process.
All of us on the AudienceManager team are very excited about TraitWeight and look-alike modeling. However, we think this just the beginning, a scratch on the surface of what this and other Adobe predictive analytics solutions can provide. In the coming months we hope to expand these capabilities by improving TraitWeight and adding new algorithms for improved modeling.