Lookalikes

Lookalike stats of a segment can be seen by clicking the cog icon next to each segment in the list view or detailed view, and then selecting "Lookalike".



In the lookalike screen you can create lookalike segments by choosing a similarity rating. We encourage you to test and identify your ideal balance between reach and performance. To do so, you can create multiple lookalike segments of the same base segment using different similarity ratings.

Some segments will not show Lookalikes. Reasons for that can be found here:
https://konsole.zendesk.com/hc/en-us/articles/115001099847-Lookalike-Modeling-Availability-FAQ


Below is a breakdown of the 5 Key Differentiators of Salesforce DMP Lookalikes:

1.  Algorithms.  
Salesforce DMP's team has a long history of developing algorithmic approaches to high-dimensional problems in pricing and inventory prediction where you have lots of data, lots of noise, too little signal.  Our approach to Lookalikes draws on several of the techniques we've used along the way for statistical inference and probabilistic modeling.  

2.  Tunability.  
Other methods for Lookalikes follow the "You get what you get and you don't throw a fit" approach:  users receive a pile of lookalikes without the ability to tune or control them in any way. Salesforce DMP Lookalikes puts modeling control directly into the hands of our users. You can decrease or increase reach by turning your statistical confidence level up or down respectively.  No PhD-level handling of thetas, sigmas, or other messy statistical parameters required.

3.  Interpretability.  
Other lookalike methods are black box, in the sense you're given an increased pool of cookies against which to target your campaign, but without the ability to persuade yourself that they make sense, much less the ability to convey their value to an advertising buyer. Salesforce DMP doesn’t turn back a pile of cookies; we give you segments that you can unpack and understand at the most granular level.  Because of our use of Bayesian methods, all of our lookalikes are interpretable in terms of their constituent parts – the rules and attributes comprising an audience segment.  

In simpler terms:  The other guys give you binaries, a bunch of 0's and 1's; Salesforce DMP gives you source for evaluation and iteration on the segments.

4. Actionability.  
Because other solutions rely on outboard partners (think Datacratic) for their Lookalike results, you have to wait for the back-and-forth between your DMP and their partner before you can put your Lookalikes to work.  At Salesforce DMP Lookalikes are native to our platform, so there are no hops, no delays, no latencies.  You can take immediate action on any Salesforce DMP Lookalike for O&O, extension, or connection to partners.

5.  Higher Signal-to-Noise Ratio by Design.  
Because other DMP's have outsourced their Lookalikes to partners, the resulting computations have only as much signal as the DMP passes to the lookalike specialist. Salesforce DMP computes Lookalikes based on clicks and conversions, but also based on all the data and all the traffic we see, across the 1.2B global uniques flowing through our platform.  That's a gigantic amount of signal upon which to train our lookalike algorithms.  

Moreover, the datasets we draw on include but are not limited to:  (a) first-party page and user attribute data; (b) programmatically generated first-party segments based on browsing behaviors, search keyword history, event pixels, etc; (c) geo and technographic data; (d) second and third party sources for additional validation.

Have more questions? Submit a request

0 Comments

Please sign in to leave a comment.