Retargeting can be an expensive undertaking and the chances of success are different for different users. Not all visitors are equal. The predictions of our artificial intelligence can help to classify users based on behavior into users who are very likely to convert, likely to convert, and unlikely to convert.
This classification can be used to fill segments in other platforms like Facebook, Yahoo Gemini, Microsoft Bing Ads, and many more. Once you have segments based on their likelihood of conversion you can choose how much you want to bid on these users or how you want to modify the bid, e.g. you could increase the bid on users who are very likely to convert, leave the bid unmodified for users likely to convert and stop bidding on users unlikely to convert. Or you could stop bidding on users likely to convert, since they might convert without any additional advertising, bid heavily on those with a medium conversion probability, and place small bids on users with low conversion probabilities, just hedging your bets. The right strategy depends very much on your business.

Besides the conversion probability, our artificial intelligence also provides predictions about customer lifetime values and next basket values, giving you further options to evaluate users to modify your bidding behavior on different platforms.

This matrix along the categorizations of expected customer lifetime value and conversion probability visualizes a possible bid strategy. High-value users receive higher bids in general, accentuated based on their conversion probability, whereas mid-value users receive lower bids and low-value users receive the lowest bids, each time modified by their conversion probability.
In order to make this work, the segments need to be transferred to different platforms via Audience Stream and you'll have to create the segments and or lists on the target platform and post the corresponding remarketing pixel in the audience stream configuration. Once the audience stream is configured, you can populate them with the segments. If 9 segments seem too much, you can also start with four segments and just classify high/low conversion probabilities vs. high/low customer lifetime value and leave everything in the middle unmodified.

Caveats

Audience streams are transferring the information to the target platform when the user is active on your website, the transfer does not work for users, who are not coming to the website anymore. Predictions are also available in real-time based on the current session, so that you can immediately classify new users based on their current ongoing session, although, of course, predictions based on more than one visit are usually more accurate.