What do master composers and algorithms have in common? They both master the technicality and precision required to make a complex system work seamlessly.

With our latest version of the BidIQ algorithm, named Chopin, timing is everything. Since our last release, Beethoven, our data science team has been hard at work figuring out the best time to serve an ad to a user, and when it’s the right time to stop serving ads.

As a result, campaigns have been much more efficient — we’ve increased click-through-rates by 39% and lowered CPCs by 35% on average over the last 4 months.


Our customers have noticed the difference.

OneStop, a leading digital agency that represents premium retail brands such as Frye, 7 For All Mankind, and Dolce Vita, drove an average 40% lower cost per acquisition for their advertisers. Click-through rate almost doubled, while CPC went down by 28%.

“AdRoll’s performance is driving strong incremental sales for our retail brands. We’re really happy with the ROI we’ve been able to drive through AdRoll’s platform, all while making sure we’re being smart about brand safety and frequency capping.”

– Josh Miguel, Digital Marketing Manager at OneStop

Through advanced machine learning, Chopin creates smart user models to understand the relationship between site activity, frequency of ads shown, and propensity to buy. As a result, campaigns are more efficient by reducing wasted impressions and reallocating budget to users who are more likely to click and convert.

We’re using some of the most advanced machine learning methods in the industry to drive better performance across our platform. Read on for a deep dive on some of the modeling we’ve implemented in Chopin to drive these performance improvements.

Page category modeling

This feature allows BidIQ to better understand the “digital body language” of users as they browse through your site. BidIQ builds up to ten million models of how a site visitor behaves, and through advanced machine learning figures out which of those models is most likely to click and convert. After users leave your site, BidIQ tunes its bidding according to each cohort’s particular browsing behavior on your site.

Intelligent frequency capping

You don’t want to overwhelm customers with your ads. We’ve further refined our understanding of ad fatigue at an individual user level, so we can taper the amount of ads to those who are unlikely to buy. This improvement helps brands’ reputations while using the saved budget to reach more potential customers.

Factorization machines

This is a sophisticated machine learning method, better explained by Matt Wilson, one of our data science engineers. “The FM model allows us to consider interactions between all predictive variables, rather than only certain manually selected interactions. As a result, we have seen that our advertisers are collecting almost 7% more clicks for only 2% more cost.” Translation: we’re now considering exponentially more variables about a user when calculating a bid, and it’s allowing AdRoll to be more efficient with your media dollars.

These are just a few of the many updates we’ve made with the Chopin release to make your campaigns perform better than ever. We’re constantly improving our bidding algorithm to help deliver the most efficient, brand safe campaigns possible.

To stay up to date on the latest AdRoll products, as well as industry news and trends, subscribe to the AdRoll blog today!: