Case Study: How Barkyn Doubled Revenue from Retargeting with Velocidi

A DTC brand takes machine learning by the horns

Barkyn is a direct-to-consumer (DTC) pet-care company. Like many DTC companies, social media is a primary acquisition channel for them. However, audience insights available on those platforms are too opaque to get a good understanding of the buyer journey. As a result, trying to get the best possible ROI from those platforms is often a matter of guess-work.

So, when Barkyn implemented their Velocidi private CDP, our first job was to help them take campaign ROI into their own hands.

Actually, this case study was born out of the standard onboarding process that Velocidi provides every client. Testing machine-learning audience segments in real campaigns is how we make sure the CDP is delivering ROI as fast as possible. For Barkyn we decided to test out the CDP on their retargeting campaigns.

To be more specific, Barkyn’s team used the built-in “likelihood to buy” machine-learning model for this retargeting experiment. Our team guided them to make sure the machine-learning model was being trained properly using their first-party data. It took about two weeks to produce consistent and reliable results, and soon enough, Barkyn was able to create audience segments based on each website visitor’s predicted likelihood to buy.

By the way, Barkyn is actually a smaller brand than most people would assume is able to make use of machine learning. It’s true that a high volume of data is the first requirement for for machine learning to be meaningful. However, a DTC brand’s direct relationship with their customers makes them naturally data-rich. All they need is the tools to make their data actionable.

Higher relevancy leads to higher sales volumes

Once the machine-learning audiences were ready, we did an A/B test on a Facebook retargeting campaign. An unsegmented audience of website visitors was used as the ‘control’ group. An audience pre-segmented using Velocidi’s machine-learning model served as the ‘test’ group.

After two iterations of the campaign, the machine-learning audience had produced double the return on ad spend (ROAS). And we also saw 1.9 times the sales volume, and 5 times the purchases per 1000 visitors compared to the control group.

The success of the machine-learning segment is attributed to the fact that ad delivery was focused on only the most relevant audience. The size of the audience receiving ads in the test group was only one-third the size of the control group. This not only led to higher conversion rates, but higher total sales as well.

The full case study is available for download below, and an abridged version can be accessed at www.velocidi.com/case-study-barkyn.