Case Study – Barkyn


Download Now

Get the full case study now in a printable format or continue reading below.

The Challenge

As a brand that lives and breathes personalization, Barkyn wanted their retargeting strategy to go beyond the basic assumption of, “You visited our site, therefore you must be a potential customer.” They needed a more precise way to segment their retargeting campaigns in order to deliver relevant messaging to those customers who are most likely to convert.


Company Overview

Barkyn is a direct-to-consumer (DTC) pet food brand. They offer a monthly subscription box of dog food, snacks, and toys, personalized for each client. Barkyn’s online store is also stocked with healthy dog products for purchase on demand.

The Solution

With the implementation of a private customer data platform from Velocidi, Barkyn got access to a unified customer view for each of their visitors including multi-channel visitor behavior such as logins, page views, cart activity, email clicks, email opens, ad clicks, and others.

After training the CDP’s ready-made machine-learning models with first-party data, Barkyn was able to automatically segment visitors whose behavior indicated higher intent to buy. This knowledge enabled Barkyn to activate an audience segment of selected customers in an anonymized format to Facebook Ads Manager.

The Campaign

To test the effectiveness of Velocidi’s machine-learning segment, Barkyn conducted an A/B test on a live Facebook campaign.

The Audience

The campaign was divided into two audiences with an equal budget:

Control group: a sample of website visitors with no machine-learning or
segmentation applied.

Test group: the audience optimized by machine learning to reach only the “high intent” customers.

The Results

The A/B test was conducted twice to confirm validity. In both tests, the test group, using Velocidi’s optimized segment, outperformed the control group.

With the help of their private CDP, Barkyn was able to create a segment that concentrates their most likely buyers in an audience that’s only one third the size of the control group.

Increasing ROAS is an expected result of targeting a smaller audience on the same budget. The high disparity in sales values and purchases per 1000 customers is what really tells the story.

It’s safe to assume that the percentage of “Likely buyers” was roughly equal in the control and test groups.

So, why is the revenue generated by sales 1.9x higher in the test group?

The results of this test demonstrate that using the same budget on select portion of the audience causes the likely buyers to be exposed to a higher frequency of ads, Therefore, applying more ad pressure on only the most receptive audience, rather than engaging everyone equally, leads to higher conversion rates and a higher total sales volume.


Get the full version of the use case with four pages of detailed information.

Start growing your brand now

Get a demo and see how Velocidi takes the guess-work out of marketing by making machine learning practical and accessible.