Training the Data

Barkyn is a Portugal-based DTC pet food service that offers its customers a monthly subscription service. Like any number of similar services in other categories, customers (the pet parent, as opposed to the furry end user) receive their personalized assortment of goods based on previous orders. They can always adjust the order if they choose. In this subscription model, customers generate a steady flow of revenue over time. For Barkyn, acquiring new customers comes with a big payoff.

Andre Jordão, CEO of Barkyn, recently told me: “Barkyn is a fast growing brand. We are expanding across multiple EU markets. Our main product offering is personalized subscription boxes of pet care items, and we try to bring a personalized touch to every part of our business, including marketing.”

In the DTC space, social media plays a significant role in acquisition. Communicating over social provides the personal and authentic engagement that make for successful DTC brands and loyal followers. But for a highly-personalized brand, a walled garden like Facebook also presents a challenge to a brand’s getting to know its customers, and especially prospective customers, better. So, when Barkyn decided to pivot and expand its following, it sought a data-focused solution with New York-based platform Velocidi.

“Basically,” Jordão explained, “if we make better use of our data, we serve our customers better than anyone, which is our ultimate goal. We wanted to jump in head first with machine learning and make our data useful for as many marketing objectives as possible, including increasing retargeting efficiency while reducing ad waste. We also wanted to focus on longer term KPIs such as lifetime value instead of immediate clicks or purchases, and we now have the tools to do that as well.”

To keep up with customer engagement across all channels, Barkyn had Velocidi implement a CDP (customer data platform), which pulled together and unified the data from customers across multiple touchpoints. The CDP includes machine learning models that are trained with historical and updated real-time data that includes customer behavior like cart activity, email opens, email clicks, ad click, logins and other activities.

Paulo Cunha, CEO of Velocidi, told me, “Velocidi has built campaign testing and machine-learning training into the onboarding process for all clients. We want to make sure our clients are getting ROI from their tech investment as fast as possible. In Barkyn’s case, we decided a retargeting campaign was the fastest and simplest way to demonstrate that the system is working properly, and to show how it can be used to improve return on ad spend (ROAS) and conversion rates.”

The real-time and historical data, especially, can inform advertisers about the intent of existing customers and prospects. This, in turn, determines the kind of relevant messaging the brand delivers to consumers. If the message isn’t timed well and relevant, it could push buyers away. But if intent is high, not sending enough ads a consumer’s way could be blowing a missed opportunity.

“Making messaging relevant to a customer’s intent level has partly to do with timing and frequency, and partly to do with the creative messaging,” Cunha said. “A customer who is at a higher-intent stage can be served ads more frequently, with messages that remind them of the purchase they intended to make. Lower-intent customers will be more likely to get annoyed by frequent ads, so they can be served ads less frequently, with offers of discounts or other incentives to nudge them toward the purchase point.”

This additional effort on the part of Barkyn and its data platform allows the brand to curb its spend and make it more efficient over affective social media platforms.

“The private CDP gives Barkyn an in-house system for collecting first-party data from website, email, advertising and e-commerce channels,” Cunha stated. “Then it has built-in machine-learning models sticking predictive behavior attributes to each visitor and customer profile so that those customers can easily be segmented by their marketing team according to their predicted behavior.”

Using this strategy, advertisers don’t have to come up with their own approximations, but instead can rely on the machine learning to make solid, actionable predictions, according to Cunha. For example, Barkyn didn’t have to invent some abstract category, like visitors who have clicked on a product multiple times without actually purchasing it. Instead, they are able to segment visitors based on the likelihood that they would buy, using all the data that had been collected and unified. The segment is crafted in advance of Barkyn taking this actionable data to a social channel like Facebook.

“Once a segment is created, it can be activated directly on Facebook, or any other ad platform,” Cunha said. “The private CDP protects Barkyn’s proprietary data by using anonymized identifiers, and limiting the data exposed to Facebook’s pixel to only the data necessary for the campaign. Barkyn can then go into Facebook Ad Manager and apply their pre-segmented audience to their retargeting campaign.”

He added, “Traditionally you would create a retargeting audience by giving Facebook the entire customer journey and demographic data. This is unnecessary data sharing and just enriches Facebook’s own user insights.”

In the three-week retargeting test, Barkyn used two audiences of equal sizes to A/B test a control segment during a live Facebook campaign advertising dog food products. One audience was the control, while the other was divided into low-intent and higher-intent. The higher-intent segment of the test audience was only about a third of the group, and it was this audience that Barkyn paid to advertise, using a smaller, more efficient ad spend.

The results in reaching this concentrated audience were a 2X increase in ROAS and five times as many sales per 1,000 people reached.

“We now know each individual visitor’s likely future behavior and lifetime value, instead of just having a historical view,” Jordão concluded. “And we can use this to improve the dialog between us and our customers while maximizing the value we deliver to them. That, together with the ability to slice and dice this information across different geographical regions, is having a positive impact on our growth strategy. Now when we are prospecting new audiences, we can figure out much faster who is resonating with our brand and allocate budget in a more meaningful way.”

Total
0
Shares
Related Posts