Find Harmony Among Your Data, Technology, and Customers with Custom ML/AI Enabled through CDP
Duluth Trading Co. is an American outdoor and workwear retailer, focused on triple-stitching a better business through humanity, community, and sustainability. They create sustainable products for hard-working, hands-on people for the love of doing. Duluth engaged with Bounteous to embark on a comprehensive digital transformation through a Customer Data Platform, with an end goal to improve their customers’ overall digital experience, increase engagement, and drive revenue. Through the implementation of multiple Adobe Experience Cloud solutions (Adobe Experience Platform + Adobe Campaign Standard), Duluth and Bounteous have focused on personalization and insights to drive toward realized value.
Bounteous built and deployed a custom product recommender using machine learning to drive 1:1 personalization with refined audience segmentation and content targeting to improve email campaign relevance and engagement. This advanced model allowed Bounteous to contribute to Duluth Trading Co.'s high-level objectives including amplifying existing client wallet share, increasing individual basket size, improving sales conversion, and expanding customer reach & awareness.
Historically, Duluth viewed their customers in siloes based on email segments, browsing history, and events and did not have a 360-degree view of their customers or a unified customer profile. Because of these silos, the experiences they delivered were static, lacking the flexibility to shift with their customers. Data was being collected, but they lacked the digital flow to act upon their data while continuously learning.
Duluth had implemented AEP as their Customer Data Platform, but now needed to act upon their Adobe investment to see the full return on their platform investment.
In doing so, Duluth had several goals to keep in mind. Duluth wanted to leverage the Unified Profile and all of its data to expand the reach and relevance of their email program. In their existing Adobe Campaign workflows Duluth wanted to create personalized recommendations and build out more Dynamic Content Blocks. Finally, Duluth wanted ONE model to be versatile enough to use for omni-channel orchestration.
However, Duluth had multiple challenges when activating their Customer Data Platform. Duluth was using Adobe Target’s out-of-the-box Product Recommender Model for onsite personalization, but there was no obvious approach to extend this level of personalized content into their email program. In Adobe Campaign, Duluth was using complex business rules, but these rules were not sophisticated enough for their use case and were a challenge to maintain at scale. Additionally, to measure the value of these personalization efforts, Duluth needed A/B testing that would allow for the preservation of test vs. control audiences across multiple sends (something that extends beyond the capabilities of ACS).
Through a strategic discovery at the onset of the engagement, we determined that we would apply the machine learning model to Duluth’s most prominent marketing channel first: their email program. Bounteous then built a custom product recommender model in AEP, leveraging the Unified Profile and all purchase history stitched together by user. The product recommender model leverages historical user behaviors including both online and offline purchases and recommends up to 20 relevant products for each of Duluth’s 4.4MM customers. The model applies various filters downstream that even further customize the products being shown to match the different intents of the various email campaigns at hand. Because results are fed into the Unified Profile, Duluth can activate across multiple channels in the future.
Deploying this model in AEP not only enabled layering personalization capabilities within a dynamic email but also the added capability for A/B testing. We had two use cases where we leveraged A/B testing capabilities to determine the success of the model in-market. We structured A/B testing within AEP to designate each user randomly into a test vs. control group and maintain each user’s designation over time. Thus, if the test email campaign extends over various sends, each user receives the same experience every time. Not all tools with built-in A/B testing capabilities will do this, but our solution removes confounding signals and bias from the testing results.
The first use case was a Daily Promotional Email Campaign with a Dynamic Banner surfacing relevant products for each individual user. This campaign has a wide breadth and drives upper-funnel engagement and awareness of Duluth brands and products. This test measured the impact of our product recommender versus no personalization, which drove 10% more users to purchase over non-personalized content, leading to an additional $245k in revenue from just three email sends.
The second use case was a Post-Purchase Retargeting Email Campaign with a Dynamic Banner surfacing relevant products for each individual user. This campaign has a smaller reach, but drives repeat engagement for users to purchase more products, more frequently. This test measured the impact of our machine learning model up against a rules-based model, which drove a 5% lift in basket size over rules-based personalization for users who purchased.
Now, we have proven ROI on using advanced machine learning techniques in personalization, as well as proven incremental value of these sophisticated techniques over business rules. We have also demonstrated the value of integrating AEP with ACS to optimize the capabilities of the two platforms. This is only the beginning for Duluth on their journey to finding harmony among their data, technology, and customers with custom ML/AI-enabled through CDP.