Soon after launching the Gemini ad platform with Yahoo!, we saw a proliferation of -- what sentiment surveys indicated were -- low-quality, robotic ads that weren't connecting with our customers in a meaningful way. In turn, ad engagement metrics were on the decline, and I felt the need to experiment with new ad formats. The goal: Set a new, higher bar for our ad formats (maybe win some awards along the way) that would measurably improve engagement metrics between our end-user customers and our brand customers, with a focus on mobile native.
Leverage native app features to create nuanced conversational experiences between brands who advertise on our platform and the customers who value connecting with them.
Yahoo!'s acquisition of Tumblr provided a new, high-engagement user flow that could facilitate a conversational ad flow. The Tumblr app had an extremely high open rate with push alert's turned on when users were given signals to their post likes. Hooking into this mutually beneficial behavior, the concept above illustrates how top-tier brands would show up as a real Tumblr "user", with their own real Tumblr account who likes their content.
To ensure this signal felt authentic, the brand and end-user content alignment required high relevance. The back-end that powered this match-making, started by indexing the user-created tags. A similar dataset was created for brands based on their own content on the Tumblr platform. When the user-generated content had a high-confidence relevancy match for a brand offering available, tapping the brand "like" would reveal the offering in a delightfully human way. Once tapped, the brand doesn't necessarily show a traditional ad format but could layer on top of the end-user's content as shown in the far-right panel.