Situation
I joined the Drift team during a crucial phase when we were working hard to find product-market fit. The team had just introduced Lead Bot, a sales-centric chatbot that customers installed on their websites to generate leads for sales. The release of this feature was exciting to customers because it provided them with a new channel for pipeline generation.
Lead Bot -- the original Drift chatbot builder, circa 2018, when I had just joined the team
Iterating on our flagship product
Given the enthusiasm surrounding Lead Bot, our primary objective was to grow adoption. It quickly became apparent to us that the chatbot builder user experience was a barrier to adoption, as it contradicted our users’ mental models. The user interface presented the conversation flow in a linear fashion, even though we know that conversations are anything but linear.
We’d heard anecdotal feedback that customers envisioned a decision tree builder, but I wanted to validate this before jumping in. First, I deconstructed several customer chatbots, which helped me identify and visualize the shortcomings of the existing UI. Then I conducted interviews with customer-facing teammates and customers to understand pain points and gather feedback on low-fidelity mockups.
Deconstructing customers' chat flows
Scrappy proof-of-concept
At this time, new competitors were emerging and we wanted to be first to market with this new user-friendly and powerful UI. Aside from being a challenging design task, our team was very aware and mindful of the significant technical investment that this would require. So, to mitigate technical risk and provide value to our customers as quickly as possible, we opted for a scrappy and incremental proof-of-concept for the first version we built, even though there were some clear usability issues we’d have to address sooner rather than later.
First incremental update -- prove out that we can render lines connecting questions
Early learnings set us up for a successful second iteration
Because our approach accounted for translating existing chatbots into this new layout, we were able to switch our customers access to this new version overnight. Through this, we received an abundance of useful feedback. For one, we learned that the side drawer was too disorienting, and not at all conducive to editing. Another important bit of feedback we received was the ability to preview as much of the conversation as possible without having to click into a modal.
All of this feedback was invaluable as I began designing the next iteration of the chatbot builder. As we continued to get strong signals that customers were craving decision tree builder with path branching, the substantial technical investment was becoming more apparent, and we had some important (and expensive) decisions to make. To ensure the new UI was thoroughly validated before development began, I worked on countless iterations and conducted dozens of usability testing sessions. Nailing down interactions was incredibly important with this project, so I focused much of my time there, continuously testing the designs with users and soliciting technical feasibility feedback engineers. When all was said and done, I had designs that the whole team was confident about.
Customers couldn't get their hands on it fast enough
After a brief closed beta, we officially released the new chatbot builder in an open beta in May 2019, and customers were very eager to start using it. 98% of organizations opted in to the UI within 4 months of the initial rollout. While we have made subsequent improvements since then, this UI remains an important part of the daily workflow for Drift customers.
