introducing a chatbot 🤖
to the Spotify support page, using mixed methods

⬇contact rate, 🔝Costumer statisfaction

Problem

With over 300 MAU (2020) users, you can imagine Spotify receives a tremendous amount of support requests every day. They wanted to lower this amount, while keeping the similar high quality of service and resolution rate by introducing a chatbot. 

Solution

By using multiple research methods I guided the team step by step on where to put their focus. 

From understanding the users' attitude towards different topics, to concrete flows, to design and interaction design. 



Impact

🦉Helped the team develop more user centered (instead of technology first).

⏱️ A better customer satisfaction score then before the chatbot, since simple problems are quicker resolved. 


How to build a great chatbot 

A lot of this work falls under NDA, so I can not write or show as much as I would like to: but let's give it a go, since this has been one of my most succesful projects up to this date. 

Between you and me, I was not really sure what to deliver on this project when starting. The team (6 people) working on this project was based in Cambridge, and were convinced that a chatbot would help them, but wasn't sure how to make it work. Instead of creating one big boom, I focussed on different rounds of research influencing the roadmap (and my own research) along the way: creating sparkles along the way ✨

Team: Kim

Method: Qualitative interviews (London)

Delivery: 5 info posters + presentation for the CS department


2. Understand what topics do users want to get help on via a chatbot?

Team: Kim, UX writer

Method: Card sorting & qualitative interviews (Cambridge)

Delivery: 5 info posters + presentation for the CS department 


3. Location of the chatbot

Team: Kim, UX writer, CS rep, Digital support specialist

Method: Participatory design (London in UX lab)

Delivery: 3 info posters descibing the userflow (*learnings) + presentation for the CS department  + socialising the results + next steps meeting

4. User Experience of chatting with the chatbot

Team: Kim, designer, copywriter, CS rep, Digital support specialist, developers, PM

Method: usability testing + wizzard of ozz technique (London UX lab)

Delivery: insights poster + presentation for the CS department  + socialising the results + concrete suggestions on what to improve

5. Optimise the UX of [specific flow]

Team: Kim, designer, copywriter, CS rep, Digital support specialist, developers, PM

Method: usability testing + wizzard of ozz technique (London UX lab)

Delivery: posters + presentation for the CS department  + 2 day workshop to create a final delivery that could be implemented (incl timeline)




Learnings

As you might have noticed, the team (and impact) grew on the way. This team was new  to UX research , and research by research more members wanted to partake in designing and observing the results. I am very proud of how I impacted the team over time, but also of all the positive feedback I have received: without my help the product would not have been as  great as it is now. 

Rome isn't build in one day, and neither is a chatbot. By running smaller rounds of research I could help the team design  the whole roadmap. We for example pivoted and zoomed in after research  round 2: making a clear impact on the roadmap (and results). Now looking back I wish I could run the research even faster (but that is easier said then done afterwards). 

*Leaning on creating visuals: I created a quick and dirty visual to discuss with the team, before I shared out the nicely polished posters. That ugly visual has hunted me for another 2 years: the insights were so great that they were circling to the VP of CS and beyond. From now on there is WIP on my sketchy visuals to avoid this 😅