Information/content design, UX Research, Personas, Discovery, Visual Design, Visual Storytelling — ongoing project
At the end of a ride Lyft passengers have the option to leave a comment and/or “flag” drivers in four problem areas: navigation, safety, cleanliness, and friendliness.
Lyft asked me to develop a scalable approach for training drivers in these areas, hoping to drive a reduction in the number of flags drivers receive. I worked in collaboration with Julia, the program manager who spearheaded the initiative inside Lyft.
1. What’s the trouble
To gain a deeper understanding of the reasons for flags, I looked at passenger comments. An analysis of 4,089 passenger comments containing a flag in the span of 1 week across all geographical areas yielded the following takeaways:
- Knowledge of city and navigation is one of the biggest issues. 17% of comments stated that the driver was “unfamiliar” with the area (“long ride,” “driver felt unfamiliar with the area,” and “wrong turn” were most common navigation flag comments)
- Bad behavior affected ratings more than bad navigation alone. Comments that cite navigational, as well as behavioral and trust issues such as being “unfriendly”, “running the meter” or having “bad intentions” are correlated with lower rated rides than comments that only cite navigation as the issue
- Pickups are a major pain point, with many passengers reporting delayed or wrong pickups
- Many comments suggest a lack of trust. There is a perception that some drivers use tactics to increase the cost of a ride
This information helps us decide what content to cover and prioritize in the training program. Although navigational skills and attributes to do with a driver’s knowledge in general were very important, attributes related to drivers’ behavior were equally important, and in fact more likely to cause a bad rating. This makes sense, as people may not demand taxi-driver level navigational expertise from a Lyft driver, but rude or disturbing behavior is generally never excused. Both attributes would have to be addressed by the training system.
A series of driver proto-personas helps us identify additional constraints.
The Lyft driver fleet is incredibly diverse, so the program would need to accommodate different levels of tech-savvy, learning styles, and availability of time to spend on learning. The material would have to be succinct and engaging for a large variety of people — employing a variety of media types (written, image, video, click-through…) to serve people with different learning styles. Additionally, a good program would function preventatively, addressing gaps in knowledge and roadside manner for newbies in the fleet before they ever get flagged.
But what format would make the most sense? At which juncture(s) would drivers be looped into the program? Would the lessons be digital or in-person, or both? How would the content be sourced, structured, and in what style would it be presented?
2. How might we address the issue
I put together a discovery exercise for Julia and I so that we could tackle these questions over some brainstorming. The exercise would lead us to:
- Outline the challenge, capture existing knowledge and related research
- Brainstorm ideas, strategies, features; aim to exhaust possibilities
- Synthesize the brainstorm: prioritize features & ideas; consider impact & feasibility; get a sense of resources required by different approaches to the problem
I used the virtual collaboration tool Mural to capture our session, as we were in contemporary fashion across the globe in Berlin and San Francisco.
Overall the exercise yielded the following insights:
- We aught to harness expertise of rockstar drivers and collect notes on tips & tricks from them that we might pass along to rookie drivers
- An online learning platform would be the most resource-efficient and scalable way forward.
- The platform should incorporate interactive elements, such as quizzes, videos, and discussion boards
- The narrative voice and media content should directly embody the attitude we want the drivers to adopt and project to passengers — friendly, fun, warm-hearted, the attributes that set the Lyft brand apart.
3. Let’s launch and learn
Rather than build a custom platform that would require a lot of resources to create and maintain — since efficiency and scalability was one of our main goals — I recommended we proceed with a white-label learning platform. After some research we settled on SchoolKeep, the same platform Airbnb uses to provide learning materials for its hosts.
Although poor navigation skills was one of the biggest contributing factors in the number of negative flags, addressing this part of the problem turned out to be the highest-hanging fruit. The trouble is, every city would need a custom solution — custom maps, custom tips, custom quizzes.
We decided to test our content approach first with lessons we could deploy globally, geared specifically at new drivers, and plan to roll out city navigation lessons as a stage 2 of the project.
Drawing on the insights gathered in the research phase, our philosophy for the content was to keep it text-light, and image-heavy to maximize engagement and memorability. I created more than 100 graphics total for the lessons incorporating storytelling and humor, including icons, characters, comic strips, and spot illustrations. A few samples below.
As part of stage 1 we produced 9 lessons total, addressing issues that frequently came up in passenger comments as well as general topics of interest to new drivers. Topics included “acing the first ride,” best practices for pickups and dropoffs, dealing with inappropriate passenger behavior, dealing with cancellations, getting paid, understanding how the driver is insured through Lyft, and understanding Lyft Line (the carpool option).
We made sure lessons comprised a variety of content to keep drivers engaged, including lots of images, quizzes, videos, driver tips, and clickthroughs.
Julia tapped internal teams at Lyft to produce the written copy and videos for the lessons, while I created all of the graphics and pulled it all together on the SchoolKeep platform with typography and layout.
4. How are we doing
At the current time Analytics at Lyft is working on tying lesson engagement to business results.
Julia reports:
“Once a driver starts a lesson, ~70% are completing the lesson, which is pretty high. We recently surveyed drivers who have engaged in a lesson to find out what’s working and where there’s opportunity. In terms of improvements, overwhelmingly drivers said they just wanted more — more information, more topics, etc.”
Meanwhile we are gearing up to start working on navigation lessons as well as addressing requests for more content in the current lessons.
To be continued…
