MileIQ recommendations are based on previous classifications, the system recommends classification for all the recent drives. MileIQ provides an intelligent way of classifying drives in bulk with this feature.
Scroll to see details
MY ROLE
Product Design Lead
Led and executed end-to-end product design from concept to launch.
Defined long-term and short-term design strategies.
Led design reviews and kept the stakeholders involved and on board with all design decisions throughout the project.
Practices: Design reviews, design sessions, user research
Deliverables: Problem framing & solution, User-flow, interaction design, wireframes, prototype, visual design.
Team: 1 Product Manager, 1 UX writer, 1 Data scientist, and 4 engineers.
Duration: Six weeks
BUSINESS GOAL
Improving low classification rate
Product impact: The drive classification is the core function of the MileIQ app. If users don’t classify enough drives means they are not taking full advantage of the app. And this would impact the MileIQ revenue in the long run.
Business impact: Users could save up to $6,200 a year.
MileIQ Subscription: $59.99/year
RESEARCH TAKEAWAYS
We looked at the data from user interviews, surveys, feedback, and CX tickets. And recognized the highest volume recorded around classification issues.
• 71% Manual classification: There are too many drives to classify manually.
• 29% Accuracy concerns: People don’t remember the context of the drive if they don’t classify it right away.
Persona: Construction Worker
Daily drives: 8-10: Business 79% vs Personal 16%
Archetype: Construction workers, Realtors, Freelancers
Classification Freq: Weekly, Monthly, Yearly
PROBLEM STATEMENT
The classification of drives is a manual process and it is causing a low classification rate. We need to make the classification experience less manual, more efficient and accurate.
BRAINSTORM FOR SOLUTION
Manual to Automation to ML
We already had a few features in the app to automate classification by creating rules around repeated drives. However, these automation features are still not enough to improve the classification rate. So we naturally thought of building an ML model and making the MileIQ intelligent.
Current state (Manual & Automation)
Future state (ML)
OUR SOLUTION
Bulk classification based on ML recommendations.
Design
DESIGN EXPLORATIONS
How recommendations should be displayed.
We explored a lot of ideas to visualize how the Recommendation list should be displayed on the screen. It could be a List view, buckets, card, tabs, etc. We started testing these hypotheses with users and decided that it would be a list view.
Pro
Nice balance of text
Precise enough so the user can scan through it easily
Long enough to be bulk classified.
Easy-to-read displayed information
Easy to understand the call to action
Clarity on user control over recommendations
RECOMMENDATION LIST
10 drives per session
Cons
The drives only show up in 10 drives per session. If user wants to classify 40 drives they will have to have to go through 4 sessions.
Why I am recommending this
The information display makes it easy to read, make corrections, and take action.
Pro
Nice balance of text
Precise enough so the user can scan through it easily
Long enough to be bulk classified.
Easy-to-read displayed information
Easy to understand the call to action
Clarity on user control over recommendations
ENTRY POINT
FAB (Floating Action Button)
Cons
The drives only show up in 10 drives per session. If user wants to classify 40 drives they will have to have to go through 4 sessions.
Why I recommended this
The information display makes it easy to read, make corrections, and take action.
UX writing
The tone and language play a vital role in feature adoption and gaining trust from users. We wanted to make sure users understand the value of the recommendations and how they can take full advantage of them.
Iconography
I wanted to create an icon for recommendations that is something users are familiar with. We chose a star as the icon for the recommendation feature. Historically, the star icon has been used in MileIQ for visual indication of task completion.
Illustration
I wanted users to pause here and read the message as this message would help them take full advantage of the feature and only show up once. I see a great opportunity to choose an illustration style that would help us push our visual design system and evolve further.
FINAL DESIGN
Readable, actionable & clear
CONCLUSION
Outcome & Achievement
OUTCOME
The classification rate has improved significantly. Users are more confident about the accuracy of the classification of their drives.
ACHIEVEMENT
I am very proud of our work on this project. Microsoft filed a patent on this ML model. My team and I were awarded the Microsoft Patent Award.
UX writing credits: Melissa Santos