AI and machine learning
Product design and data science worked seamlessly to meet user needs and hit KPIs. If a driver gets a price within the first minute they engage more, which increase revenue and increases KPIs.
My Role: Lead Product Designer
Team: Data scientist, product manager, copy writer, marketing manager
- Quote rate increased by +145%
- Acceptance rate increased by +36%
Acheivements
Give drivers an instant accurate price to allow them to get booked in for their car repair.
Problem statement
Increase quote rates by +50%, increase acceptance rate by +20%
KPIs to gauge success
Empathise with the users
Site visits
I interviewed around 20 garages and mechanics. The main concerns for the garages are around price and making sure these are accurate.
Interviews and surveys
Do drivers expect the price to be final? How would they feel about us creating a price for them as opposed to a bespoke quote coming from the garage.
MVP for feedback
We released an MVP with 20 volunteer garages that we could then gather feedback on:
- How accurate was the price
- What factors affect prices
Concerned that the price we generate is wrong, worried about drivers accepting a wrong price and bad customer relationships leading from this.
Key comments from garages
“I dont expect the price to be final, I expect this to go up or down when they see the car”
Key comments from drivers
Define
I gathered all the results and held a session consisting of affinity diagram sessions to consolidate the results and define the start point.
- Drivers expect prices to move, however we need to ensure they know its a quote
- Garages want accurate prices:
- The AI has a tolerance of plus and minus 10% so we need to give the the power to manage and update prices.
- The parts plus labour calculations do not take into account the "Fudge factor". For example a certain model ford focus may not need a part removed to get it the timing belt.
- Garages need to update prices instantly and live.
- Garages need a management tool to view there prices and to update if they feel they need to. (They will have upwards of 300,000 potential quotes)
Some key patterns emerged
Map out full user journey
Using our customer journey map, add in the instant pricing touch points for the garage and the drvier.
Affinity diagrams
Group all the feedback and findings to notice larger patterns and fuel conversation and planning within the team.
Design and Ideate
I then set out designing the flows, wireframes, interaction designs and prototypes, so we could test and iterate. Keeping our key user stories in mind.
Map out user flow
I start with a simple user flow to start to map out touch points, interactions and get a feeling for the flows.
Wireframes flow and wirefames
After I have the simple flow I build on this and more detailed wireframe flows, and prototypes with wireframes.
Brain storm and feedback
With all the user feedback, flows, wireframes and UI I arranged a brainstorming session to discuss findings and gather feedback from the team.
The garage will want to view and manage all the quotes we provide, and we need to differentiate the quote types for drivers.
Key findings
Driver UI / interactions
Now that I have tested the wireframes and the brand, I then created the UI designs and created the prototype to then test and pass onto the devs.
All touch points need to be considered. UI design, emails designs, text messages, T&Cs and inperson correspondances with garage.
All touch points and end to end experience mapped
UI designs in sketch
- Sketch
- Invision prototype
Interaction design problem
Garages want to be able to update a price in real time, we then need to show the change to the driver.
How will we update a price in real time on the comparison page whilst keeping the user in full control at all times.
Interation problem statment
Garage management
"As a garage I want to be able to view all my prices and update and manage them easily"
User story
Management tool
The garages need to search and manage over 300,000 prices.
Interaction mapping
Working out the details of all interactions in the feature to provide as a deliverable to the devs.
"What if a driver accepts a quote that's wrong? We need to be able to update and monitor prices in live time"
Garage concern
"The AI model needs to be kept up to date to learn"
Data science feedback
Solution
I designed a tool to allow garages to update prices live update a drivers price and save for future use. This also updates the AI model everytime a user inputs a updated price.
On-boarding for new garages
New garages that do not have any quoting behaviour to use to create a matrix for would initally not be able to have smartquotes. So to get around this we would need to get them to quote on 24 pre-defined car types, and job types to then use this date to cluster the results and allow us to generate 300,000 prices. Prototype 1 Prototype 1
Usability tests with mechanics and drivers before release
Once we had a working version to test the end to end flow, I approached several garages that had indicated that they would like to be involved in the process of developing a solution.
Releases
A session with the product team was then held, to work out how we want to release, how we are going to monitor and how we are going to gather data on the release.
Decisions
- Release on two job types first
- On-boarding to allow mechanics to authorise us to start quoting on their behalf
- Set up hot-jar to monitor users live
- Set up GA dashboard
- Data team daily feedback reports
- Release on a category by category basis and slowly introduce more features and improvements based of further feedback.
Instant quote rate increased by +145%, and accept rate increased by +36%;
Results
Next steps
- Working with a local Usability Lab and research team to test the full journey from the driver side.
- Base changes on this feedback.
- Release next interaction of the management tool.