Zoomcar is India's largest personal mobility platform. The introduction of car-sharing services in 2013 and today is the market leader in the self-drive space with over 10,000 cars in its fleet. Zoomcar allows users to rent cars by the hour, day, week, or month.
Overview
The core part of being self-driving services is the freedom it offers to customers. It provides the freedom to explore and discover things independently. This freedom does come at a cost as some part of our car fleet is driven carelessly, taking a big hit on our car's health. It also hindered the customer experience of the next customer of the car.
Bad driving had a significant impact on our 3 of the core business metric.
Our objective here was to give enough information to customers on their driving behavior. This way, people can understand for themselves what they have to improve upon.
Defining Bad Driving Behaviour
How do we define what bad driving is? Is it speeding, is it fast turns?
We backtracked from our business goal of reducing Service & Maintenance Cost (SAM). What all events in customer trips impacted the car's health in the long run? After going through our past trip data, our data analytics team finalized three parameters, overspeeding, harsh acceleration & hard breaking.
Understanding Existing Driving Behaviour
Once we got metrics that contributed to the depreciation of a car's life, we monitored these metrics for a couple of weeks to get a baseline of current behavior.
Stick or Carrot
Once we were able to cast the villans for a car's health. The next step was to figure out how do we use this information to improve driving behavior?
We could punish those customers who drive badly and eventually ban them from the Zoomcar platform. Otherway is to create awareness about their driving behavior and give incentives to improve their score.
We decided to go with a mix of both approaches. We want to reward people who have excellent driving behavior while deterring those who don't. So we're creating incentives for people to drive more responsibly.
Design Sprint for Idea Generation
Here, a scale is only possible through a product solution sitting at the design/tech/analytics/business overlap. We used the diverge & converge process of the design sprint to get ideas from these company functions. It also helped us with aligning the teams at a later stage of the design process.
I defined problem statements to different stakeholders. They diverged and came with different ideas at a wireframe level. We converged on all these ideas and voted few ideas we wanted
We were able to get a couple of good sets of ideas from the sprint. Broadly the ideas presented were surrounded by three themes.
- Scoring System: The idea here is to present a score to the customer when they start a trip. We will inform the customer of what all thing he/she do to improve the score.
- Reward System: We can tie the score we have to our reward program. This can act as motivation for the customer to improve their score during the trip.
- Gamification. Use gamification strategy to maintain score over multiple trips. These include ideas like leaderboards, streaks, and community aspects of it.
Scoring & Alerting - Phase 1
Here we wanted to validate any behavior shift when users know they are tracked on the driving behavior.
As part of the initial release, we restricted the customer touchpoint to two main areas before the trip starts and during the trip. Since this project had many unknown unknowns, we wanted to build something very lean, then observe and course-correct if necessary.
Our initial wireframe discussion mostly consisted of understanding the best way to represent the score during the trip.
Education
We wanted users to set expectations on how the alerts work and make it clear what all the stuff we are going to monitor.
In Trip Scoring & Alerts
Scoring system to be a success, we need to convey the score's specifics during the trip accurately.
- Context-Based Notification: There is a reasonable probability that the customer will be driving during the notification. We need to make sure we sent only that notification that needed his attention. We divide each event into three classes based on their severity. Only when customers do a high severity event like hard braking, we notify the customer.
- Notification Sound: We wanted notification to stand apart from regular default device sounds. This way, in the span of a couple of notifications, users can differentiate from regular notifications.
Impact
Understanding the business impact with the scoring system would take us some time since it would take at least a couple of months to see the effect getting refected upon maintenance cost.
We monitored the driver score change across the different cohorts. These are the result we came up with.
Observations & Feedbacks
After three weeks of data gathering, we divided people into different cohort people who's score improved, was same, etc.
- Discovery: In the calling, we figured a couple of people was not aware driver score feature since they have skipped through the screen.
- ** Motivation:** It became clear for us to impact people who had less than average scores need more motivation to drive better.
Reward - Phase II
The main objective was to introduce the reward system to give users more motivation to drive good and fix the issues we discovered in phase 1.
Solving Discovery
After the first phase release, we identified many people who missed seeing the education screen in our research. We added couple more touchpoints for the driver score during the customer booking journey.
We wanted also wanted to experiment with driver score as the brand. It also enabled our marketing team to run campaigns with the driver's score brand, which helped with its awareness and education.
Reward
Zoomcar internally uses Z points for a reward, which customers redeem on their next trip. Here Z points worked as a variable reward, reinforcing the user's motivation for driving better.
Alerts
We did A/B testing with voice PN in which users can hear what type of alert without looking at the phone.
Given WhatsApp's high delivery date, we opted to communicate driver scores before the trip, mid, and after the trip through WhatsApp.
Impact
We rolled out the new design to 100% of our users in the span of 2 weeks. Fingers crossed all eyes on the dashboards. We closely analyzed how users interacted with the design.
Major Takeaways
The only way to know what a possible solution might be is to keep changing and experimenting.
Driver Score Representation: During the user's trip, their absolute score didn't vary significantly, especially in city drive. This demotivated a lot of users to try for a better score during the journey. We simulated a percentile scoring system during the release period to understand how scores vary during the trips. Currently, we are figuring out what is the best way to show percentile to users.
Trust is Key: In our interview couple of people did express that scoring doesn't represent their actual driving behavior. For this reason, they lost the trust in their score and motivation to improve also. Improving the accuracy of the algorithm has been an ongoing process. There have been many hurdles to overcome in fixing data availability, improving accuracy, turning weightage in the algorithm, etc.
I am currently in the process of figuring how to gamify the whole scoring system. This way, we can give more incentive to users who are doing good driving.