Mission Report · MapMyIndia [Gtropy] · InTouch, for VinFast & SM GreenProject case study
Driver Safety
A fleet manager doesn't need three sensor dashboards. They need to know which driver to look at first, and why.
- Status
- Active
- Owned
- PRD owner · UX · DMS, ADAS & driver sections
- Clients
- VinFast · SM Green
- Camera alarms
- 7 in-cab · 4 road
- Case study
- Read it on Behance
- Stack
- PRD v1.0
- Clickable prototype
- OpenAPI contract
- InTouch
ResultsBeing measured
- Time to find the riskiest driver
- Scorecard adoption
- Alert settings customised per customer
- “Score looks wrong” support tickets
- Alarm-naming support tickets
I decided
- Driver-first: who needs attention, on one screen
- One Safety score; alerts grouped by risk, not by sensor
- Data only: no coaching programme, no review workflow
- Scores per distance or hours, never raw counts
I built
- PRD v1.0 (13 Oct 2026)
- The clickable prototype, 11 design iterations
- The DMS, ADAS and driver sections in InTouch
With the team
- The OpenAPI contract with backend
- Delivery with frontend and QA
How it works
What it does, feature by feature
Safety overview
One Safety score with its trend, the risk themes, and the drivers who need attention first.
All sourcesDMS
What the driver does: 7 in-cab camera alarms, from phone use and smoking to drowsiness and a missing seat belt.
In-cab cameraADAS
What is happening on the road: 4 road-facing alarms for collision, pedestrians, headway and lane departure.
Road cameraTrip alerts
What the vehicle did: overspeeding, harsh acceleration, braking and cornering, night and non-stop driving.
GPS & motionDriver scorecard
Every driver ranked on the same scale, with a per-driver report behind each score.
All sourcesOn-demand video
Request the clip behind an alert, so a score can be checked against what actually happened.
CamerasAlert settings
Each customer hides alarms and sets their weights; weights must total 100, and hidden alarms never leak into exports.
Per customer
Sources
- In-cab cameraDMS
- Road cameraADAS
- GPS & motionTrip alerts
Scoring
- Count per distance or hoursnever raw totals
- Weight by the customer's settingsweights total 100
- Drop hidden alarmsenforced server-side
The manager sees
- One Safety score
- Risk themes
- Drivers ranked, with video
How an alarm becomes a ranked driver. A model of the product's logic, not live data.
Each alert type belongs to one risk theme; the source is a tag. A theme with no alert types for a customer is hidden, so a GPS-only fleet still sees Speeding, Harsh driving and Fatigue.
Brief
A Driver Safety module inside InTouch for VinFast and SM Green: driver monitoring (DMS), driver assistance (ADAS) and trip alerts, rolled into one score and one ranked list of drivers.
Context
Video telematics produces a flood of alarms from two cameras and the vehicle's own sensors. Sold by sensor, the product had a page per source and three separate gauges, and none of them said who to talk to.
Users
Fleet managers and safety leads at enterprise customers, who have minutes, not hours, to decide which drivers need attention this week.
Discovery
The sensor split is how the product is sold and entitled, not how a manager thinks. Alerts that all look equally important stop getting opened; the useful grouping is by risk (speeding, harsh driving, distraction, fatigue, collision risk, compliance), with the source shown as a small tag.
Problem
Three modules with similar-sounding scores (driving score, fleet safety score, driver scores) and a third page called Driving Behaviour that read like the parent of the other two.
Strategy
Name each source by what it sees (the driver, the road, the vehicle), keep exactly one Safety score, and make the overview answer three questions in order: are we improving, who needs attention, and what happened.
Prioritization
Overview and the driver scorecard first, then the DMS, ADAS and trip-alert pages as the places to check the detail, then on-demand video and alert settings. Coaching and a review workflow were cut: the product reports, the customer decides.
UX
Driver-first: the riskiest drivers are on the first screen. Scores compare per distance or per hour, never raw counts, and a driver isn't scored until they have driven enough to be compared fairly.
Design
A clickable prototype through 11 design iterations by 1 October 2026, each page hidden for a customer who hasn't bought that source.
Technology
A PRD (v1.0, 13 October 2026) and an OpenAPI contract for backend, frontend and QA. The scores, bands and risk themes are computed from event counts and the customer's own alert weights, and hidden alarms are enforced on the server so they never leak through an export.
Outcome
In build. The PRD names what will be measured, listed above as targets, not results.
Learnings
A name is a product decision. Three scores that sound alike cost more support tickets than any missing feature.