MeterMate: ending estimated electricity bills with photo-verified meter reading
An electricity distribution utility's meter readers wrote readings in books, and many bills were estimated or disputed. Complaints clogged service centres. We built MeterMate: a meter reader app with photo capture and OCR, route planning, anomaly detection for theft and faults, spot billing on the doorstep and a consumer app for bills, payments and complaints.
Estimated bills and angry queues
MeterMate's chief engineer described queues at service centres of consumers disputing bills. Many bills were estimated because a reading was missed; others had readings copied wrongly from a book. Some readings were never taken at all. Revenue leaked, and trust with consumers eroded.
“Every disputed bill costs us twice: once in revenue, once in trust.”
On the round with meter readers
We accompanied meter readers on their rounds. Routes were fixed in paper books, meters were often in dark corners or behind locked gates, and readers had hundreds of meters to cover. There was no way to verify a reading after the fact.
- Readings missed and bills estimated
- Readings copied incorrectly
- No evidence to resolve disputes
- Theft and faulty meters undetected
A photo for every reading
The meter reader app guides readers along an optimised route. At each meter, the reader photographs the display, and an OCR model reads the value; the reader confirms it. Unusual consumption compared with history triggers a check on the spot. Bills are printed on a portable printer at the door and sent by SMS.
Every reading has a photo, time and location, so disputes are resolved by looking at the evidence. Consumption patterns suggesting theft or meter faults are flagged to vigilance teams.
The team
The OCR model and billing integration were the core technical challenges.
Utility workshops and division rollout.
Meter reader and consumer apps.
Reader app, camera and printing.
Billing integration and operations console.
Meter OCR and anomaly detection.
6 people in total, working as one team.
Decisions we made
Agreed with the chief engineer and commercial department.
Trust typed readings?
- Typed readings
- Photo with OCR, confirmed by the reader
Our call: Photo with OCR, confirmed by the reader. Photos made every reading verifiable and reduced errors.
Bills at the door or later?
- Bills generated centrally later
- Spot billing printed at the door
Our call: Spot billing printed at the door. Immediate bills reduced disputes and sped up collections.
Where to host?
- Public cloud
- Utility data centre
Our call: Utility data centre. Utility policy required consumer data in its own infrastructure.
Every feature, module by module
Everything that shipped for meter readers, billing, vigilance and consumers.
- 01Optimised routes
Meters in the best order.
- 02Photo capture with OCR
Readings read from the display.
- 03Consumption checks
Unusual readings verified on the spot.
- 04Spot billing
Bills printed at the door.
- 05Offline mode
Works without signal.
- 06Meter issue reporting
Damaged or inaccessible meters flagged.
- 07Billing integration
Readings flow into the billing system.
- 08Reading validation
Photos and history checked.
- 09Estimated bill tracking
Why a reading was missed.
- 10SMS bills
Bills sent to consumers.
- 11Anomaly detection
Suspicious consumption patterns.
- 12Inspection tasks
Assigned to vigilance teams.
- 13Evidence records
Photos and history per case.
- 14Consumer app
Bills with meter photo.
- 15Online payment
UPI and cards.
- 16Complaints
Raise and track complaints.
- 17Consumption history
Monthly usage charts.
- 18Role-based access
Readers, billing, vigilance and consumers.
- 19Audit trail
Every reading and change recorded.
Rollout
One division piloted MeterMate for two billing cycles. Estimated bills fell sharply and disputes were resolved with photos. The remaining divisions followed with training for meter readers at each sub-division office.
- Weeks 1–3Discovery
Rounds with meter readers.
- Weeks 4–5Design
Reader app and OCR flow.
- Weeks 6–14Build
Apps, OCR, billing and vigilance.
- Weeks 15–19Division pilot
Two billing cycles in one division.
- Weeks 20–22All divisions
Training at sub-division offices.
What we learned
Evidence ends disputes. A photo with every reading changed conversations with consumers.
Help the reader, don't just monitor them. Route planning and OCR made the job easier as well as more accurate.
- Kotlin (Android)
- Next.js
- Java / Spring Boot
- PostgreSQL
- Meter OCR model
- Consumption anomaly detection
- Bluetooth bill printers
- Utility data centre
