Early failure warnings for neighbourhood power transformers
TL;DR Coilcare predicts which neighbourhood transformers are likely to fail, using the smart meters DISCOMs are already fitting under RDSS. On the riskiest units it adds cheap moisture and oil sensors. DISCOMs pay per transformer watched, and part of the fee is tied to a lower failure rate.
When a neighbourhood transformer burns out, the whole street sits in the dark until a replacement turns up. India loses about 1.3 million distribution transformers a year, a failure rate of roughly 10%, according to the Central Electricity Authority's chair in November 2025[1]. The spread tells you it is fixable. Kerala reported 1.9%, and the best private distributors in Delhi and Kolkata stay below 0.5%[1].
Moisture getting into the transformer is named as the cause of 90% of failures[2], followed by overloading and poor earthing. And most state DISCOMs only find out about a failure when customers start calling.
What it does, in regulatory terms: asset-health analytics and alerts for the DISCOM. Coilcare does not operate the network or switch equipment. Sensors are fitted by the DISCOM's staff or its authorised contractors under its safety rules.
The seed pitched predictive sensing for outages in general. I kept that aim and narrowed it down to the transformer, because it fails often and the data to watch it is arriving anyway.
Coilcare reads the load, voltage and imbalance data from transformer smart meters and compares each unit with its own history and with similar units. A transformer that runs hot every evening with unbalanced phases gets flagged early. For the top few percent of risky units, Coilcare adds a small box that measures moisture, oil level and temperature, since moisture drives most failures[2]. Field engineers get a weekly list - rebalance these loads, check these breathers, replace this oil before the monsoon.
The DISCOM pays per transformer under watch, from its operations or RDSS-linked budgets. A share of the fee would depend on the failure rate falling against an agreed baseline, and that structure is my assumption, not something I've seen in a contract. The cost of doing nothing is plain to see - a burnt transformer to replace, lost sales and angry customers.
RDSS requires smart meters on distribution transformers. About 5 million had been sanctioned by March 2025, but only around 0.44 million were actually communicating[3]. As the rest come online, DISCOMs will be sitting on data they don't have the staff to analyse. In November 2025 the CEA got utilities and manufacturers together to cut failure rates[1], so the problem has official attention.
There is no public figure for spending on transformer health analytics in India. Here are the firm numbers - about 14.8 million transformers in state DISCOM networks, 5 million sanctioned for smart meters with a median award of ₹28,954 per meter[3], and 1.3 million failures a year[1]. Those size the problem, not your revenue.
One AI vendor's monitoring product watched 1,000 transformers at an Indian utility and claimed at least 50% fewer unexpected outages, which is a vendor figure[4]. A large private Delhi distributor uses AI analytics on its own data to flag transformer overheating[5]. Smart-meter suppliers bundle basic dashboards. The gap is state DISCOMs with failure rates far above the best, which don't have in-house analytics and want guaranteed results instead of yet another dashboard.
| Difficulty | Hard (57/100) |
| Starting capital | ₹1.5–3 Cr |
| Time to launch | 12–18 months |
Prediction is the easier half. The hard parts are getting meter data out of a DISCOM's systems and getting field crews to act on a list. State DISCOMs buy through tenders, pay slowly and change officers often, so the one champion you have can vanish halfway through a pilot.
No licence covers analytics itself, but working on live equipment and buying through tenders bring rules.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't let AI switch equipment or decide when a crew works on a live unit. Those calls stay with the DISCOM's engineers.
Want to work out the AI layer for Coilcare? Talk to Yash →
This is right for a founder who has worked inside a DISCOM or with a smart-meter vendor, and who can sell outcomes to a managing director. If you're expecting quick software sales then look elsewhere. First prove that one DISCOM circle will share 12 months of transformer meter data and that your top-ranked units fail at several times the average rate.
Plan for roughly ₹1.5–3 Cr, because DISCOM sales are slow and sensors, data work and pilots need funding.
No sector licence for analytics, but you need WPC approval for radio boxes, DISCOM-authorised crews for installs and the eligibility tenders ask for.
Yes, transformer failures cost utilities everywhere. Check how your local utilities buy, who owns meter data and the safety rules for live equipment.
The free Coilcare validation kit has the line-by-line capital breakdown, the effort scoring, the procurement, safety and radio checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for DISCOM engineers, and 5 assumptions to test. Download the kit →
Thinking about building Coilcare or something close to it? The make-or-break bit is getting a DISCOM to share its meter data and then act on your list, and I can help you work out how to ask for both. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.