AT&C loss is the power sector's bluntest truth serum. It combines technical loss, theft, billing failure, collection failure, and bad accounting into one ugly number. India moved from loss levels above 36 percent in 2002-03 to roughly 16 percent in 2023-24, but the curve is not a victory lap. It is a thirty-year bug report. Schemes changed, meters changed, dashboards changed, and the sector still has to prove what happened at the edge.
The number nobody wants on the first slide
Aggregate Technical and Commercial loss, usually written as AT&C loss, is not just line loss. It is the part of energy input that a distribution utility cannot turn into collected revenue after technical losses, theft, metering failure, billing gaps, and collection failure are all counted.
That makes it a miserable metric and a useful one. It refuses to let the sector hide behind a single excuse. If the feeder is overloaded, the number notices. If the meter is bypassed, the number notices. If the bill is raised and never collected, the number notices. If the accounting system cannot reconcile energy input and billed energy, the number notices.
India has been trying to bring this number down since the era when distribution reform meant unbundling state electricity boards, checking feeder meters, and arguing about theft without saying too much about politics. The story is usually told as policy history: APDRP, R-APDRP, UDAY, IPDS, DDUGJY, RDSS. That framing is fine for a ministry note. It is weak for engineering.
The engineering version is simpler: the system has never had a trustworthy enough measurement layer.
Thirty years as one reform timeline
The schemes changed, but the core record problem kept returning at the edge.
2002: the baseline was brutal
The Ministry of Power said in 2011 that AT&C losses had fallen from 36.64 percent in 2002-03 to 27.15 percent in 2009-10. Read that again. The improvement was real, and the remaining loss was still more than a quarter of energy input.
This is the part of the chart that should embarrass everyone. A country can build generation capacity, transmission corridors, private renewables, payment rails, telecom scale, and rockets while still losing a large share of distributed electricity to a mixture of physics, theft, billing weakness, and collection failure.
The grid did not fail because India lacked software dashboards. It failed because the edge record was weak.
APDRP and R-APDRP: count better, lose less
The early reform playbook was not stupid. It went after feeder separation, metering, distribution strengthening, IT systems, and loss-reduction targets. APDRP and later R-APDRP understood that a utility cannot reduce a loss it cannot localize.
The pattern was familiar: establish a baseline, fund network upgrades, put meters at the right points, create an IT layer, and reward reduction. This is normal reform logic. It also exposes the structural problem. A central scheme can fund hardware and systems, but daily loss reduction happens at feeders, transformers, consumer meters, billing offices, and cash counters.
The field is always where the truth starts.
That should sound obvious. It still gets ignored. A utility headquarters can run a perfect dashboard over a measurement chain that is not trustworthy enough. A feeder reading can be late. A consumer meter can be tampered with. A DT meter can be missing. A collection entry can be wrong. The head-office number then becomes a polished aggregation of weak claims.
The old reform schemes treated measurement as an input to the program. In reality, measurement was the product.
UDAY: finance cannot rescue missing evidence
UDAY arrived in 2015 with a different center of gravity: debt restructuring, state takeover of discom debt, operational targets, loss reduction, and ACS-ARR gap closure. The Government later reported that state power distribution utilities reduced AT&C losses from 23.70 percent in FY 2015-16 to 20.73 percent in FY 2019-20 under UDAY and other measures.
That is progress, but it is not escape velocity. UDAY improved the financial story without permanently fixing the operating record. The sector learned this lesson the expensive way. You can refinance a discom. You cannot refinance a missing meter reading into existence.
Reform eras and the measurement problem they attacked
UDAY also proved the limit of headline reform numbers. A national average can improve while state-level variance remains absurd. One utility can behave like a modern distribution business. Another can behave like an accounting accident with wires attached.
This is why AT&C losses are not just a finance metric. They are an operating quality metric. They tell you whether the system can observe itself.
The state spread is the real story
The PFC 2023-24 performance report says national AT&C losses worsened from 15.11 percent in 2022-23 to 16.12 percent in 2023-24. The national average is not the most important part. The spread is.
AT&C loss range, selected FY 2023-24 utilities
A national policy discussion tends to talk as if India has one distribution problem. It does not. It has a portfolio of distribution realities. Gujarat's best utility and Nagaland's power department do not need the same intervention. They share a category, not a condition.
This matters for EV charging and distributed energy. The next wave of load will not arrive politely in the states with the cleanest loss curves. Chargers, batteries, rooftop solar, and flexible industrial loads will plug into the grid that exists. Some sites will sit in low-loss distribution environments. Some will sit in systems where the measurement layer is already contested before the charger arrives.
If the operator's software depends on the discom's record being clean, the operator inherits the discom's problem.
Why averages keep fooling the sector
The national average is useful for Parliament and almost useless for a field engineer. It compresses a thousand different operating realities into one number that sounds like a strategy.
Take a utility with high agricultural consumption, long rural feeders, political pressure against disconnection, and weak feeder-level metering. Then compare it with an urban private distribution utility with dense load, better collection, shorter lines, better instrumentation, and a regulator that can enforce tariff discipline. Calling both "the Indian discom problem" is like calling a burnt regulator and a missing invoice the same failure because both show up in the monthly loss report.
This matters because averages encourage average fixes. Average fixes create scheme documents. Scheme documents create procurement. Procurement creates meters, transformers, cables, software, dashboards, and compliance reports. Some of those investments help. Some become artifacts in a file. The field loss remains local.
The better diagnostic starts one level lower. Which feeder is leaking? Which transformer has input energy that cannot be reconciled to consumer bills? Which consumers have suspicious load shapes? Which meters stopped reporting after a firmware update? Which collection areas show billed energy but weak cash recovery? Which exceptions repeat for months because nobody owns the closure loop?
That diagnostic needs instrumentation, but it also needs receipts. A smart meter event that disappears inside a head-end is not enough. A feeder-loss report that cannot be traced back to the underlying reads is not enough. A tamper flag that never becomes a field action is not enough. The system needs a chain from observation to decision to action to closure.
AT&C loss is a lagging number. Evidence at the meter is a leading surface. India has spent too much time admiring the lagging number after the year is over.
RDSS: the smart meter decade
RDSS is the most serious attempt to pull measurement closer to the edge. The scheme targets pan-India AT&C losses of 12 to 15 percent and elimination of the ACS-ARR gap by 2024-25. Public replies cite approved works of more than Rs. 1.21 lakh crore for loss reduction and more than Rs. 1.30 lakh crore for smart metering.
The smart meter target is huge: 250 million prepaid smart meters, plus feeder and distribution-transformer metering. If it works, the sector gets monthly energy accounting, faster tamper detection, better consumer billing, remote operations, and a cleaner base for tariffs.
If it does not work, the country gets a very large electronics deployment with the same old dispute layer underneath.
This is where the LinkedIn-chart-of-the-week genre becomes irritating. A slide can show 250 million smart meters and call it digital transformation. The hard question is whether the readings become evidence that a counterparty can trust. Smart meters are instruments. They are not automatically audit trails.
There is another quiet problem: rollout success and operating success are not the same thing. A meter can be installed, commissioned, connected, and still not change loss behavior if the downstream process is weak. Someone has to use the reads to localize loss. Someone has to investigate exceptions. Someone has to handle consumer complaints without turning prepaid metering into a trust crisis. Someone has to reconcile feeder, DT, and consumer data often enough that the utility is not waiting for a quarterly surprise.
This is where power-sector reform starts looking like software operations. The device fleet is only the visible layer. Under it sits identity, firmware, connectivity, retries, time sync, security, data contracts, exception queues, field service, and audit. If any of those layers is vague, the meter rollout produces numbers but not confidence.
The sector also has to separate two kinds of belief. A discom may believe its meter because it owns the system. A consumer, fleet operator, lender, or regulator may not. For a marketplace, internal belief is not enough. The record has to travel. It has to survive outside the system that produced it.
The distribution sector is really the weak part.
Ahluwalia was describing the power distribution segment, and that is where the reform story keeps returning. India kept refinancing, reorganizing, and relaunching around the same weak point. The deeper problem was operational: the sector still struggled to prove energy input, billed energy, and collected revenue at the edge.
The answer is that loss reduction is not a single intervention. It is a habit across the measurement chain. The meter has to read. The network has to account. The bill has to issue. The consumer has to pay. The exception has to be investigated. The tamper event has to be visible. The field crew has to act. The regulator has to believe the number.
Any weak link leaks into AT&C loss.
"The next UPI is energy" needs a meter-grade record
The next UPI is energy.
Nilekani's phrase is useful because it raises the bar. UPI did not scale because a payment app looked good. It scaled because the system had a shared transaction model, identity rails, bank participation, dispute rules, settlement, and enough trust for a tiny payment to move instantly across parties.
Energy is harder. A kWh is physical. It crosses wires. It is measured by devices installed in hostile and messy environments. It is priced by tariffs, subsidies, time windows, sanctioned loads, demand charges, and sometimes political promises nobody wants to put in a schema.
An energy marketplace that mirrors UPI needs a record that behaves like settlement infrastructure. A rooftop producer, an EV depot, a battery operator, a discom, and a regulator cannot all be reading screenshots from separate dashboards. They need signed, replayable, time-aligned records at the edge.
That is the missing bridge between smart metering and market design. Without signed evidence, the energy marketplace becomes another reconciliation queue.
Why AT&C loss is a software problem now
For most of its life, AT&C loss was treated as an electrical, theft, billing, and governance problem. It still is. But the next reduction path is also a software problem.
Not software in the "new dashboard" sense. Please, no more dashboards that turn red when the thing everyone knows is broken continues to be broken. Software in the control-surface sense: protocol adapters, identity, signed events, meter provenance, tamper records, policy versions, offline sync, replay, and independent verification.
The discom's measurement chain used to be mostly internal. EV charging changes that. Rooftop solar changes that. Batteries change that. Open access and peer-style energy transactions change that. AI dispatch will make it weirder. A cloud optimizer can issue actions that affect money, grid load, and compliance. If the action does not leave a receipt, it becomes another unbilled loss pathway waiting for a name.
This is why JouleBridge starts at the site. The signed runtime does not replace the smart meter rollout. It assumes smart meters matter and asks for a stronger record around them: reads, commands, policy decisions, rejected actions, chain heads, and exportable evidence packs.
The cleanest example is an EV depot. The discom meter says the site imported one number. The charger platform says sessions consumed another. The fleet operator says vehicles were present for a third pattern. The tariff window says peak pricing began at a specific time. If those records disagree, the dispute is not solved by asking which dashboard has better colors.
It is solved by replaying the records. Which meter event was signed? Which charger command was accepted? Which policy bundle governed the session? Which timestamp source was used? Did the chain continue without gaps? Did any command get rejected? Did the evidence pack cover the full billing period?
This is why the AT&C loss story belongs inside a signed-edge-runtime corpus. The old grid could pretend the record was internal. The new grid cannot. Every new controllable asset increases the number of parties that need the record to be checkable.
What the old reforms got right
The old reforms got one major thing right: loss reduction starts with measurement.
They were also right to push feeder metering, DT metering, consumer metering, IT systems, and financial discipline. They were right that a utility cannot stay solvent while losing a quarter of its input energy. They were right that theft is not the only problem and that billing and collection failures deserve equal attention.
The missing piece was not intent. It was verifiability.
A field record that cannot survive a dispute is not enough for the next grid. A prepaid smart meter that reports a reading into an opaque head-end is better than an unread analog meter, but it is still not the same thing as a signed event stream. A dashboard that shows feeder loss is better than no visibility, but it is still not proof.
The next layer should let an operator ask: which meter read was used, who signed it, what policy applied, what previous event did it chain to, and can an external verifier replay the pack?
That is not overengineering. That is what settlement systems look like when the market grows up.
The finding
If thirty years of AT&C loss reform were written as one engineering finding, it would be this: the Indian distribution system has never had a measurement layer that can prove energy input, billed energy, and collected revenue at the edge. Schemes changed the inputs. The deeper bug stayed in place. The next decade will decide whether the new smart-meter layer becomes real evidence or another expensive electronics deployment with the same disputed paper trail underneath it.
Sources
- PFC Report on Performance of Power Utilities, 2023-24
- PIB, AT&C losses reduced from 2002-03 to 2009-10
- PIB, Implementation of UDAY
- PIB, progress under RDSS
- Power Ministry Rajya Sabha reply on RDSS
- PRS overview of power sector issues
- CSEP, DisCom Billing Losses
- Times of India, Nandan Nilekani on energy as the next UPI
- Business Standard, R. K. Singh on bringing losses below 15 percent
- Business Standard, Montek Singh Ahluwalia on power distribution losses