Arnab Mitra
A few years into audit, I stopped thinking of Ind AS as a rulebook. It behaves more like a language, the grammar stays the same, but the sentences it produces depend entirely on who's speaking. Sit with an NBFC's provisioning model on Monday, a mine's closure obligation on Wednesday, and a real estate developer's revenue recognition on Friday, and you start to see the same handful of standards stretch, bend and occasionally strain to fit realities their drafters could never have imagined side by side. Four sectors in, here's what that actually looks like on the ground.
NBFCs: The standard that lives inside a model, not a memo
My first NBFC audit reset how I thought about Ind AS 109. Expected credit loss isn't a bad-debt estimate bolted onto the balance sheet; it's a live model, and the audit conversation shifts from “did this happen” to “is this reasonable,” which is a much harder question to answer. (Worth flagging for context: NBFCs have been living this ECL model since Ind AS became mandatory for them; banks will only move onto the same framework from FY 2026-27, so for now this is very much an NBFC-specific discipline, not a banking-sector one.)
The ECL model itself is where most of that difficulty lives. Every exposure first has to be staged
- Stage 1 for accounts performing as expected,
- Stage 2 for accounts where credit risk has increased significantly since origination even if nothing has actually gone wrong yet, and
- Stage 3 for accounts that are credit-impaired.
That staging decision alone is a judgement call dressed up as a mechanical rule: a days-past-due trigger is easy to test, but a qualitative trigger a covenant breach, a sectoral stress flag, a restructuring request depends on the NBFC's own watchlist processes being complete and current. Auditing ECL means testing whether accounts that should have moved from Stage 1 to Stage 2 actually did, not just whether the ones that moved were staged correctly.
Once staging is settled, the provision itself is built from three moving parts probability of default, loss given default, and exposure at default each derived from historical loss data, then adjusted for forward-looking macroeconomic scenarios and overlaid with management judgement for risks the historical data hasn't seen yet.
Testing this means pulling apart the PD curves segment by segment, checking whether the loss history window used to calibrate them is long enough to be credible, and pressure-testing the macro overlay: is the scenario weighting genuinely forward-looking, or is it a plug that happens to produce a number management is comfortable with?
The most useful audit work here isn't recomputation it's asking why a particular overlay assumption moved from one quarter to the next, and whether that movement is explained by an actual change in outlook or by a need to keep the provision within a familiar range.
A model can be internally consistent, well-documented and still be wrong in a way that only shows up when you challenge the assumption feeding it, not the arithmetic built on top of it. That's the real discipline ECL demands: Staying close enough to the model to challenge it without becoming the modeller yourself.
Mining: Where the ground itself becomes an accounting estimate
Move to a mining audit and Ind AS 16, Ind AS 36 and Ind AS 105 take over. Capitalisation of development expenditure, stripping costs, and increasingly, as groups restructure discontinued operations, all hinge on assumptions nobody can fully verify from a desk: future ore prices, mine life, the point at which a lease genuinely stops being an operating asset and starts being a wind-down.
The engagement I remember most clearly involved reconciling a large freight cost line against a 35,000-row sales register spanning a full year of dispatch data. On paper it was a routine cut of the numbers. In practice it meant building the reconciliation from first principles dispatch quantity, freight rate, destination and watching a gap emerge that didn't close no matter how the data was sliced, eventually resolving into a reclassification running into tens of crores that had sat quietly between two expense heads for long enough to matter under fraud-risk considerations, not just presentation.
What made it a mining lesson rather than a generic revenue-testing lesson was the scale of the underlying operation: six entities, one flagship company managed by a small team, and capital-intensive assets a project under construction that was ultimately trimmed of cost that shouldn't have been capitalised, once the ineligible additions were traced project-by-project. In mining, materiality isn't an abstraction you calculate once at planning and move on from. It's something you keep re-testing, because a single line item can be large enough to swing the opinion on its own.
Real estate: Revenue recognition as a legal reading exercise
In real estate, Ind AS 115 is where the real work happens, and it rarely starts with the trial balance it starts with a contract. Whether revenue is recognised over time or at a point in time comes down to a close reading of the buyer agreement: does the developer have an enforceable right to payment for work completed, and does the asset have an alternative use? Joint development agreements complicate this further, because revenue share, land cost allocation and control transfer often depend on clauses buried deep in an executed agreement rather than anything visible in the general ledger.
One project I worked on was built on land brought in under exactly such an arrangement a landowner contributing the land, a developer contributing construction, and a contractually fixed revenue share sitting at the heart of every unit sold. The developer's own MIS reported the split correctly at a headline level. What didn't hold up was a deduction buried inside that split: brokerage costs were being netted off before the revenue share was computed, at a rate several times higher than the cap the underlying agreement actually permitted. Nothing about the MIS looked wrong on a first read the formula was internally consistent, it just wasn't consistent with the contract it was supposed to be implementing. Catching it meant going back to the original registered agreement, rebuilding the revenue-share mechanism clause by clause, and finding that the landowner's consideration had been understated by a figure running into several crores as a result. It's a good reminder that in real estate, the general ledger is downstream of the lawyers, and an audit that stops at the ledger will miss exactly the kind of finding that matters most.
Manufacturing: Where the "simple" standards stop being simple
Manufacturing looks deceptively straightforward next to the above until inventory valuation, cost absorption and asset classification all start moving at once, particularly during a divestiture. A held-for-sale classification under Ind AS 105 sounds like a single reclassification entry. In practice it forces a question through every line of the balance sheet: if depreciation has been suspended because an asset is held for sale, does the fixed asset register agree with that decision, or is the register still quietly running its own depreciation calculation in the background as though nothing had changed?
That exact question came up on a plant undergoing a sale process. The asset register, run independently of the statutory books, kept computing depreciation as if the classification had never happened a gap running into tens of crores between what the register implied and what the books actually recognised, on assets that were supposed to have stopped depreciating altogether. Alongside it sat a subtler finding: a finished-goods net realisable value provision that looked properly booked at first pass, until ageing analysis showed most of the affected stock was less than thirty days old, which is an unusual profile for a write-down that's normally driven by stock that's sat too long. Neither finding was about a standard being misapplied in some dramatic way. Both were about the numbers behind the standard quietly drifting out of sync with the judgement the standard demanded which, across every sector, has turned out to be the real pattern worth watching for.
The thread that ties it together
None of this is really about the standards changing. It's about where judgement is forced to live in each sector in a system field in NBFC lending, in a project-wise cost sheet in mining, in a registered agreement in real estate, in an asset register in manufacturing. The text of Ind AS 109 or Ind AS 115 doesn't change from one engagement to the next. What changes is where the number that matters is actually sitting, and how far back you have to trace it before the standard's real question not its textbook question reveals itself. That, more than any single technical update, has been the real masterclass: learning to ask the sector-specific question before reaching for the sector-neutral standard.
Curious which sector has surprised you most with how differently it reads the same standard I'd love to hear it in the comments.
CAclubindia
