Content authored by practitioners with experience at Amazon India, Intuit QuickBooks, and the Tata Group. Meet the team →
Indian finance teams running 2–6 person reconciliation desks against tens of thousands of monthly transactions absorb structural, repeatable revenue losses across TDS credits not appearing in Form 26AS or Form 168, ITC lapsed under Rule 36(4) or reversed under Rule 37, undisclosed MDR and platform settlement fees, unreconciled NACH bounce-charge recoveries, sub-rupee rounding compounded across millions of rows, partial payments closed as full at month-end, and a residual catch-all of unexplained variance written off as a JV. Without a named class for each pattern, every month-end closes leakage by guesswork and the recovery trail is permanently broken.
Apply a seven-class variance taxonomy at the point of reconciliation: FEE_DEDUCTION for platform and bank fee variance against contracted rate, TAX_DEDUCTION for TDS receivable not in Form 26AS or 168, DISCOUNT_APPLIED for discount applied without authorisation tag, ROUNDING for sub-₹10 paise variance, PARTIAL_PAYMENT for invoice short-settled against its outstanding, PENALTY_OR_INTEREST for Section 50 interest, Section 416 interest, NACH bounce charge or late-fee debit, UNEXPLAINED for residual variance with no rule match. Each class carries a regulator anchor, a detection signal, and a recovery action so the Discovered Money register can age recovery by class.
Variance taxonomy seven-class enum applied at reconciliation engine output, regulator-anchor table mapping each class to the operative section (Section 393/394/413/416, Rule 36(4)/37, NACH circular, contract clause), Discovered Money register keyed by class and ageing bucket, week-one detection-signal checklist for CFO baseline, recovery-action library by class with owner and standard SLA, audit-trail field on every reclassified line capturing the original class and the corrected class.
A monthly Discovered Money register that decomposes total reconciliation variance into the seven classes with regulator anchor and recovery owner per row, a Recovery Aged Trial Balance by class showing stuck / at-risk / recoverable rupees, a quarterly leakage-class trend report for the audit committee, and a permanent reduction in the unexplained-variance write-off line once classified variance moves to the recovery queue instead of the JV.
A CFO at a ₹85 crore Bangalore IT-services firm closes the FY26 books and finds, on a year-on-year comparison, that ₹37 lakh has moved across three lines of the trial balance without an audit-defensible story. ₹14.2 lakh sits as “TDS receivable — older than 24 months.” ₹9.4 lakh has been written off as “Razorpay settlement variance.” ₹6.1 lakh appears as “GSTR-3B Table 4(B)(2) reversal — Rule 36(4).” ₹4.8 lakh is sitting under “NACH bounce charges — recoverable.” And ₹2.5 lakh, the smallest line, is the most dangerous: “Misc reconciliation adjustment.” No regulator owns that line. No customer is being chased on it. It is forgiven leakage.
This is what revenue leakage looks like in an Indian finance team: not a single catastrophic event, but seven repeatable classes that compound silently across every month-end close. This guide names each class, ties it to the regulator that governs it, and shows the detection signal a CFO can act on in week one — before any reconciliation platform is configured.
Quick reference: the seven classes of revenue leakage
| Class | Regulator / source | Detection signal | Typical leakage band | Primary recovery action |
|---|---|---|---|---|
| Fee & commission leakage | Contract, CBIC (GST on MDR), RBI (bank charges) | Settlement fee column does not equal contracted rate × gross | 0.05–0.4% of platform volume | Fee-level reconciliation against contract, monthly chargeback claim |
| Tax-deduction leakage | CBDT (Section 393/394, Form 26AS / Form 168) | Books TDS receivable greater than 26AS / 168 record | 8–14% of TDS-bearing receivable | TDS receivable ageing, deductor follow-up, rectification request |
| Discount leakage | Internal pricing policy, Section 15 of CGST Act | Discount applied without authorisation tag on invoice | 0.1–0.6% of gross revenue | Discount authorisation gate, double-discount detection |
| Rounding & truncation leakage | Internal, Ind AS 8 materiality | Sub-₹10 paise variance per transaction | ₹0.20–₹2 per transaction | Engine-level rounding policy, period-end rounding sweep |
| Short-settlement leakage | Contract, Section 73 Indian Contract Act | Partial payment closed as full settlement | 0.4–1.8% of receivable | Open-residual queue, payment-allocation engine |
| Penalty & interest leakage | CBDT (Section 416), CBIC (Section 50), NPCI (NACH circular) | Statutory interest paid that was avoidable | ₹50,000–₹6 lakh per ₹10 crore turnover | Calendar-driven statutory clock, automated dispatch |
| Unexplained variance | Internal, CARO 2020 Para 3 | ”Adjustment” or “write-off” lines above ₹10,000 | 0.1–0.7% of revenue | Variance taxonomy enforcement, no-JV-without-class rule |
What does “revenue leakage” mean in practice?
Revenue leakage is the rupee value a business has earned, billed, or is statutorily entitled to claim, that never lands in its operating cash account or its tax-credit register. The framing matters because the three things it is not are routinely confused for it.
It is not bad debt. Bad debt is the conscious credit decision to extend payment terms to a customer who fails. It is provisioned, reported, and audit-defensible.
It is not fraud. Fraud is malicious — an employee, vendor, or counterparty acting with intent to misappropriate. It triggers a separate investigation track, often legal.
Leakage is structural and repeated. It is the rupee lost because the platform did not disclose the fee, the deductor did not file the return, the supplier did not raise the credit note, the controller did not have time to chase the residual, or the engine had no rule for the variance and the JV team closed it to clear the books.
Seven classes cover almost every real instance, and they map cleanly onto regulator-anchored detection signals.
Class 1 — Fee & commission leakage
The biggest visible class for any business that takes more than 30% of revenue through a payment gateway, marketplace, or aggregator. The pattern is contractual opacity at the settlement layer: a 2.0% MDR rate is contracted, but the settlement file applies 2.05% on certain card types, 2.1% on UPI-Premium, or layers a “platform fee” the merchant cannot reconcile to the underlying transaction. The variance never gets surfaced because the finance team trusts the aggregated daily settlement number rather than the per-transaction fee column.
Detection signal: take any 30-day settlement file from your payment gateway. Sum the gross transaction value. Multiply by your contracted MDR. Compare the product to the actual “fees” column total on the same file. Any non-trivial delta is fee leakage. For a D2C brand running ₹4 crore of monthly Razorpay volume on a contracted 1.95% MDR, a 0.07% undisclosed delta is ₹2.8 lakh a year of fee leakage.
Recovery action: monthly fee-level reconciliation, chargeback claim filed within the platform’s dispute window (typically 60–90 days), contract review at 6 months.
Class 2 — Tax-deduction leakage
The biggest invisible class for any services-led business. Section 393 and Section 394 of the Income Tax Act 2025 require customers to deduct TDS at source on most service payments — typically 2% under Section 393(1)(a) payment code 1002 for professional services, 1% for individual contractors. The deductor is supposed to deposit the amount, file the quarterly TDS return, and the credit lands in the deductee’s Form 26AS / Form 168 record within the year.
In practice, 8–14% of TDS-bearing receivable for a typical mid-market services business never converts to a claimed credit. Causes include wrong-PAN filings (the deductor used an old PAN or an incorrect one), wrong-section code (Section 393(1)(a) code 1002 entered as 1003), wrong assessment period, deductor never filing the quarterly return, or the credit appearing in 26AS but ageing past the rectification window.
Detection signal: pull every TDS receivable on the books older than 90 days. Cross-check each against Form 26AS / Form 168 by PAN, period, and section code. Any books-side TDS receivable not appearing in 168 within the deductor’s filing window is leakage in progress.
Worked example. A ₹50 crore IT-services firm has ₹85 lakh of TDS-bearing receivables flowing through customer side annually (effective rate 1.7% on professional-services revenue mix). A 12% leakage band on this is ₹10.2 lakh per year of TDS credit silently lost.
Quantify TDS credit leakage on your receivables
Plug in revenue, deduction rate, and 26AS / 168 match band to get a class-wise leakage estimate before any reconciliation deployment.
Open the TDS Mismatch Estimator →Class 3 — Discount leakage
Indian B2B and D2C businesses run two parallel discount channels: contracted discounts (volume slabs, MOU rates, channel partner terms) and tactical discounts (festival, clearance, retention). Without a discount-authorisation gate at the invoice line, a customer can be granted a 4% MOU discount and a 3% festival discount simultaneously on the same SKU. The invoice prints fine; the margin reconciliation in the next quarter shows the gap.
Detection signal: pull every invoice line where total discount exceeds your highest contracted discount slab. Each line is either authorised (find the approval mail) or unauthorised (discount leakage by definition). Under Section 15 of the CGST Act, only discounts established at or before the time of supply and recorded on the invoice are deductible from taxable value — the rest land as GST-payable surprises.
Class 4 — Rounding & truncation leakage
The class everyone dismisses until the volume hits scale. A payment gateway truncates paise on every settled transaction at the merchant-side credit, rounding 23.47 to 23.45 systematically. On any single transaction this is invisible. On 380,000 monthly transactions, 0.02 rupees per transaction is ₹91,200 a year — entirely lost to engine-level rounding policy that no contract disclosed.
Detection signal: take the per-transaction gross from the platform, compute the contracted-rate net to the paise, compare to the per-transaction credit in the settlement file. The systematic per-transaction delta is rounding leakage.
Class 5 — Short-settlement leakage
A ₹4.2 lakh invoice gets ₹4.05 lakh credited against it. The AR controller, under pressure to clear the open-invoice list before MIS day, marks the invoice closed. The ₹15,000 residual stops being chased. Across a hundred similar closures in a year, this is a 0.4–1.8% of receivable line that never makes it to the bank.
Detection signal: pull every invoice marked “settled” in the last 90 days. Check the cash-receipt allocation against the invoice gross. Any residual closed without a credit-note authorisation is short-settlement leakage. The Indian Contract Act under Section 73 actually preserves the recoverable right — but only if it has not been formally written off.
Class 6 — Penalty & interest leakage
Indian statute makes interest the silent compounding cost of operational delay. Section 416 of the Income Tax Act 2025 charges interest on late-deposit of TDS at 1.5% per month. Section 50 of the CGST Act runs 18% per annum on ITC claimed without GSTR-2B support or on ITC reversed under Rule 37. NPCI’s NACH circular framework lets sponsor banks debit ₹300–₹750 per returned mandate without surfacing the recovery against the end-customer.
These are not occasional events; they are calendar-driven. A finance team without a statutory-clock dashboard pays Section 416 interest on every Q3 TDS deposit that crosses the 7th, Section 50 interest on every period where the supplier filed GSTR-1 late, and NACH bounce recovery is never billed onward.
Worked example. A ₹40 crore NBFC running 18,000 monthly NACH presentations with a 6.8% return rate sees roughly 1,224 monthly bounces at an average ₹450 sponsor-bank charge — ₹5.5 lakh a month, ₹66 lakh a year, of which the standard recovery clause in the loan agreement allows about 80% to be recharged. Without a recharge-tracking workflow, the entire ₹66 lakh stays as a finance-side cost line.
Class 7 — Unexplained variance
The catch-all, and the most dangerous. Every reconciliation engine, manual or automated, produces a residual: rupees that did not match a rule, did not fit a known class, but had to be cleared so the books could close. The traditional treatment is a JV to a generic “reconciliation adjustment” account.
Every rupee written off this way is leakage that has been formally forgiven. The CARO 2020 Para 3 reporting requirement around inventory and receivable reconciliation actually picks this up at audit; the auditor’s working paper invariably asks for the variance schedule, and the JV-to-adjustment lines come under scrutiny.
The cure is a no-JV-without-class rule: every reconciliation variance must be classified to one of the other six classes (and if it cannot be classified, it stays in the queue for the next period, not in the P&L).
Putting the framework to work — a four-step CFO playbook
Step 1: Run the week-one detection-signal checklist. The five signals at the top of this guide are designed to take a CFO from “I suspect we leak” to “here is the rupee number” in 4–6 hours of data pulls.
Step 2: Build the Discovered Money register. Every variance classified into one of the seven classes, with a regulator anchor, a recovery owner, and a standard SLA. This becomes the single artefact the audit committee asks for at every quarterly review.
Step 3: Migrate from JV-to-adjustment to class-coded queue. The unexplained-variance bucket should shrink monotonically as the rule library grows. A mature reconciliation engine running on the 51% → 88% match-rate band leaves an exception queue of 12–18%, of which more than 90% should hit one of the six rule-based classes inside the standard SLA.
Step 4: Quarterly trend reporting by class. The class-wise leakage trend is the single most useful artefact the CFO can put in front of the audit committee — it tells them which class is structural (fix the engine), which is procedural (fix the team), which is contractual (fix the platform), and which is statutory (fix the calendar).
Continue reading on the leakage backbone
For the regulator-by-regulator deep dive, continue with TDS credit leakage and Form 26AS / Form 168, ITC leakage under Rule 36(4), and working capital leakage from reconciliation delays. For the executive-summary view of the framework, see the Stop Revenue Leakage pillar page.