सीमा-अपवंचन
Bloodhound
Threshold evasion. Finds payments and awards broken into tranches that sit just under a sanction, tender or delegation ceiling, including sequences spread across weeks or across vendors.
Audit Intelligence
Forensic financial audit intelligence for Indian government departments
सरकारी वित्तीय जाँच मंच — a reference brief on what a forensic money-trail engine can establish from records a department already holds, what it cannot establish, and how findings are presented so a reviewer can act on them.
Definition
What NIRNAE is (सरकारी वित्तीय जाँच मंच)
NIRNAE is a forensic money-trail engine built for government audit, finance and vigilance functions. It takes expenditure records a department already produces: treasury payment registers, procurement and tender awards, subsidy and beneficiary lists, grant releases, loan disbursal statements. It merges them into a single working ledger, runs nine detection agents across that ledger, and returns a ranked set of findings with the underlying rows and the arithmetic attached.
It is deliberately not an integration product. There is no connector to treasury systems, PFMS, core banking or departmental applications. The department decides what to export and what to submit, which keeps the data boundary inside the institution's own control and removes the procurement and security burden that live-access tooling carries.
Every output is advisory. A finding is an indicator that a specific set of transactions deserves verification. It is not a determination of fraud, and it does not substitute for a statutory audit, departmental enquiry or investigation.
Method
How it works (कैसे काम करता है)
Four steps, no integration, no field data collection.
01
Export
The department exports the registers it wants examined, in whatever format the source system emits.
02
Normalise
Files are merged into one working ledger: dates, amounts, payees and identifiers reconciled across sources.
03
Detect
Nine agents run across the same ledger, each testing a distinct pattern and citing the rows behind each flag.
04
Rank
Findings are consolidated per entity into one severity score, ordered so verification starts with the strongest case.
Detection
The nine detection agents (नौ जाँच एजेंट)
Each agent tests one pattern that recurs across public expenditure. They run on the same ledger so their findings corroborate each other.
विक्रेता मिलीभगत
Spider
Collusion rings. Links suppliers that share an address, telephone, bank account, tax-identifier series or directors, and shows where those linked firms bid against each other.
फर्जी बिलिंग
Phantom
Ghost billing and phantom works. Surfaces vague or non-measurable line items, bills repeated across financial years, and payees with no other trace in the book.
समय-संबंधी विसंगति
Clockwork
Timing anomalies. Year-end fund dumping, same-day sanction and payment, approvals recorded outside working hours, and unusual clustering before a closing date.
भौगोलिक असंगति
Compass
Geography mismatch. Vendors located implausibly far from the sanctioned work site, and awards falling outside the sanctioning authority's jurisdiction.
लेखा-शीर्ष दुरुपयोग
Ledger
Head-of-account misuse. Expenditure booked to a scheme or budget head that does not match the described purpose of the payment.
लाभार्थी इतिहास
Origin
Beneficiary and vendor history. Firms incorporated shortly before a large award, dormant entities reactivated, and beneficiaries recurring under near-identical particulars.
चक्रीय मार्ग
Conduit
Pass-through and circular routing. Layered intermediaries that add no discernible work, and funds that return towards their point of origin.
समेकित वरीयता-क्रम
Verdict
Consolidation and ranking. Merges every corroborating finding into one severity score per entity, so a reviewer starts with the strongest case rather than the first row.
Value
Why audit teams use it (सरकारी लेखापरीक्षकों के लिए लाभ)
Whole book, not a sample
Manual audit samples because time is finite. The engine reads every row that was exported, so a pattern spread thinly across districts still surfaces.
Arithmetic on the face of the finding
Every flag carries the rows it came from and the calculation that produced it, so a reviewer can re-derive it without asking us how it was computed.
No system integration
Nothing is connected to treasury, PFMS or banking systems. The department exports what it wants examined; IT effort is close to zero.
Ranked, not exhaustive
Output is an ordered inspection list. Officers spend verification capacity on the entities with the most corroboration, not on whatever came first alphabetically.
Patterns
What irregularity looks like in the data
These are the shapes that repeat across scheme funds, procurement, subsidy and loan books, whatever the department.
Threshold evasion
A work that should have gone to open tender is issued as several smaller orders, each just below the ceiling. In isolation each order is defensible. In sequence — same payee, same head, same fortnight — the pattern is visible arithmetically.
Vendor collusion
Competing bidders that share a bank account, a registered address, a phone number or a director. The CAG's own audit findings on state e-procurement have recorded suspect bids traced to shared machines, duplicate PANs and fabricated email identifiers — exactly the linkage a collusion agent tests for.
Ghost billing and phantom works
Bills certified without measurable quantities, the same measurement abstract raised in two financial years, or a payee that appears once for a large amount and never again.
Timing anomalies
March-end surges that exhaust an allocation in days, sanction and payment recorded on the same date, and approvals timestamped outside working hours.
Circular routing
Funds moved through intermediaries that add no discernible work before returning towards their origin, often across scheme boundaries.
Comparison
Against the alternatives (तुलना)
| Approach | Scope | Where it falls short |
|---|---|---|
| Manual audit and consultancy review | Sampled records, delivered as narrative reports | Slow, non-repeatable, and rarely leaves a re-runnable analytical trail |
| Generic fraud and AML analytics | Built for banking and enforcement data models | No notion of sanction ceilings, budget heads, scheme structures or work orders |
| Existing government portals | Query and reporting over transactions already recorded | Report the data faithfully but do not test it for cross-entity patterns |
| NIRNAE | Exported departmental records, nine detection agents, ranked findings | Advisory only — indicators for verification, with no live system access |
Deployment
Which institutions this is for (लक्ष्य एजेंसियाँ)
Any office that accounts for public money and can export the records it already holds.
- 01State finance departments and treasuries
- 02Directorates of audit and local fund audit
- 03Vigilance and anti-corruption cells
- 04Union ministries releasing scheme funds
- 05Development finance and refinance bodies
- 06Research and defence procurement bodies
- 07Public sector banks reviewing loan books
- 08Urban local bodies and district administrations
- 09Panchayati raj and rural development wings
Regions
State-level briefs
The same engine, described against each state's own audit and treasury structure.
Next
Conclusion and next step (निष्कर्ष एवं अगला चरण)
Public money leaves a trail in the records departments already keep. The constraint has never been the absence of data — it is that reading all of it, consistently, across departments and years, is beyond a manual sampling process. NIRNAE closes that gap without asking an institution to surrender live access to anything.
The product is in pre-deployment. Engagements today are supervised reviews on exported or anonymised datasets, and every finding is an indicator for verification. See how evidence is presented or read the data-handling boundary.
Disclosure
Operating boundary. NIRNAE holds no live access to treasury, PFMS, banking or departmental systems. The department exports the records it wants examined and submits them. The product is in pre-deployment: engagements today are supervised pilots on exported or anonymised datasets, and every finding is an indicator for verification, not a determination. Full answers in the FAQ.
Interested?
Leave a departmental email and we will arrange a briefing on a sample dataset.
Request a briefing