India’s insolvency record is public — and messy. Registers contradict themselves, order PDFs bury fee amounts next to claim amounts, and aggregators misattribute cases between similarly-named companies. This page is how we turn that into data you can act on: where each number comes from, how conflicts get settled, and what happens when we are wrong.
Counts as of 19 Oct 2026 · 8,910 companies in corporate insolvency (CIRP) and 2,721 solvent voluntary liquidations, which are tracked on their own rail
Every field we publish is traceable to a source, and every source sits in a trust tier. A fact is publishable only once a Tier-A or Tier-B document has been opened and read. A listing, a tag or a search result is a claim, never proof.
Tribunal order PDFs, IBBI per-case process pages (public announcement, claims lists, Form-G rounds, appointed professionals), MCA master data, regulator orders on the regulator’s own domain. When Tier A speaks, the question is closed.
Rating-agency rationales, auction notices on lender domains, tribunal cause lists. These settle facts inside their domain and no further — a reserve price is a price marker, never a valuation.
Third-party registry mirrors are used only for narrow facts such as directorships and charges, always with the aggregator named. They never establish that a proceeding exists — we have caught rupee-identical figures attributed to two “different” companies that were one case, misfiled.
Financial press points us at documents; it is never itself the source of a figure. A tribunal PDF hosted by a news site is the document itself — we cite the PDF and note the host.
Primary sources contradict each other more often than outsiders expect. A later document is not automatically a replacement — it may be deciding a different matter entirely. Before one document is allowed to overwrite another we check that they are about the same company, the same insolvency round, the same table and the same units.
A document’s host and the authority that issued it are also different things, and only the second one counts.
Public insolvency data does not just have gaps; it has active contradictions. Two professionals listed against one case. A register saying one thing and the case file another. The resolution rules are fixed in advance and applied the same way every time.
IBBI’s per-case process pages record who was actually appointed, with commencement dates. Per-professional register pages are partly self-reported and can double-list. When they conflict the case page wins — and the contradiction itself is logged as a finding.
Every case-to-professional link in our spine is checked against a separately scraped layer built only from per-case appointment records. A link that no source supports does not stay in the data — it goes to review.
India has many professionals sharing a name. We match on the IBBI registration serial. Companies are matched on the CIN registration number, which survives renames, conversions and state changes.
Rupee amounts inside tribunal order PDFs are usually fees, costs or bid figures — not claim amounts. We never fill admitted claims or liquidation values from an order PDF. Claim figures come only from claims-list filings.
In 2016–17 IBBI issued a cohort of provisional registrations that all expired by mid-2017. Their register pages, however, stayed online — and over the years those dead pages silently accumulated the case mandates of later, unrelated professionals.
The result: a professional’s public register page could show dozens of cases that were never theirs. Anyone scraping the register at face value inherits that pollution and reports inflated, misattributed track records without knowing it.
We found the class by cross-examining our own data. Duplicate-appointment audits surfaced impossible patterns, per-case records disproved them, and we purged roughly 4,900 phantom case links across 528 expired registrations. Those records are now flagged, excluded from every directory and ranking, and where the professional re-registered their page points at the current registration. A standing weekly guard stops the pollution re-entering on any future refresh.
A headline plan figure can cover more than payments to creditors. Reading it as a recovery is the single most common mistake made with Indian insolvency numbers.
The approved creditor allocation, including committed deferred payments. This is the only number that belongs in a recovery.
Working capital or business investment established separately in the plan. Real money, but not creditor money.
CIRP costs and allowances, shown apart from both of the above.
Use creditor payments and admitted claims for the same creditor population. Approved-plan recovery includes upfront and committed deferred amounts; cash actually received is a separate measure and we label it as one.
₹100 Cr admitted claims · ₹10 Cr upfront + ₹25 Cr committed later
₹35 Cr ÷ ₹100 Cr = 35% plan recovery
100% − 35% = 65% haircut
If the same plan also provides ₹8 Cr to the business and ₹2 Cr for costs, neither amount enters that numerator. Conditional, unquantified or non-cash treatment gets its own explanation rather than a number.
Order PDFs can establish claims and valuations where the specific table has been verified — but fees, costs and bids elsewhere in the same PDF must never be mistaken for them. A missing value stays unknown, including an unverified allocation to the company.
Every figure on this platform arrives with its date, its scope and something to measure it against. That pairing is the product; a number alone is already published by IBBI.
“Typical” means median, never mean. A sector mean is dragged by one large case; the median is the case in the middle, which is the one a reader is actually comparing against. Each benchmark states its cohort and how many cases it was computed over.
Recovery against liquidation value and recovery against admitted claims measure different relationships and are never subtracted from one another. Nor does either, by itself, measure a professional’s skill: asset quality, creditor mix, case history and process constraints all sit inside the outcome.
Field coverage varies. A calculated ratio identifies its inputs, and an unavailable or mismatched denominator does not get to produce a percentage.
Clean-once data rots. Ours is defended by recurring checks with defined pass conditions, and the dates are published at /freshness — the same page we use internally to hold ourselves to them.
We store what the source printed and format at display. If IBBI prints a contradiction our database faithfully contains that contradiction — labelled, not silently fixed.
A blank is stored as unknown and shown as a dash — never ₹0, never “No”. A professional whose accreditation is unpublished shows “not published”, not “lapsed”.
We do not sum, infer or extrapolate figures the primary record states directly. Where it states nothing, we say nothing.
Concluded cases are never overwritten by a stale feed, existing records update in place, and the expired-registration guard runs at the end of every cycle.
Duplicate appointment slots, multi-claimed roles, spine links unsupported by any source, unflagged expired registrations. The pass condition for the critical checks is zero rows, and the counts are logged.
Financial-creditor basis and total basis, CIRP-stage and liquidation-stage claims, reserve and realisation. The gap between two bases is a definition, not an error, and the page says which one it is showing.
With records on 11,630 companies compiled from imperfect public sources, errors will exist. What separates a data platform from a data dump is what happens next.
Case pages and professional profiles both carry a “Report an error” form. Anything else reaches us through contact. One paragraph, and your email only if you want a reply.
Reports land in a queue and are read by a person. Nothing is auto-applied, and nothing is auto-discarded.
The correction is checked against the Tier-A record. The document decides — not the report, and not our existing data.
Verified corrections are applied to the underlying record, so every page, ranking and benchmark that touches it moves together.
Professionals can go further: claiming your profile is free, lets you flag mandate-level issues directly, and puts your corrections at the front of the queue.
The first entries in every directory and the cases we feature are open to everyone. Read them against the primary record — that is what we do all day.