Every figure on this page is a count taken from the database on the stated date. Nothing here is a projection, a target, or a rounded marketing number. Where a figure changes, the as-at date changes with it.
Measured counts as at 30 August 2026. Seven agencies live; earliest rating action held is from May 2019.
As at 30 August 2026, the TatvaRatings corpus holds 70,337 rating actions across 45,528 borrowers, 194,471 rated facilities and 179,857 lender-exposure rows naming 1,501 distinct lenders.
All figures on this page are as at 30 August 2026.
The corpus, in numbers
- Rating agencies live
- 7
- Rating actions parsed
- 70,337
- Borrowers
- 45,528
- Rated facilities
- 194,471
- Lender-exposure rows
- 179,857
- Distinct lenders named
- 1,501
- Earliest rating action held
- May 2019
A worked reading of one of those numbers: State Bank of India is named in 5,476 borrower annexures, as at 30 August 2026. That is a count of ingested annexures naming SBI. It is a floor, not a total. Why →
The seven agencies
TatvaRatings ingests publicly published credit rating press releases from seven SEBI-registered credit rating agencies operating in India:
- CARE Ratings (CareEdge)
- CRISIL Ratings
- ICRA
- India Ratings and Research
- Acuité Ratings and Research
- Brickwork Ratings
- Infomerics Valuation and Rating
These seven are the established houses whose published archives make up the rating output in circulation; they are not the whole of SEBI’s register, which listed nine registered credit rating agencies as at 22 August 2026, the other two being recent registrants we do not ingest. (SEBI register)
All seven are live. Each agency publishes only its own book, in its own layout, under its own notation, and with its own archive structure — which is why cross-agency work requires a single schema rather than seven readers. Direct links to each agency’s public archive are on Sources.
What “live” means here
An agency is live when its published press releases are being ingested on a continuing basis and parsed into the same schema as every other agency, with borrower and lender identities resolved across the whole corpus. It does not mean the corpus holds every release that agency has ever published — see the date range and the parse-outcome notes below.
Date range
The earliest rating action currently held is from May 2019. Coverage runs from there to the present.
The agencies themselves maintain longer archives — SEBI’s framework requires each agency to keep an archive of its issuer-specific disclosures on its website for at least ten years — and historical backfill continues. Where the corpus deepens, the earliest-action date on this page moves and the as-at date moves with it.
What is extracted from each release
Each ingested press release is parsed into six record families:
| Record family | Contents |
|---|---|
| Borrowers | Entity identity, prior and erstwhile names resolved, sector, state |
| Rating actions | Agency, rating, outlook, action type, date, and the transition from the prior action |
| Rated facilities | Instrument type, facility type, amount, long or short term |
| Lender-wise exposure | The annexure, parsed row by row — which lender, which facility, how much |
| Key financials | The figures the agency itself printed in the release |
| Rating drivers | Strengths, weaknesses and rating sensitivities, as published |
Provenance travels with every record: source agency, source document, and the date the action was published.
Annexure availability
The lender-wise annexure is the record family the rest of the product rests on, and it is also the one with the largest structural gap.
46.2% of rating actions publish a lender-wise annexure — 32,498 of 70,337, measured as at 30 August 2026. Disclosure of lender details depends on the issuer’s consent, which is why the share is well below 100%, and it varies enormously by agency:
| Agency | Rating actions | With an annexure | Share |
|---|---|---|---|
| Acuité | 4,856 | 4,467 | 92.0% |
| Infomerics | 8,808 | 5,863 | 66.6% |
| Brickwork | 76 | 45 | 59.2% |
| CARE | 12,403 | 6,501 | 52.4% |
| CRISIL | 22,046 | 10,136 | 46.0% |
| ICRA | 12,658 | 5,486 | 43.3% |
| India Ratings | 9,490 | 0 | 0.0% |
| All seven | 70,337 | 32,498 | 46.2% |
India Ratings’ 0.0% is genuine non-publication rather than a parsing gap: their annexure is instrument-wise and carries no lender column at all, verified across all 9,490 of their rating actions in this corpus — each one backed by its own source document. Where no annexure was published, there is no lender row to hold — the gap is in the source document, not in the parse.
The corpus holds 179,857 lender-exposure rows in total.
Every count on this page is a floor, not a total. It is the amount evidenced by published documents. The true position is larger, by a margin that varies by bank, sector and period, and that margin is not estimated here. Full specification of limits →
Refresh cadence
Ingest is continuous. New press releases are collected from all seven agencies on an ongoing basis as they publish, rather than on a fixed batch date.
Three separate clocks govern how current any given record is, and they should not be confused:
- Ingest lag — the interval between an agency publishing a release and TatvaRatings holding it parsed. Short, and the only one we control.
- Publication lag — the interval between a rating committee’s decision and the agency’s publication. Set by the agency.
- Review cycle — the interval between one rating action on a borrower and the next. A cooperating issuer’s rating is reviewed periodically rather than continuously, broadly once a year. This is the dominant clock, and it means a typical annexure record describes a position a year or more old.
An annexure is a dated snapshot, not a feed. It reflects the sanctioned position as at the rating action date — not today, and not an outstanding balance. No amount of ingest frequency compresses the review cycle. TatvaRatings is not exposure monitoring, and the refresh cadence of the corpus should not be read as the freshness of the underlying facts.
Backfill and parse outcomes
Two honest notes about how the corpus was built.
Backfill is incremental and ongoing. Historical releases are recovered agency by agency, and depth of history is not uniform across the seven. The single earliest-action date above is the floor of the whole corpus, not a guarantee of equal depth per agency.
Not every document parses cleanly. Source releases vary in structure, and some are scanned or malformed in ways that defeat extraction. Documents that fail extraction are recorded as failures and replayed, rather than silently dropped or partially loaded. A release that produced no annexure record is distinguishable in the data from one where no annexure was published — those are different facts and are not merged.
What is not covered
Stated here in brief; specified in full on Data limits.
- Unrated borrowers. The universe is borrowers with a published rating action from one of the seven agencies. Most Indian borrowers have never been rated.
- Unrated facilities of a rated borrower. Only facilities covered by a rating action appear.
- Outstanding balances. The annexure gives a sanctioned position, not a drawn one.
- Non-bank lending relationships that never appear in a rating annexure.
- Any non-public source. No supervisory returns, no bureau data, no confidential lender submissions.
Common questions
How far back does the data go?
The earliest rating action currently held is from May 2019, and coverage runs from there to the present. The agencies' own archives run longer — SEBI's framework requires each agency to keep issuer-specific disclosures on its website for at least ten years — and historical backfill is incremental and ongoing, so depth of history is not uniform across the seven agencies. Where the corpus deepens, the earliest-action date moves and the as-at date moves with it.
How fresh is the data?
Ingest is continuous — new press releases are collected from all seven agencies as they publish, not on a batch date. But three clocks govern currency: ingest lag (short, the only one we control), the agency's own publication lag, and the review cycle — a cooperating issuer's rating is reviewed broadly once a year, so a typical annexure record describes a position a year or more old. An annexure is a dated snapshot of the sanctioned position as at the rating action date, not a feed. Data limits specifies this in full.
Why are the figures floors, not totals?
Every count is the amount evidenced by published documents. Two filters sit between it and the real position: only 46.2% of rating actions publish a lender-wise annexure (disclosure depends on the issuer's consent), and the rated book is a minority of any lender's borrowers. The true position is larger, by a margin that varies by bank, sector and period, and that margin is not estimated. The full specification covers both filters.
Do you cover unlisted companies?
The boundary is rating status, not listing status. The universe is borrowers with a published rating action from one of the seven SEBI-registered agencies — 45,528 borrowers as at 30 August 2026 — and a company does not need to be listed to be in it. What is excluded is unrated borrowers, unrated facilities of a rated borrower, and any non-public source. Universe boundaries lists the exclusions.
Related
- Methodology — how the parse works
- Data limits — the full specification
- Sources — each agency’s public archive
- Talk to us — we will tell you what current coverage can and cannot answer for your question