Open data
Dataset releases
Frozen, dated snapshots of the four open datasets. A release URL never changes its contents, so a citation made today still resolves to the same numbers next year.
Which URL should you cite?
The endpoints under /data are live: they serve the current catalogue and will return different numbers tomorrow. They are the right thing to build a tool against, and the wrong thing to cite in something that has to stay checkable.
The releases below are immutable. Each one is a fixed copy at a permanent URL, published under CC BY 4.0. Cite these — a reference that cannot be retrieved is worse than no reference, and it is the first thing a careful reader will test.
What has changed since v2026-09-17
v2026-09-17 is frozen and always will be — that is what makes it citable. The live endpoints under /data/ have moved on since 2026-09-17, and this is how far. Computed on every rebuild against the release files themselves, so it cannot drift from what the release actually contains.
cards 280 cards frozen, 280 live — 0 added, 0 removed, 231 changed.
Fields that moved: lastVerified on 231, forexMarkupPct on 14, maxAgeYears on 11, minSalaryRs on 8, minAgeYears on 5, network on 2, and 3 more fields (of 18 compared).
Present on both sides and deliberately not diffed (nested or list-valued, where a reordering is not a change): perks, rewards.
transfers 156 edges frozen, 161 live (+5) — 8 added, 3 removed, 6 changed.
Added: AXIS_EDGE_MILES -> ACCOR_ALL, AXIS_EDGE_MILES -> ASIA_MILES, AXIS_EDGE_MILES -> HILTON_HONORS, AXIS_EDGE_MILES -> MILES_AND_MORE, AXIS_EDGE_RP -> ACCOR_ALL, AXIS_EDGE_RP -> ASIA_MILES and 2 more
Removed: AXIS_EDGE_MILES -> CLUB_ITC, AXIS_EDGE_RP -> CLUB_ITC, SBI_TRAVEL_CREDITS -> FLYING_BLUE. A row leaving is not a mistake in the release — the release is a snapshot of a day on which it existed.
Fields that moved: cap on 6, lastVerified on 6 (of 15 compared).
valuations 231 card rows frozen, 236 live (+5) — 5 added, 0 removed, 1 changed.
Added: au-tejas, jupiter-edge, rbl-patanjali-swarn, rbl-patanjali-vishisht, sc-manhattan-platinum
Fields that moved: programName on 1 (of 8 compared).
changes 1,629 events frozen, 1,735 live (+106) — 113 added, 7 removed, 3 changed.
Added: amex-centurion | 2026-09-30 | Reward Multiplier discontinued; ShopWise e-vouchers keep their bonus, amex-gold | 2026-09-30 | Reward Multiplier discontinued; ShopWise e-vouchers keep their bonus, amex-platinum-charge | 2026-09-30 | Reward Multiplier discontinued; ShopWise e-vouchers keep their bonus, amex-platinum-reserve | 2026-09-30 | Reward Multiplier discontinued; ShopWise e-vouchers keep their bonus, amex-platinum-travel | 2026-09-30 | Reward Multiplier discontinued; ShopWise e-vouchers keep their bonus, axis-atlas | 2026-09-22 | Accor ALL returns as a transfer partner, in Group A and 107 more
Removed: axis-supermoney | 2026-01-01 | 3% super.money UPI cashback throttled to 1× base earn; sub-₹100 spends excluded, axis-supermoney | 2026-01-01 | Flat ₹500 monthly cashback cap removed — 1% base now uncapped, cred-sovereign | 2025-12-21 | Complimentary air-accident insurance withdrawn, jupiter-edge | 2026-03-21 | Shopping 10% category re-capped and gift cards / bills / groceries excluded, kiwi | 2024 | Yes Bank 'Klick' becomes the mainstream issuer, utkarsh-supercard | 2025-08-11 | Base cashback doubled to 1% on UPI and non-UPI spends and 1 more. A row leaving is not a mistake in the release — the release is a snapshot of a day on which it existed.
Fields that moved: after on 2, verificationTier on 2, verificationUrl on 2 (of 18 compared).
Cite the frozen URL when the number has to stay retrievable, and the live endpoint when you want today’s figure. This section exists so that choice is informed rather than a guess — and so nobody has to download both and diff them by hand.
| Dataset | Rows | Data as of | Schema | Download |
|---|---|---|---|---|
| India Credit Card Facts Dataset | 280 cards | 2026-09-17 | v6 | JSON · CSV |
| India Credit Card Change History Dataset | 1,629 events | 2026-09-10 | v6 | JSON · CSV |
| India Credit Card Points Transfer Graph Dataset | 156 edges | 2026-09-10 | v6 | JSON · CSV |
| India Credit Card Point Valuations Dataset | 231 cards | 2026-09-05 | v6 | JSON · CSV |
Citation for india credit card facts dataset. Each file carries its own citation string in the payload.
Source: <a href="https://cardadvisor.in/data/v/2026-09-17/cards.json">India Credit Card Facts Dataset v2026-09-17</a>, CardAdvisor (CC BY 4.0).
v2026-08-16
/data/v/2026-08-16/| Dataset | Rows | Data as of | Schema | Download |
|---|---|---|---|---|
| India Credit Card Facts Dataset | 280 cards | 2026-08-15 | v3 | JSON · CSV |
| India Credit Card Change History Dataset | 1,617 events | 2026-08-13 | v3 | JSON · CSV |
| India Credit Card Points Transfer Graph Dataset | 153 edges | 2026-07-02 | v3 | JSON · CSV |
| India Credit Card Point Valuations Dataset | 233 cards | 2026-07-02 | v3 | JSON · CSV |
Errata — the release stands as published; the correction is recorded here
2026-09-04 · data · transfers carried `confidence: HIGH` on every edge whose confidence was never authored. The loader defaulted an absent value to HIGH and the enum's own documentation said absent MEANS HIGH, so silence published as the strongest claim available: "verified against the programme's own published terms". Measured on 2026-09-04 against the same source files, 61 of 156 edges had no authored confidence and 7 of those had no sourceUrl either. Readers of this release should treat any HIGH without a sourceUrl as unverified. From schema v4 the default is the new value UNVERIFIED, and only an authored value can say HIGH. Resolved in vschema v4.
2026-09-04 · data · the transfers dataset published one `asOf` for the whole graph, taken from a hand-maintained `lastReviewed` key (on the live endpoint the citable `version` token was derived from it too; a frozen file's `version` is its release date). Every edge already carried its own `lastVerified` and the dataset dropped it. On 2026-09-04 the graph-level date was 2026-07-02 while the 156 edges spanned eight dates up to 2026-08-27, so the published date was wrong for 151 rows — understating freshness here, and capable of overstating it after any later edit. From schema v4 each edge carries its own `lastVerified`. From schema v5 the dataset-level date is a window derived from the published rows — `asOf` is the newest per-row check and `verifiedFrom` the oldest — so no hand-typed key sets it. Resolved in vschema v4 (per-edge dates); schema v5 (dataset date window).
2026-09-17 · data · valuations published the same hand-maintained `lastReviewed` date, 2026-07-02, as its dataset-level `asOf`. On 2026-09-04 that date described 8 of the 36 partner-programme valuations these rows are derived from; 2 had been checked earlier and 26 later. From schema v5 the dataset-level date is a window derived from those partner checks — `asOf` is the newest and `verifiedFrom` the oldest — so no hand-typed key sets it. Resolved in vschema v5.
2026-09-04 · schema · the field dictionary described `lastVerified` as "Date this card record was last verified against issuer sources". It is a CHANGE stamp: the V29 database triggers re-stamp it on any edit to the card or its reward, perk and related-card rows, so a corrected typo re-dated the record and published it as a fresh verification. No re-read of an issuer document is implied by any value in this column, in this release or any other. The catalogue does carry a real verification stamp — `verifiedOn` with the document it was checked against in `verifiedAgainst`, split out by migration V142 precisely because the two had been one field — and as of 2026-09-04 no active card carries one, so no card in this release had been verified in that sense. Readers should treat `lastVerified` as last-modified. The live dictionary was corrected on 2026-09-04; frozen releases are never rewritten. Resolved in vlive dictionary 2026-09-04; next frozen release.
Citation for india credit card facts dataset. Each file carries its own citation string in the payload.
Source: <a href="https://cardadvisor.in/data/v/2026-08-16/cards.json">India Credit Card Facts Dataset v2026-08-16</a>, CardAdvisor (CC BY 4.0).
v2026-08-15
/data/v/2026-08-15/| Dataset | Rows | Data as of | Schema | Download |
|---|---|---|---|---|
| India Credit Card Facts Dataset | 280 cards | 2026-08-14 | v2 | JSON · CSV |
| India Credit Card Change History Dataset | 1,613 events | 2026-08-05 | v2 | JSON · CSV |
| India Credit Card Points Transfer Graph Dataset | 154 edges | 2026-07-02 | v2 | JSON · CSV |
| India Credit Card Point Valuations Dataset | 231 cards | 2026-07-02 | v2 | JSON · CSV |
Errata — the release stands as published; the correction is recorded here
2026-08-16 · schema · cards carried `expertRating`, `welcomeBonusValueRs` and `perks[].estimatedAnnualValueRs`, and valuations/transfers carried route-level point values (`bestPaisaPerPoint`, `cashPaisaPerPoint`, `voucherPaisaPerPoint`, `travelPortalPaisaPerPoint`, `transferPaisaPerPoint`, `valueGapPaisaPerPoint`, `bestRouteName`, `valuePaisa`, `paisaPerPoint`, `destValuePaisa`) — CardAdvisor's editorial valuation model, which the datasheet stated was excluded. The datasheet was right about the intent and the files were wrong. From schema v3 the open dataset carries facts plus one point value per card/row; the model is published on the site. Resolved in v2026-08-16.
2026-09-04 · data · transfers carried `confidence: HIGH` on every edge whose confidence was never authored. The loader defaulted an absent value to HIGH and the enum's own documentation said absent MEANS HIGH, so silence published as the strongest claim available: "verified against the programme's own published terms". Measured on 2026-09-04 against the same source files, 61 of 156 edges had no authored confidence and 7 of those had no sourceUrl either. Readers of this release should treat any HIGH without a sourceUrl as unverified. From schema v4 the default is the new value UNVERIFIED, and only an authored value can say HIGH. Resolved in vschema v4.
2026-09-04 · data · the transfers dataset published one `asOf` for the whole graph, taken from a hand-maintained `lastReviewed` key (on the live endpoint the citable `version` token was derived from it too; a frozen file's `version` is its release date). Every edge already carried its own `lastVerified` and the dataset dropped it. On 2026-09-04 the graph-level date was 2026-07-02 while the 156 edges spanned eight dates up to 2026-08-27, so the published date was wrong for 151 rows — understating freshness here, and capable of overstating it after any later edit. From schema v4 each edge carries its own `lastVerified`. From schema v5 the dataset-level date is a window derived from the published rows — `asOf` is the newest per-row check and `verifiedFrom` the oldest — so no hand-typed key sets it. Resolved in vschema v4 (per-edge dates); schema v5 (dataset date window).
2026-09-17 · data · valuations published the same hand-maintained `lastReviewed` date, 2026-07-02, as its dataset-level `asOf`. On 2026-09-04 that date described 8 of the 36 partner-programme valuations these rows are derived from; 2 had been checked earlier and 26 later. From schema v5 the dataset-level date is a window derived from those partner checks — `asOf` is the newest and `verifiedFrom` the oldest — so no hand-typed key sets it. Resolved in vschema v5.
2026-09-04 · schema · the field dictionary described `lastVerified` as "Date this card record was last verified against issuer sources". It is a CHANGE stamp: the V29 database triggers re-stamp it on any edit to the card or its reward, perk and related-card rows, so a corrected typo re-dated the record and published it as a fresh verification. No re-read of an issuer document is implied by any value in this column, in this release or any other. The catalogue does carry a real verification stamp — `verifiedOn` with the document it was checked against in `verifiedAgainst`, split out by migration V142 precisely because the two had been one field — and as of 2026-09-04 no active card carries one, so no card in this release had been verified in that sense. Readers should treat `lastVerified` as last-modified. The live dictionary was corrected on 2026-09-04; frozen releases are never rewritten. Resolved in vlive dictionary 2026-09-04; next frozen release.
Citation for india credit card facts dataset. Each file carries its own citation string in the payload.
Source: <a href="https://cardadvisor.in/data/v/2026-08-15/cards.json">India Credit Card Facts Dataset v2026-08-15</a>, CardAdvisor (CC BY 4.0).
Notes for data consumers
- Release date is the cut date, not the data date. The four datasets refresh on different cycles, so each file carries its own
asOf— the table above shows them separately rather than averaging them into one number. - Releases are never rewritten. The cutter refuses to overwrite an existing date. If a figure in a published release turns out to be wrong, the correction lands in the next release, the old one stays as it was, and an erratum is recorded against it above (and in the machine-readable index).
- A machine-readable index of every release lives at /data/v/index.json.
- Licence: CC BY 4.0 — use it commercially, just credit CardAdvisor.