The India Points Devaluation Index
What did Indian points programmes take from you this year?
A named, recurring number computed from two datasets nobody else keeps: every dated ₹-per-point mark this desk has published, and 376 sourced programme changes going back to 1961. One figure per financial year, one method, frozen and published. Recompute it yourself — the YAML and the formula are both here.
what this index is, and what it deliberately does not measure fold that back up
The events behind it are on the devaluation tracker; the marks are on the valuation index.
Method v1.0 · 376 events, 1961-01-01 to 2026-09-30 in force · 164 currencies, 241 dated marks · the method
the series · every edition · what it is not · the method
FY2026 · April 2025 – March 2026 · cut share
78.9%
Of everything this desk logged happening to Indian card and loyalty programmes in FY2026, 78.9% of it — weighted by the severity weights we already publish — went against the cardholder. 40 cuts against 12 gains: one cut every 9.1 days. That is the highest of the 4 financial years in this archive carrying at least 20 logged changes.
And it is not one bad year. Cut share has risen every financial year for 3 running: FY2024 66.7% → FY2025 78.9% → FY2026 78.9%.
52
dated changes logged · the denominator
52.5
net severity points against the reader
−₹600
on 10,000 points in each of 69 currencies
1
currency whose ₹ value we could quantify moving
FY2027 so far (April 2026 – March 2027, still running): 71.4% cut share over 119 changes. It gets no page until it closes — a period that can still change is not an edition.
- This counts what we logged, not everything that happened. Every figure rests on 376 events this desk dated against a source. A change nobody published is not in here, and the number can only understate.
- The ₹ line is a floor, not a census. A currency only carries a dated ₹-per-point move where the archive QUANTIFIES one, so a programme that devalued without us pricing the move reads as flat. FY2026 carries 1 quantified move.
- Fee hikes count in the share and not in the rupees. A fee destroys real value, so it counts as a cut at its published weight — but it is not a ₹-per-point move, and pricing one into rupees would need a spend assumption this desk does not have.
The series
One row per Indian financial year, 1 April to 31 March. Cut share is the headline; the rupee column is the basket. A year carrying fewer than 20 changes is in the table and out of every superlative — a 90% share over nine events is true arithmetic and a false sentence.
| Financial year | Changes | Cuts | Cut share | Basket | ₹ / 10,000 pts |
|---|---|---|---|---|---|
| FY2027 running | 119 | 79 | 71.4% | 69 | +₹12,800 |
| FY2026 | 52 | 40 | 78.9% | 69 | −₹600 |
| FY2025 | 22 | 16 | 78.9% | 62 | — |
| FY2024 | 26 | 15 | 66.7% | 56 | −₹4,000 |
| FY2023 | 14 | 6 | 52.1%† | 50 | −₹1,200 |
| FY2022 | 21 | 9 | 38.6% | 44 | — |
| FY2021 | 7 | 2 | 38.5%† | 44 | — |
| FY2020 | 17 | 9 | 63.1%† | 41 | — |
| FY2019 | 13 | 6 | 53.2%† | 40 | −₹1,000 |
| FY2018 | 9 | 8 | 95.9%† | 39 | — |
| FY2017 | 10 | 6 | 58.6%† | 37 | — |
| FY2016 | 11 | 5 | 57.0%† | 36 | — |
the earlier archive — 19 financial years, 1961 onward
| Financial year | Changes | Cuts | Cut share | Basket | ₹ / 10,000 pts |
|---|---|---|---|---|---|
| FY2015 | 8 | 3 | 45.5%† | 1 | — |
| FY2014 | 4 | 1 | 16.7%† | 1 | — |
| FY2013 | 7 | 4 | 74.5%† | 0 | — |
| FY2012 | 5 | 3 | 53.1%† | 0 | — |
| FY2011 | 2 | 1 | 71.4%† | 0 | — |
| FY2010 | 2 | 0 | 0.0%† | 0 | — |
| FY2009 | 4 | 1 | 16.7%† | 0 | — |
| FY2008 | 3 | 0 | 0.0%† | 0 | — |
| FY2007 | 4 | 0 | 0.0%† | 0 | — |
| FY2006 | 3 | 0 | 0.0%† | 0 | — |
| FY2002 | 1 | 0 | 0.0%† | 0 | — |
| FY1999 | 3 | 0 | 0.0%† | 0 | — |
| FY1998 | 1 | 0 | 0.0%† | 0 | — |
| FY1994 | 3 | 0 | 0.0%† | 0 | — |
| FY1993 | 1 | 0 | 0.0%† | 0 | — |
| FY1987 | 1 | 0 | 0.0%† | 0 | — |
| FY1981 | 1 | 0 | 0.0%† | 0 | — |
| FY1980 | 1 | 0 | 0.0%† | 0 | — |
| FY1961 | 1 | 0 | 0.0%† | 0 | — |
† Fewer than 20 logged changes: the share is real arithmetic over a field too small to rank, so the row is in the table and out of every superlative on this site.
A "—" in the rupee column is a period in which no currency in that period's basket carried a dated ₹-per-point move. It is not a zero we measured; it is a move we could not quantify.
Every edition
A period gets a permanent page once it CLOSES and once it carries at least 10 dated changes. Below that the numbers are still in the series above — a ratio over four events is noise wearing a percent sign, and it does not get a headline. Annual editions are the record; quarterly editions are the cadence.
- FY2027 Q1 Apr–Jun 2026 79.5% cut share · 36 changes
- FY2026 April 2025 – March 2026 78.9% cut share · 52 changes
- FY2026 Q4 Jan–Mar 2026 75.2% cut share · 21 changes
- FY2026 Q1 Apr–Jun 2025 91.2% cut share · 16 changes
- FY2025 April 2024 – March 2025 78.9% cut share · 22 changes
- FY2024 April 2023 – March 2024 66.7% cut share · 26 changes
- FY2023 April 2022 – March 2023 52.1% cut share · 14 changes
- FY2022 April 2021 – March 2022 38.6% cut share · 21 changes
- FY2020 April 2019 – March 2020 63.1% cut share · 17 changes
- FY2019 April 2018 – March 2019 53.2% cut share · 13 changes
- FY2017 April 2016 – March 2017 58.6% cut share · 10 changes
- FY2016 April 2015 – March 2016 57.0% cut share · 11 changes
What this index is not measuring
It is not weighted by how many readers hold what. Nobody publishes Indian membership by programme, and a weight built from card counts measures product supply, not wallets. Every currency counts once. Each edition prints how many cards earn a currency and how many transfer into it, as context you can re-weight by hand — never as a weight applied behind your back.
It is not a forecast. The desk grades a programme's forward nerf risk separately, on the methodology page. This is a backward-looking statement about a closed period, and the two must not be read as one.
It is not the size of the damage. Cut share counts how much of a period's change went the wrong way; it says nothing about how far. The rupee line is the size, and it is deliberately conservative: the worst single move in the whole archive is Marriott Bonvoy, 2022-03-29, −47.6% — −₹5,900 on 10,000 points.
It will not always agree with the tracker's seismograph, and here is exactly why. That trace names 2026 the heaviest year in the record — 150.6 severity points of cuts, and this index reproduces that figure to the decimal from the same events. But heaviest is a VOLUME, and volume grows with how much a desk logs. 2026 is also the year we logged the most GAINS (58.8 points of them), so on the ratio this index reports it sits #3 of 5, at 71.9%, behind 2024's 81.3%. Both numbers are right. Only one of them is about the market rather than about us — and the seismograph counts calendar years while the index counts financial ones. The build asserts the two reduce the same events to the same severity loads, so they can differ in what they ASK and never in the arithmetic.
It is restatable, and that is intended. Editions are recomputed from the archive every time the site builds. Back-dated coverage — a currency indexed today whose series reaches 2015 — can move a past edition's basket, exactly as a real index restates. The method page carries the version; this one carries the archive's own vintage in the head above.
Reproduce it: the whole series as JSON · the formula at /devaluation-index/method · the inputs are data/changelog.yaml and data/valuation-history.yaml in the open repository.