SMVaddadi/propensity-data-fca-value-measures

fca_value_measures — mirror

A convenience mirror of a public dataset, pushed by propensity.data.hf_mirror so that benchmark runs do not depend on a flaky upstream origin.

This is a mirror, not a grant of rights

Copyright and licence remain with the original source named below. Mirroring moves bytes; it does not and cannot change the terms they arrived under. If you intend to redistribute this further, or to use it commercially, read the original terms — not this card, and not the license: field above, which is our best reading and may be wrong.

Provenance

• Source: FCA General Insurance Value Measures, the five modern editions Jul-Dec 2021 through 2025. Contains public sector information licensed under the Open Government Licence v3.0 (https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/) via clause 2.2 of fca.org.uk/legal, read in prose on the rights holder's own page. Clause 2.2 also reserves third-party copyright material. WHY THIS ONE MATTERS: it is the ONLY source in this collection where GAP (Guaranteed Asset Protection) insurance is a separately reported product -- GAP insurance (Add-on) and GAP insurance (Stand-alone), 157 firm-level rows across the five editions -- and the only one where an ordinal target is READ from the data rather than invented, because the FCA published the banding. WHAT IS PUBLISHED: firms arm 4,526 rows x 16 cols, one row per firm x product-category x period-end year, 301 casefold-distinct free-text firms, 53 product categories, four BANDED outcomes plus a derived ordinal rank per band. products arm 266 x 12, product-category x year, seven genuinely numeric measures over 39 categories -- the continuous counterpart the Firms tab cannot provide. LICENCE SCOPE, AND IT HAS A REAL COST: only the numeric Firms and Products/Product Table tabs are OGL under clause 2.2. Each workbook's narrative Information tab is OUTSIDE the grant under clause 2.3 and is NOT mirrored -- so the FCA's own 3,000-average-policies reporting floor, five-firm suppression rule and reporting-window statements DO NOT TRAVEL with the data and are restated here instead. A firm's absence from a product is therefore not evidence that it does not write that product. FIVE MEASURED CAVEATS. (1) No firm-level value is ever a number -- only the band -- so regression on the underlying quantity is impossible. The *_ordinal columns are lower-edge/step RANKS, not values, and no midpoint is supplied or implied. The 5pp / GBP500 grid is a rule, not an invariant: across 119 distinct band strings the widths are 5pp x301, GBP500 x191 and 10pp x3, the exception being '50 - 60%' which appears 7 times in the 2022 edition only and collides with '50 - 55%' on the ordinal grid; those rows are flagged. (2) The entity key is free-text Firm Name with NO FRN anywhere in the workbook, and the FCA re-cases three firms between editions, so 304 raw strings reduce to 301 casefold-distinct and a raw-name join would split each of those three into two entities. (3) Editions overlap by one year and RESTATE it: 1,605 of 4,526 firm rows repeat a (firm, product, year) key across editions and the editions disagree -- on the products arm 58 of 228 overlapping cells differ by over 1% and 25 by over 5%, far beyond the rounding difference. (4) The column ORDER swaps at the 2024 cut-over (acceptance/frequency and complaints/payout both trade places), so a positional read puts an acceptance band into the frequency column and nothing raises. (5) The 2024 edition's suppression marker is a NON-BREAKING SPACE on 86 band cells where the other editions leave the cell empty, so testing only isna() keeps it as a band label. WHAT THIS IS NOT: not claim-level or policy-level, not a training corpus (4,526 censored aggregate rows), not a price or premium-rate table, not regression-ready at firm level, not longitudinally joinable on a stable firm identifier, and NOT the 2016-2019 FCA pilot publication -- that is a different workbook family with four products and no GAP at all, refused by name. Contains no personal data.
• Origin URL: https://www.fca.org.uk/data/general-insurance-value-measures-data
• Loader: propensity.data.public_loaders.load_fca_value_measures
• Extract: full pull from origin
• Reported source: fca_value_measures
• Rows pushed: 4,526
• Columns: 16

Licence and redistribution

• Licence as found: other (uk-ogl-v3)
• Redistribution verdict: open
• Basis: FCA, UK Open Government Licence v3.0, SCOPED PER TAB. Read 2026-11-10 on the rights holder's own page https://www.fca.org.uk/legal (HTTP 200, 189,082 bytes; page states "Last updated: 12/10/2026"). Clause 2.2 verbatim: "The UKOGL replaces conditions 3.2 to 3.5 of these terms in relation to any charts, figures, graphs, diagrams, tables and numerical datasets published within the Data section of this site and in respect of which the FCA owns the copyright. Where we have identified any third-party copyright information, you will need to obtain permission from the copyright holders concerned." Doubly conditional. Clause 2.3 withholds it from prose, verbatim: "conditions 3.2 to 3.5 of these terms continue to apply to any narrative or explanatory text in any documents or pages published in or made available via the Data section of this site." Conditions 3.2-3.5 limit use to "personal use or for use within an individual firm or organisation" and 3.2.3 says extracts "must not ... be distributed, sold or otherwise circulated to third parties". SCOPE ENFORCED IN CODE, not noted: only the numeric Firms and Products/Product Table tabs are mirrored. Each workbook's narrative Information tab (Data Note in the 2021 edition) is excluded under 2.3, and the loader RAISES on arm in {information, Information, data_note, notes} before any I/O, quoting the clause. An earlier draft silently served the Product Table under the name information because dispatch fell through, which is why the check sits at the entry point. THE EXCLUDED TAB IS SUBSTANTIVE, and this is a real consequence of the scope: it is where the FCA states the 3,000-average-policies reporting floor, the five-firm suppression rule and the reporting window. A redistribution of this data therefore CANNOT carry the publisher's own caveats, so the card restates them. A firm's absence from a product is not evidence it does not write it. Per-edition caveat recorded in code: clause 2.2 is conditional on publication "within the Data section of this site", and the 2022 edition is indexed under /publications/data/ rather than /data/. Recorded per edition rather than assumed. No personal data: firm-level banded aggregates.

Citation

Financial Conduct Authority. General insurance value measures data, Jul-Dec 2021 to 2025. Contains public sector information licensed under the Open Government Licence v3.0 (fca.org.uk/legal clause 2.2). Firm-level outcomes are published as bands, not values.

Loading

from propensity.data.hf_mirror import load_mirrored

result = load_mirrored("fca_value_measures", org="SMVaddadi")
result.frame.head()

load_mirrored falls back to the original loader (origin, then its own offline replica) if this mirror is unreachable, so it is safe to depend on.