MarketOS runs two independent repair loops. They use different technologies to solve different failure classes, and are never combined into a single "self-heals" number.
Loop COLLECTOR — Bright Data Scraper Studio repairs the archive-index scraper
Bright Data platform
HEAL PROPOSED → DONE
the platform's own success report
MarketOS
REJECTED
measured against an independent catalog oracle
MarketOS oracle
INVALID
measured archive coverage
measured coverage after the heal
0.00%
Collector ID
c_mt1iteyvmsxv50eri
Failure class
NAMING_CONVENTION
Before the heal
modern PPFAS_Monthly_Portfolio_Report_* filenames only (coverage 75.9%)
Proposed repair
all .xls/.xlsx anchors, every naming convention (coverage 0.0%)
Measured coverage
0.00%
Detected
2026-08-20 14:38:11.900548+00:00
Bright Data reported this heal as successful. MarketOS rejected it.
The repaired scraper returned a different document series — fortnightly debt disclosures rather than the monthly portfolio archive — so measured coverage against the MarketOS catalog oracle collapsed instead of improving. A repair is accepted on measured evidence, never on the platform's own success report.
The repaired scraper returned a different document series — fortnightly debt disclosures rather than the monthly portfolio archive — so measured coverage against the MarketOS catalog oracle collapsed instead of improving. A repair is accepted on measured evidence, never on the platform's own success report.
Loop INGESTION — MarketOS repairs its own workbook parser
Adapter
ppfas-mf
Parser issue
STRUCTURE
Strategy before
default
Strategy after
legacy-2018-percent-weights
Validation score
0.5000
Status
ACCEPTED
Bright Data involvement
none — this loop never touches a Collector ID
A repair changes the parse strategy, never a validation threshold, and the result is gated by the same four-rung ladder as an ordinary run.
Ingestion runs
25 clean · 5 suspect. A suspect run publishes nothing, so those periods are absent from the dataset rather than partially loaded.
| Period | Status | Rows | Validator detail |
|---|---|---|---|
| 2026-07-01 | SUCCESS | 596 | |
| 2026-07-01 | SUCCESS | 596 | |
| 2026-07-01 | SUCCESS | 596 | |
| 2026-07-01 | SUCCESS | 6420 | |
| 2026-07-01 | SUCCESS | 596 | |
| 2026-07-01 | SUCCESS | 596 | |
| 2026-07-01 | SUCCESS | 596 | |
| 2026-07-01 | SUCCESS | 596 | |
| 2026-07-01 | SUCCESS | 596 | |
| 2026-07-01 | SUCCESS | 6420 | |
| 2026-07-01 | SUCCESS | 596 | |
| 2026-06-01 | SUCCESS | 600 | |
| 2026-06-01 | SUSPECT | 0 | MT: 66% of non-money-market holdings are new, above the 60% limit - this indicates a parse failure rather than a portfolio decision |
| 2026-06-01 | SUCCESS | 600 | |
| 2026-06-01 | SUCCESS | 600 | |
| 2026-06-01 | SUSPECT | 0 | MT: 66% of non-money-market holdings are new, above the 60% limit - this indicates a parse failure rather than a portfolio decision |
| 2026-05-01 | SUCCESS | 615 | |
| 2026-05-01 | SUCCESS | 6453 | |
| 2026-05-01 | SUCCESS | 6453 | |
| 2026-05-01 | SUCCESS | 615 | |
| 2026-04-01 | SUCCESS | 612 | |
| 2026-04-01 | SUCCESS | 612 | |
| 2026-04-01 | SUCCESS | 612 | |
| 2026-04-01 | SUCCESS | 6499 | |
| 2026-04-01 | SUCCESS | 6499 | |
| 2026-04-01 | SUCCESS | 6499 | |
| 2026-04-01 | SUCCESS | 612 | |
| 2026-03-01 | SUSPECT | 0 | LQ: non-money-market holdings count moved from 27 to 15, outside the +/-40% band; NY: non-money-market holdings count moved from 20 to 7, outside the +/-40% ban |
| 2026-03-01 | SUSPECT | 0 | PPFCF: non-money-market holdings count moved from 82 to 47, outside the +/-40% band |
| 2026-03-01 | SUSPECT | 0 | LQ: non-money-market holdings count moved from 27 to 15, outside the +/-40% band; NY: non-money-market holdings count moved from 20 to 7, outside the +/-40% ban |
Data quality
Known conditions in the current dataset, queried from the database rather than asserted. MarketOS shows these so the product never presents false certainty.
10 period(s) excluded by validation
These periods failed a validation rung and were not published, so the ingested history skips them.
| 2018-12-01 | AdapterValidationError: Equity Scheme Portfolio: disclosure date not found in the first 6 title rows |
| 2024-12-01 | PPAF: weights sum to 1.6997999999999999, expected 1.0 +/- 0.02 |
| 2026-03-01 | PPFCF: non-money-market holdings count moved from 82 to 47, outside the +/-40% band |
| 2026-03-01 | PPFCF: non-money-market holdings count moved from 82 to 47, outside the +/-40% band |
| 2026-03-01 | PPFCF: non-money-market holdings count moved from 82 to 47, outside the +/-40% band |
| 2026-03-01 | LQ: non-money-market holdings count moved from 27 to 15, outside the +/-40% band; NY: non-money-market holdings count moved from 20 to 7, outside the +/-40% band |
| 2026-03-01 | PPFCF: non-money-market holdings count moved from 82 to 47, outside the +/-40% band |
| 2026-03-01 | LQ: non-money-market holdings count moved from 27 to 15, outside the +/-40% band; NY: non-money-market holdings count moved from 20 to 7, outside the +/-40% band |
| 2026-06-01 | MT: 66% of non-money-market holdings are new, above the 60% limit - this indicates a parse failure rather than a portfolio decision |
| 2026-06-01 | MT: 66% of non-money-market holdings are new, above the 60% limit - this indicates a parse failure rather than a portfolio decision |
1 scheme(s) appear under more than one code
The AMC renamed the scheme's sheet code mid-archive. Fund identity keys on that code, so history is discontinuous across the rename.
| Parag Parikh ELSS Tax Saver Fund (An open ended equity linked saving s | PPETSF, PPTSF |
2011 arbitrage holding(s) classified as equity
Accumulation signals therefore count hedged exposure alongside directional exposure.
3 industry label(s) split by footnote markers
The AMC appends footnote markers to some industry labels, so one industry can appear under more than one spelling and its sector total is split. This affects foreign holdings only; no affected label reaches the top sectors shown on the Command Center.
| Computer Software: Programming, Data Processing | 2 | 44 |
| Catalog/Specialty Distribution | 2 | 22 |
| Computer Software: Prepackaged Software | 2 | 22 |
166 synthetic instrument(s) excluded from company questions
TREPS, net receivables, cash and fund units are not companies and are filtered out of every research query.
MarketOS doesn't blindly trust the scraper. It verifies the scraper.
Collector output is scored against an independent catalog oracle, and parser repairs are gated by the same validation ladder as ordinary ingestion.
Collector output is scored against an independent catalog oracle, and parser repairs are gated by the same validation ladder as ordinary ingestion.