Skip to content

Historical coverage of the corpus

This page documents how far back the neutral INAPI corpus reaches and from which year it is dense. Honesty is part of the contract: a consumer deserves to know "until when" the data is reliable before building on it. The corpus is neutral — zero consumer-specific fields or cross-references.

The figures below are a production snapshot of 2026-10-07 08:59 UTC (read-only, no sync running). The corpus grows every day: the live date is given by each response's meta.dataAsOf and meta.coverageThrough, not by this page. Corpus total in that snapshot: 845,426 trademarks (one per application number), all with a filing date (from 1994-09-09 to 2026-10-05).

The coverage boundary, in one sentence

The corpus is dense from 2009, in transition between 2006 and 2008 and residual before (1994-2005). This is neither a gap to fill nor a defect: it is the real shape of INAPI's public source, and we document it as-is so nobody is surprised.

Density by filing year

Span Coverage Volume per year
1994-2005 Residual from 1 to 474 filings/year
2006-2008 Transition 1,886 (2006), 11,438 (2007), 17,645 (2008)
2009 → today Dense between 34,527 (2018) and 52,743 (2012) filings/year
Year Rows Year Rows Year Rows Year Rows
1994 1 2003 121 2011 51,051 2019 34,584
1995 3 2004 214 2012 52,743 2020 43,508
1996 2 2005 474 2013 48,918 2021 50,079
1997 6 2006 1,886 2014 48,194 2022 42,739
1998 1 2007 11,438 2015 47,555 2023 46,042
1999 5 2008 17,645 2016 50,556 2024 44,525
2000 36 2009 39,935 2017 39,987 2025 50,547
2001 23 2010 45,106 2018 34,527 2026 42,911 (through 2026-10-05)
2002 64

Key reading: an analysis starting before 2009 will see fewer rows because the source has fewer rows, not because ingestion is missing.

How the read contract exposes this boundary

Every collection of the contract carries freshness in meta, with the same convention across resources:

{
  "results": [ /* … the collection rows … */ ],
  "limit": 20,
  "next": null,
  "hasMore": false,
  "total": 14,
  "meta": {
    "dataAsOf": "2026-09-20T19:05:06Z",
    "coverageThrough": "2026-09-17"
  }
}
  • dataAsOf — ISO timestamp of the last successful sync: how far ingestion ran.
  • coverageThrough — the enriched boundary: the most recent filing date among the already-enriched applications (with class/holder). Beyond that date live the fresh stubs from INAPI's record pages, seen but not yet enriched — reporting THEIR newest date would overstate coverage. It is not a bulletin date. One exception: GET /reports/daily-bulletin ≡ get_daily_bulletin carries in meta.coverageThrough the latest ingested Estado Diario bulletin, the same value as the coverage.through of its 422.

Both fields are always present in meta; null means "we do not know" (never filled with a plausible value). On individual objects freshness travels in headers (Iris-Data-As-Of, Iris-Coverage-Through). REST and MCP return the same freshness by construction (one shared query layer).

Catalogs: the date of THEIR source

In the catalogs, dataAsOf is not the trademark sync: it is the last successful load or check of that catalog's own source (since 2026-10-07). It is never the date of your request: that would claim a check that did not happen.

catalog meta.dataAsOf meta.coverageThrough
GET /nice-classes ≡ list_nice_classes last successful check of the headings against WIPO null
GET /legal-norms ≡ list_legal_norms last successful sync with BCN null
GET /holidays ≡ list_holidays last successful load of the calendar last day with a declared calendar
GET /sections ≡ list_sections the day the code legend was last read in an INAPI publication (start of the day, UTC) null

coverageThrough: null on a catalog means it has no "through": it is a list, not a series that moves forward in time. The exception is the holiday calendar: deadlines that end on or before that date are counted with declared holidays (see Holidays). The fee schedule and the economic indicators already carried the date of their source (INAPI and SII; see Document base).

Historical scoring inherits the boundary

What is computed over the corpus history —a class's statistics (get_class_stats ≡ GET /nice-classes/{niceClass}/stats), a party's outcomes (get_party_outcomes) or the indicative index (get_registration_signals)— inherits this same boundary: before 2009 the evidence is thin, and that is expected, not a gap. That is why those responses declare the period and the n they were computed on.

  • Every trademark has a status in the 2026-10-07 snapshot (0 rows without one; ~29.7k in July).
  • The 9-digit numbering series starting with 99 (21,671 rows, mostly foreign marks) is legitimate —consistent with international registrations / the Madrid Protocol into Chile— and is read like the normal series.

In short: the corpus is honest about its own boundary. Dense from 2009, residual before 2006, and every collection response states its date via meta.coverageThrough and meta.dataAsOf.