# New podcast feeds by hosting platform and start year

Five files behind the report "Spreaker passed Anchor in 2025; 73% of its new
feeds hold 3 or fewer episodes".

Source: the Podcast Index feed dump, table `podcasts`, 4,711,475 rows, dated
2026-08-16. Reference timestamp 1786838400 (2026-08-16 00:00 UTC). Every file is
produced by `tools/report_spreaker_facts.py` in one pass over the full table.

## new_feeds_by_host_year.csv
One row per hosting platform and start year, 2015 to 2026, for the 20
platforms with the most 2025 starts, plus an `(other)` row holding every other
platform. The 20 are picked by 2025 feed count alone, so the list includes
domains that are not commercial hosting platforms.
Columns: host, host_domain, start_year, feeds, le1_episodes, le1_share_pct,
le3_episodes, le3_share_pct, active_90d, active_90d_share_pct,
episode_counts_reliable. `le1` is episodeCount at one or zero, `le3` is
episodeCount at three or fewer. Buzzsprout rows carry feeds and active_90d
only, with the two episode columns empty and `episode_counts_reliable` set to
`no`.

## cohorts_all.csv
One row per start year, 2005 to 2026, for the whole index.
Columns: start_year, feeds, le1_episodes, le1_share_pct, le3_episodes,
le3_share_pct, active_90d, active_90d_share_pct, active_365d,
active_365d_share_pct, then the same feeds, le1, le3 and active_90d columns
with every Buzzsprout row removed. The first block counts every feed in the
cohort, Buzzsprout included, so the feeds column is the cohort itself; the
`_ex_buzzsprout` block is the one to use for any episode figure.

## cohort_2020_by_category.csv and cohort_2025_by_category.csv
One row per first Apple category for the 2020 and 2025 start cohorts.
Columns: category1, feeds, le3_episodes, le3_share_pct, feeds_ex_buzzsprout,
le3_episodes_ex_buzzsprout, le3_share_pct_ex_buzzsprout, active_90d,
active_90d_share_pct.

## language_2025.csv
The 15 languages with the most 2025 starts.
Columns: language_code, language, feeds_2025, feeds_ex_buzzsprout,
le3_episodes, le3_share_pct, active_90d, active_90d_share_pct,
spreaker_feeds, spreaker_share_pct, anchor_feeds, anchor_share_pct. The two
episode columns divide by feeds_ex_buzzsprout; the host shares divide by
feeds_2025.

Cells under 5 print `<5`. A share whose numerator is masked or whose base is
under 100 is left empty in these CSV files and prints n/a in the report tables.

Read METHOD.md before using any figure: the start year is the year of the
oldest episode still in the feed, not the year the show launched.

Data: CC BY 4.0, Podcastquery, September 2026