Skip to main content
Every day we publish an edition of each entity chart: a ranked list of the knowledge-graph entities podcasts talked about most over the window. The per-category charts order entities by how many distinct shows mentioned them in ordinary dialogue; the composite all chart draws on a wider curated pool of entity types — people, books, music, brands, companies and organizations, sports, events, products, and more, as well as the three published categories — and blends that reach with each entity’s 52-day elevation, so a story that is unusually big for its subject can outrank a perennial with more shows; the guests chart ranks people by the distinct shows they appeared on. Each entry carries its movement since the previous edition; on the mention charts’ weekly editions, a momentum classification as well; and — once the narrative pipeline has run and there is something worth explaining — a one-line explanation of why it moved, backed by a verbatim quote you can play from the source episode. Fields that do not apply to an entry are omitted from the JSON rather than sent as null. Think of a music chart where the artists are entities and the airplay is podcast dialogue. The category charts are volume-ranked: an entity that is mentioned on hundreds of shows every week sits near the top every week, and all tempers that reach with elevation rather than replacing it. Acceleration is a separate axis, reported as accel_multiple on entries with enough history (never on the guests chart) and available as an ordering with sort=acceleration, so a from-zero breakout deep in the chart can be found without it having to outrank the perennials. Use these endpoints when you want to:
  • Render a “what podcasts are talking about” leaderboard, overall or per category.
  • Track an entity’s chart position and its movement since the previous day’s edition.
  • Surface the story behind a move with a receipt: the explanation, the quote, the episode, and the offsets to play it.
  • Find the guests appearing across the most shows this week, with their appearances and a highlight clip.

​
List the charts

Response (truncated)
Every chart lists with the headline state of its current edition and the entity at rank 1. Pass window=24h to summarize the daily editions instead of the weekly ones. The category_slug is the {category} path parameter of the chart endpoint. leader is omitted on charts the caller may not read (see Access).

​
Get a chart

Response (truncated)

​
Query parameters

An unknown {category} is a 404 that points you back at the list endpoint.

​
Reading an entry

​
The guests chart

guests has source: "guest_appearances": it ranks people by the distinct shows they appeared on in the window, so several fields read differently. subject.type is person; mention_count is always 0 and accel_multiple and signals are absent, because guest appearances are counted per episode, not per dialogue line. In their place each entry carries:

​
Access

An API key reads every chart at full depth, with every field described above. A session without paid Radar access — signed out, or signed in on plans that do not yet provide it — reads a preview instead:
  • the composite all chart only, 10 entries deep regardless of limit;
  • signals keeps state but omits the trend_* and elevation_* operands;
  • the per-category charts are refused, and the list endpoint omits their leader. A signed-out session gets 401 auth_required; a signed-in one gets 402 paid_plan_confirmation_required, whose resolve.action says whether to confirm the plan it holds or add Radar.
Treat the operand fields and signals itself as optional keys when rendering, and do not assume a response holds as many entries as you asked for.

​
Editions and windows

Charts are recomputed once a day. An edition’s ranking and counts are fixed at computation, but the narrative pipeline keeps writing to it afterwards: status, enriched_at, and entries’ why fill in over the following minutes, so re-read a computed edition rather than caching it for the day. edition_date is the Eastern-time calendar date the edition describes; computed_at is when positions were fixed. status is computed while the narrative pipeline is still running and enriched once the pipeline has processed every candidate — entries it skipped still have no why, and an entry’s why may already be present on a computed edition. methodology_version increments when the ranking formula changes; editions across a version change are not directly comparable. Two windows are published for the category charts: 7d (the default, a weekly view) and 24h (a daily view). The composite all chart is weekly only.

​
How the charts are computed

  • Distinct podcasts, not episodes. One show that reads a book aloud over ninety episodes counts once. Episode and mention counts are shown but never sort.
  • Organic mentions only. An episode counts when its dialogue names the entity outside ad reads. This is the same gate the mentions endpoint applies, so following an entry through to its mentions always finds results.
  • Read from persisted editions. A request never touches raw mention data, so reads are fast and yesterday’s chart is exactly what yesterday’s readers saw.
  • Explanations carry receipts. A why that quotes an episode names it and, while that episode still exists (verifiable), links it with playback offsets; an explanation written without a quote carries text alone. Entities whose mentions look mislinked are flagged in caveats, not silently dropped.

​
Choosing the right endpoint