How we count
Every figure on this site is a count of something specific. This page says what, names every rule we apply, lists every row we threw away, and shows you how to check any of it yourself.
One relationship, not one mention
Our unit is one show × one advertiser. Within that, the figure we call appearances is the number of distinct publish dates we saw the advertiser on — not the number of episodes, and not the number of times the name occurs.
This matters more than it sounds. A sponsor block pasted into the footer of every episode description will produce thousands of mentions on a single show. Counted as mentions, that one relationship outweighs a hundred real ones. So we count the relationship once and record how many separate days it ran.
Whenever you read a sponsorship figure anywhere, the first question worth asking is which of these three it counts: shows, publishers, or mentions. The three answers can differ by a factor of a thousand.
Publishers, not feeds
One operator often puts out several shows, and a single advertiser buying that operator once will appear on all of them. So coverage is counted in publishers. Each board says both numbers, and where they differ it says so on the page.
What we keep out
We read the Podcast Index public feed database — all 4,711,475 feeds in it — and keep 30,936 (0.66%): English, updated in the last 120 days, at least 40 episodes, in the business-and-technology categories.
That is a deliberate choice and it is the main limit on this site: if your show is in another language or another category, we do not have it. We would rather cover a slice we can keep clean than a corpus we cannot.
What we threw away after reading it
These are separate rules with separate reasons. We report them separately on purpose — a single “we filtered N rows” would tell you nothing about whether we filtered the right things.
| Rule | Rows | Why |
|---|---|---|
| Templated show notes | 983 | Three or more advertisers on one show sharing an identical count and identical first and last date. Real sponsors do not start and stop together on the same days. |
| Standing footers | 458 | A block that appears on nearly every episode, dropped as a group rather than one row at a time. |
| The show’s own links | 1,412 | A domain on >=85% of a show’s episodes is that show’s own site, not a sponsor. 1,725 domains are filtered this way. |
| Broken links | 487 | Strings a publisher typed that are not addresses at all (http://support/, www.club). |
Where we say “at least”
7,127 relationships begin on the first day we were watching, which means they had already been running and we cannot see the start. Their length is a floor, not a measurement, so those panels read “at least 4.4 years” rather than “4.4 years”. We flag them rather than drop them: a floor is still information.
What a person decided
706 judgements in this data were made by a human, one row at a time, each with written evidence: whether a domain is an advertiser or a citation, whether two domains are the same company, whether a string is a typo for a real address, whether two feeds are the same show. No rule in this project is allowed to make those calls on its own — every one of them has produced a false positive we had to go back and undo.
What we do not have
No contact details. No rates. No spend figures. No audience or download numbers. If you need those, this is not where they live, and we would rather say so than imply otherwise.
We also do not rank shows, and we will never tell you a show has no advertisers of some kind. We can report what we saw; we cannot report what was absent.
Our recall, as best we can state it
We read what publishers write in their own episode descriptions. When a sponsor is only spoken aloud and never written down, we miss it. This failure is one-directional and that is why we can live with it: a missed sponsorship makes a list shorter, not wrong. So every name you see is real; the list is simply not exhaustive.
How to check us in five minutes
You do not have to take any of this on trust, and we would rather you did not. Pick any company on any board, note the years we show, open an episode from inside that window, and read the description — the sponsor line is usually there verbatim. Then try an episode from outside the window; it should not be there.
That procedure works on any sponsorship claim, not just ours, and it is worth knowing properly. The full five-minute method, with the four things that look like a sponsorship and are not →
Read through 2026-09-22. Numbers on this page come from the same export that builds the boards, so they change when the data does. Back to the boards →