What a conference agenda actually proves
Every record in this directory rests on a published agenda page. Here is exactly how much weight that page can carry, and the three places it gives way.
The rule, and why it is narrow on purpose
Only talks that have already happened are in this dataset. Somebody announced for a conference three months from now has been booked, not heard — and the difference is the whole product. Bookings fall through, line-ups change, and a directory that counts an announcement as a talk is a directory that tells you somebody spoke when they did not.
So a record needs a public page, published by the event, naming the person and the session. No page, no record. There is no "probably spoke" tier, and adding one would quietly convert the thing people come here to check into the thing they came here to avoid.
The same rule is why every entry shows the URL it came from rather than a verification badge. A badge asks you to trust us. A link lets you skip us.
Where it gives way, in the order we hit it
First: agendas are written for attendees, not for archives. A lot of them give a session a date only at the level of the whole event, so "spoke on the 14th" is often really "spoke during a conference that ran the 13th to the 15th". We store day precision only where the page genuinely stated a session date, and the rest carry the event's start date — which is why the site says "last spoke" rather than pretending to a calendar entry.
Second: a good half of agenda pages give a speaker a name and nothing else. Where the page published a talk title we store it as the event's own words; where it did not, the label you see is ours, and the record says so rather than presenting our phrasing as the conference's.
Third, and worst: agenda pages disappear. A conference re-skins its site for next year and last year's programme becomes a redirect to the new homepage. When the source goes, the evidence goes — so the entry is removed rather than quietly kept as an unverifiable claim. A record that existed last month may legitimately be gone today, and that is the method working rather than failing.
What we deliberately never take from the page
Contact details, fees and availability: none of it is here, and none of it is coming from an agenda anyway. A minority of records carry a link out to a LinkedIn profile and that is the entire extent of it — we have never read, copied or enriched from LinkedIn, and we do not know what anybody charges.
Where someone lives is the other one. An agenda tells you where the stage was. It tells you nothing about where the person flew in from, and the fields are named for the event's country rather than the speaker's specifically to stop that inference being made by accident. For an agency booking in Berlin, "has taken a German stage" is usually the better question anyway.
How to check any of this
Pick a speaker, open the source link on one of their talks, and search the page for their name. That is the entire verification procedure, and it is deliberately one you can perform without an account, an API key, or our cooperation.
If the page does not name them, that is a defect and we want it: the correction route is on every record, and so is the removal route for anyone who would rather not be listed at all.