# The data

What we hold, how we counted it, and what we know is wrong. Every number on this site is counted the way this page describes; where we cannot do something, it says so and names the issue. To query it yourself, see [the API guide](/developers).

The corpus was measured 10/22/2026, 12:47:28 PM. Freshness, further down, is read live when you load this page.

## Coverage

Distinct events starting from now onward, counted once per platform per metro, with hosts on the suppression register left out. A metro is where our region boxes placed the event, not where the platform said it was.

302490 distinct events, from 341586 stored listings.

| Metro | Events |
| --- | --- |
| [New York](/metros/new-york) | 36428 |
| [Bay Area](/metros/bay-area) | 33941 |
| [Los Angeles](/metros/los-angeles) | 30702 |
| [Miami](/metros/miami) | 26291 |
| [Atlanta](/metros/atlanta) | 17398 |
| [Washington, DC](/metros/washington-dc) | 16311 |
| [Dallas–Fort Worth](/metros/dallas) | 14390 |
| [Chicago](/metros/chicago) | 13712 |
| [Toronto](/metros/toronto) | 11902 |
| [Boston](/metros/boston) | 11088 |
| [Denver](/metros/denver) | 9423 |
| [Phoenix](/metros/phoenix) | 9153 |
| [San Diego](/metros/san-diego) | 8710 |
| [Austin](/metros/austin) | 8593 |
| [Seattle](/metros/seattle) | 7546 |
| [Baltimore](/metros/baltimore) | 7176 |
| [Vancouver](/metros/vancouver) | 6956 |
| [Minneapolis–Saint Paul](/metros/minneapolis) | 4385 |
| [Portland](/metros/portland) | 3581 |
| [St. Louis](/metros/st-louis) | 2356 |

## Duplicates

Two listings with an agreeing coordinate and the same start instant are read as one event when they also share a title or a host, on the same platform or across two. Collapsed is how many listings that hides within each platform; distinct events counts an event on two platforms once. Held is how many we still store and can still show you: nothing is merged or deleted, because a merge made on an inference is data loss with a plausible cover story (D33). Listings worded differently under differently named hosts are not matched, so some cross-platform duplicates remain.

| Listings held | Distinct events | Collapsed | Rate |
| --- | --- | --- | --- |
| 341586 | 302800 | 38786 | 11.4% |

## Descriptions

How much prose we hold, per platform, over the same distinct events. Undescribed means the row has no description at all; brief means 1 to 199 characters, which is what a search feed hands out; full means longer than that, which is what an event page carries.

| Events | None | Brief | Full | Described |
| --- | --- | --- | --- | --- |
| 302800 | 4869 | 273667 | 23954 | 98.3% |

## Where the events are

How much of the corpus we could place. A coordinate is a latitude and a longitude from the source. Inside a region box means our own bounding boxes placed it; in no metro means neither a box nor a city row could, so the event is stored unplaced and shown here rather than guessed at.

| Of 302490 distinct events | Events | Share |
| --- | --- | --- |
| Has a coordinate from the source | 299142 | 98.9% |
| Placed by one of our region boxes | 257300 | 85.1% |
| Placed in a metro, by a box or a city row | 280042 | 92.6% |
| In no metro — stored unplaced, never guessed at | 22448 | 7.4% |

## Freshness

Hours since each source last completed a run that kept its data, read live at the moment you loaded this page rather than from the nightly row. A source that has never completed one says never. Past 36 hours the row is red.

95 of 177 sources have not succeeded in 36 hours.

## Hosts we withhold

Hosts an operator ruled out, and how many upcoming events each is currently withholding from every public count and list. Nothing is deleted: the events are still stored, still countable, and one row deleted here puts them all back.

| Host | Reason | Withheld | Why |
| --- | --- | --- | --- |
| iCertGlobal | certification-mill | 1770 | 800 events across 52 cities. The largest single host in the corpus, larger than any real organizer in it. Templated exam-prep syllabi (PMP, CISSP, CAPM, Six Sigma) republished per city. |
| Toturon | certification-mill | 1462 | 131 events across 10 cities. |
| KOGNIORA GLOBAL | certification-mill | 433 | 65 events across 21 cities. |
| Certification Planner, LLC | certification-mill | 431 | 39 events across 2 cities. The smallest of the seven and the one nearest the line: the city spread is a venue's, but the titles are the mill's ("CISA Guaranteed to Run ... Exam Prep In-Person Training"). |
| Mangates | certification-mill | 146 | 162 events across 12 cities. |
| Learners Ink | certification-mill | 0 | 179 events across 11 cities. |
| Skelora Edu Tech | certification-mill | 0 | 218 events across 30 cities. |

## What we know is wrong

Defects we know about and have not fixed, each one an open issue you can read. Counts are as of the date beside them, not as of now.

-   [#39](https://github.com/foundingstudio/eventing/issues/39)Events in no metro appear in the default list934 as of 2026-09-01
    
    With no city chosen, the list includes events our region boxes could not place in any metro, sorted in among the ones it could. They are correctly stored and correctly shown as unplaced elsewhere; they should not be in a list headed by a count of metros.
    
-   [#40](https://github.com/foundingstudio/eventing/issues/40)One event cross-posted within a platform used to render as three rows3 as of 2026-09-02
    
    Three groups on one platform posting the same expo produced three rows at the top of the listing. The read-path collapse now folds them where the coordinates agree, and the issue stays open until the operator confirms the case that opened it reads as two rows rather than three.
    
-   [#21](https://github.com/foundingstudio/eventing/issues/21)A coordinate that three disagreeing cities claim is a fabrication we do not catch in general
    
    Some feeds hand out one placeholder coordinate for every event in a country. We catch the pairs we have found by hand and the cases where the street address names the city; we do not yet catch the general shape, so some events are pinned to a point nobody is holding an event at.
    
-   [#59](https://github.com/foundingstudio/eventing/issues/59)The offline test gate is red on the main branch1 as of 2026-09-05
    
    One migration file carries a comment naming a platform, which the rule that keeps vendor names out of the neutral layers rejects. The fix changes an applied migration, so it needs a database repair alongside it and has not been taken yet.
    
-   [#18](https://github.com/foundingstudio/eventing/issues/18)Three frontend test suites have never run3 as of 2026-08-25
    
    A shared test helper imports a file that does not exist and never has, so three suites fail to load rather than fail to pass. Nothing they would have covered is currently covered.
    

## Verdict thresholds

The numbers behind the sentence on an event card. They are constants, not a model: no language model runs anywhere in this product.

| Axis | Reads as | When |
| --- | --- | --- |
| Price | Cheaper than usual | 25.0% or less of the metro's priced week is cheaper |
| Price | Pricier than usual | 60.0% or more of it is cheaper |
| Host | Steady host | 10 distinct events and 180 days since we first saw them, both |
| Host | Known host | 3 distinct events or more, below the steady bar |
| Night | Quiet night | 2 or fewer other events within 2 hours either side, same metro |
| Night | Busy night | 12 or more in that window |