Your webinar program runs on inherited format decisions nobody remembers making: forty-five minutes, three speakers, slides, Q&A at the end, "we'll send the recording." Attendance says the format is tired; nobody knows which variable to change, so the next webinar changes the topic and keeps the format — the one variable that wasn't the problem.

Your industry runs hundreds of webinar experiments a quarter and publishes the recordings. Panel versus demo versus workshop, cold open versus housekeeping open, twenty minutes versus sixty — the format landscape is out there, testable against the one signal that survives in recordings: what the content does and how audiences engage with it.

deep_search assembles the format corpus: your industry's public webinars, filterable by length and type, with density verdicts flagging which ones respect the audience's time (a webinar classifying low-density is showing you its drop-off curve).

deep_query runs the comparative read: "across these, which formats and structures come up in the well-executed ones — how do they open, pace, and handle interaction?" — answered with citations to the webinars themselves. The findings turn your next webinar from a topic change into a designed experiment: format variables chosen against the industry's published evidence, not against the memory of how it's always been done.