We achieve this with recommendations steered by live-updating **pipeline estimates of the expected number of attendees in each diversity category**, based on: 1. funnel metrics, 2. conversion ratios between funnel stages, and 3. AI predictions based on scraped social media profiles. These are translated into specific introduction requests of each attendee. For example, one diversity constraint for the climate conference was that we wanted representation from each of 19 climate-related skillsets. We generated this list by counting the top skillset tags in [Frontier's Carbon removal knowledge gaps database](https://gaps.frontierclimate.com/). For example, for someone with a chemistry background, a request might be: "Do you know any electrochemists that you want to be more like in some way?" We generate a handful of these and include them on the application form. We ask for the minimum number of recommendations that ensure that our virality coefficient is greater than 1. It depends on the conversion ratios between funnel stages for that event, but it's usually 2 or 3. The first version of this was pioneered at [[Califlorence Climate]]: > The other piece of this strategy was to [enumerate our recruiting goals](https://coda.io/d/_dO1l5UMPCzf/Camp-Califlorence-whos-coming_suAsa) and share them before meetings. They were complex: we were curating people that pass the high bar of “I want to be more like this person in some way”, most of whom are plausibly open to coliving in the Bay Area, and who complement the demographic, technical, and [“idea machine”](https://nadia.xyz/idea-machines) diversity of the group. To track progress towards these diversity goals, I included live-updating dashboards that predict the expected number of attendees for each of these categories, given our funnel status and conversion metrics. Lo and behold, this dashboard improved the quality of people’s introductions dramatically. Recommendations even became self-correcting: when we dipped too low on any category, folks would start recommending 75%+ people in that category. > > Over a period of about two months, a snowballing group of 40+ enthusiastic climate builders collectively submitted 300+ individual recommendations. Of those, my EA researched 206 and we invited 152. 92 of them expressed interest in attending, and 61 ultimately attended. > > We aimed for 50% nonwhite and hit 50.8%, and for 50% nonmale and hit 37.7% (we’ll do better next time by splitting out per-group response rates, which varied dramatically). For “[i](https://nadia.xyz/idea-machines)dea machine” diversity, we aimed for 75% operators, 10% funders, and 15% “scene builders”, and hit 76.4%, 13.9%, and 23.6%, respectively. We were looking for folks with 19 different types of deep technical skillsets relevant to climate and got people with 18 of them. Despite us not explicitly seeking them, **54.1% of the attendees were venture-backed founders**. Lastly, 71 of the 152 folks invited were open-minded to coliving in the Bay Area, producing an expected value of 39 housemates. > > This is all the more magical to me because I wasn’t even well-connected in this industry — I only had half a dozen friends in climate a few months ago. But I already knew or was quickly introduced to superconnectors like Helena Merk, Westley Dang, Eugene Kirpichov, Jason Yosinski, Paul Reginato, Tom O’Keefe, and Candice Ammori, and that made all the difference. > > “I was just really shocked by the quality of the people. Uh, in a good way, obviously.” — Jamie Wong, sharing their group’s reflections on the conference at Sunday’s brunch