Amego's recommendations engine (the same one behind the "For You" tabs) looks at and uses attendee activity, session content, and attendee matching to surface sessions, speakers, sponsors, and connections attendees are likely to care about. It doesn’t require the organizer to set anything up by hand, but instead runs in the background from the moment Enable Recommendations is turned on.
We learn what attendees care about by building a picture of each attendee's interests by observing what they do in the app. Favoriting a session, adding it to their agenda, and checking in, which is the most important step, all send a signal. Someone who checks into three sessions about analytics is treated as more interested in analytics than someone who only favorited one session and never attended. These recommendations remove the requirement for surveys and questionnaires, and instead provide an attendee profile that gets sharper the more someone uses the app, so recommendations consistently improve the more the attendee interacts with the app and the event.
Recommendations use AI to automatically figure out session content. It reads the agenda (title, description, and other session details) and groups sessions into topics on its own, eliminating the need for organizers to tag everything by hand. A session titled "Scaling Your Sales Pipeline with AI" might get grouped under topics like sales strategy and applied AI without anyone touching a tagging field. Sessions are recommended by comparing an attendee's topics against each session's topics and rank sessions by how strongly they overlap. The more topics a session shares with an attendee's profile, the higher it ranks in their For You list. While reviewing these sessions, specific sessions are filtered out, such as those already on your agenda, finished sessions, and any sessions that aren't visible to that attendee's type.
Speaker and sponsor recommendations are integrated into session recommendations rather than being calculated separately. For example, a speaker is recommended because the sessions they're presenting match the attendee, and the same rule applies for sponsors. If a speaker or a sponsor is tied to several matching sessions, they rank higher. For example, a keynote speaker presenting three sessions that all match an attendee will outrank one presenting a single matching session.
For networking recommendations, we combine three signals into a single match score: shared interests, mutual connections, and belonging to the same company or email domain. Two attendees from the same company who also share several session interests will score higher than two strangers with only one thing in common. Only attendees who've opted into networking show up as suggestions, and any blocks between attendees are always respected, so a blocked attendee never appears in someone's networking recommendations, no matter how strong the match would potentially be.
Organizers stay in control of how recommendations behave for their event. They can hide specific content from recommendations entirely, exclude certain attendee groups from networking, and set rules such as showing buyers to sellers first, so the highest value connections surface at the top.
Recommendations still work even when content is limited. For a new event, or an attendee who hasn't used the app much, we fall back to showing what's popular or what's coming up next, based on activity from everyone else at the event. Sidekick is truthful about presenting a fallback list and recognizes that the suggestions aren't personalized.