Understanding Session Recommendations
Sessions in Grip can be recommended, but do not have to be. Sessions not listed as being recommendable will not be recommended. Recommendations will appear in a users agenda, with the ability/option to then add them to your schedule.
Recommendations are produced when specific changes happen, or daily if no relevant changes occur. A maximum of 10 recommendations is made in a day, to stop users from being deluged and thus ignoring recommendations.
What are Recommendations based on
Recommendations are primarily based on expressions of interest. These expressions can come in various forms.
Connections with Speakers
This is relatively rare, but if a person has a connection with a speaker at a session, this is considered a strong expression of interest. Not only are you interested in the speaker, but they are interested in you, and thus likely a first choice attendee. Any sessions they speak at would likely be recommended if possible.
Interest swipe on Speakers
It is more common for a person to express interest in a speaker - since that does not also require the interaction of the speaker. This is also considered an expression of interest and will be used to create recommendations. Sessions they speak at will be recommended to that individual if possible.
AI-based interest in speakers
Perhaps more abstractly, recommendations of sessions are also made if the AI has identified a reason you would be interested in the speaker. Typically it needs more information about you to make this decision to the certainty needed here, so it is rarer. However, some recommendations are created based on this expected interest.
Content-Based:
If you provide a list of topic subjects a person is interested in, that is another expression of interest. This is relatively new and currently underused functionality. For now, it only involves matching the explicitly listed topics of interest and the session topics given.
This method uses session tags; descriptions of session topics given to us and stored in our database. Session tags can be used for multiple things, and store various data about the session, however, if using the key 'topic-tags', these are interpreted as content tags for matching.
These 'topic-tags' values will be matched with values in the profile metadata field key 'session_topics_interested', thus for a recommendation to be made there must both be topic tags for the session, and the matching values in the profile's metadata with the key 'session_topics_interested'.
By including both of these key-value pairs into an event and into specific sessions (with exactly matching values), the organiser can gain some control over session recommendations going to this specific set of people. This can be used to send sessions to people interested in a topic (note: only a maximum of the top 10 of these will be sent to an individual each day), or potentially to everyone in a particular group if the values were automatically placed into the metadata by the organiser.
These 'session_topics_interested' are not visible in profile recommendations as reasons to meet, nor are they searchable, filterable or free-form text. They are partially used to help identify similar users when making profile recommendations but are not used directly to make these recommendations.
Speed Recommendations:
Speed sessions are a little different from other types of sessions as they do not have speakers and are instead sets of one:one auto-generated meetings for the people attending the session.
These will be specifically recommended if any of your connections, or people you have expressed interest in have added these sessions to their schedules. That is a good expression of interest as not only is your connection/interest swipe present, but it is likely that there will be people similar to them present as well - making it likely that you will have an interest in others attending this session as well.
Speed sessions will also be recommended preferentially to those who can attend as a form of backfill since experience has shown these sessions can be a great return on time for many session attendees, often producing numerous good meetings & connections in a short interval and often used as an introductory ‘getting to know you’ feature though they are also well matched to particular topics of interest.
Similar People:
We also make recommendations based on the actions of people similar to you. Similar in this context can mean a variety of things.
- Similar metadata values (e.g. services interested, product categories interested, etc).
- Similar sessions added to schedule / attended
- Have shown similar interest preferences - e.g. made similar swipes on other users/products etc.
In all cases, we can use these similar people and by using any sessions they have added to their agendas we can make recommendations for the user of interest.
As a final addition:
We have recently decided to add a final ‘backfill’ to the session recommendations. Previously despite all of the above, some users received no recommendations, while it was also possible for some sessions to not be recommended to anyone. To reduce these occurrences, we now say that if a user has no recommendations they are given a single session as a recommendation (if available), with a preference for sessions rarely or never recommended previously.
FAQ
Are companies/sponsors recommended as well?
Currently no, though there are plans to take sessions & session recommendations more into account for profile recommendations in the near future.
How can we “tag” sessions, so that they can be recommended according to the answers to the onboarding questions
Session tags for recommendations should always have the key 'topic_tags'. The matching metadata in profiles should be
Is there any way to manipulate the session recommendations in such a way as to be made to everyone in a specific group
Theoretically yes - if a specific metadata value e.g. 'key_group_1' was inserted into the 'session_topics_interested' field for a profile and the same was included as a 'topic-tag' for a session, then this session would be preferentially recommended to people in the group.
It is however possible some of them will not receive it - as there are other parts to the recommendations which may lead to them getting full sets, not including those sessions. However, it is likely most if not all members would get these recommendations.