Date of slack thread: 7/24/24
Anonymous: Our react application is a chat application with two classes of users (lets call them Type A and Type B).
tore (Statsig): Are there separate identifiers for this? Like B_id and A_id?
Anonymous: Yep
Anonymous: Providers and clients are more concrete than Type A and Type B
tore (Statsig): Im not sure I fully understand yet, but I wouldd probably model this as two custom ids then, and initialize statsig with both/set both on the user object. Your experiment will be based on one of those ID types, so checks will ignore the other ID. So you can set it up like this
{
customIDs: {
bID: <type_b_id>
aID: <type_a_id>
}
}
tore (Statsig): Wait is a user only ever type a or type b, not both?
tore (Statsig): Can you explain in less abstract terms? Its hard to follow
Anonymous: A user is either a provider or a client. A provider joins a video session with a client.
Anonymous: If a client gets put in the experiment group, I want the provider to also be put in that same group
Anonymous: In the case of a client, we would know their associated provider at initialization time. In the case of a provider we would not because a provider can have many clients.
Anonymous: LMK if I can provide additional info/context <@U01RAN2FKJP> thanks!
tore (Statsig): This sounds to me like you’re actually experimenting on a session, and the unit of experimentation should be the session_id, not the client_a or client_b id. Each client would be a part of the same session_id and would get the same experiment results.