Running a CMAB experiment
Once the experiment is set up, you have a unique experiment_id
assigned to it.
You can now use this experiment ID to randomly choose an arm or variant to present to each new end-user by calling the draw_arm
endpoint.
This works similarly to the draw_arm
endpoint for MABs, but crucially require the context to be input in addition to the experiment_id
. You can also optionally input a unique draw_id
, or the system will autogenerate one.
You can then use this ID to record the outcome of presenting the arm you drew to the user, using the update_arm
endpoint, similarly to how it works with MABs.
We're currently working on visuals for an experiment card corresponding to CMABs and on a dashboard to display more detailed information across experiments.