Class TwoArmExperiment

java.lang.Object
ai.promoted.delivery.client.TwoArmExperiment

public class TwoArmExperiment extends Object
Represents a two arm Experiment configuration. WARNING - while ramping up an experiment, do not change numControlBuckets or numTreatmentBuckets. This will likely produce bad results. For the buckets, the treatment buckets are after the control buckets.
  • Constructor Details

    • TwoArmExperiment

      public TwoArmExperiment(String cohortId, int numActiveControlBuckets, int numControlBuckets, int numActiveTreatmentBuckets, int numTreatmentBuckets)
      Creates a two-arm experiment config with the given parameters.
      Parameters:
      cohortId - the cohort (experiment) name
      numActiveControlBuckets - number of active control buckets
      numControlBuckets - total number of control buckets
      numActiveTreatmentBuckets - number of active treatment buckets
      numTreatmentBuckets - total number of treatment buckets
  • Method Details

    • create5050TwoArmExperimentConfig

      public static TwoArmExperiment create5050TwoArmExperimentConfig(String cohortId, int controlPercent, int treatmentPercent)
      Factory method for a 50/50 experiment.
      Parameters:
      cohortId - name of the cohort id (experiment)
      controlPercent - percent of the total to activate into control, range [0, 50]
      treatmentPercent - percent of the total to activate into treatment, range [0, 50]
      Returns:
      the configured experiment
    • getCohortId

      public String getCohortId()
    • getNumActiveControlBuckets

      public int getNumActiveControlBuckets()
    • getNumControlBuckets

      public int getNumControlBuckets()
    • getNumActiveTreatmentBuckets

      public int getNumActiveTreatmentBuckets()
    • getNumTreatmentBuckets

      public int getNumTreatmentBuckets()
    • checkMembership

      public CohortMembership checkMembership(String userId)
      Evaluates the experiment membership for a given user.
      Parameters:
      userId - the user id to check
      Returns:
      the correct cohort membership if the user is activated into the experiment, or null if not.