Experimentation storage allows for shadow deployments on any change to run safely and produce data in an isolation folder/container. It then combines the shadow deployed data (experiment group) with production data (control group) and makes setting up and running experiments easy and straight forward.
Experiments are defined in yaml file like the one in the example below.
- hypothesis: The newly added features improve the model, it results in higher user satisfaction which can be seen with more daily active users and longer sessions.
- table: /mnt/silver/cool-table/data
- scope:
- start: 03/11/2024
- end: 04/05/2025
- filter: id > 1000 and id < 2000
- evaluation:
- table: /mnt/silver/cool_metrics
- metrics: dau, logins, avg_session_duration
Like other Storage objects ExperimentationStorage can be turned on with spark config.
Experimentation storage allows for shadow deployments on any change to run safely and produce data in an isolation folder/container. It then combines the shadow deployed data (experiment group) with production data (control group) and makes setting up and running experiments easy and straight forward.
Experiments are defined in yaml file like the one in the example below.
Like other Storage objects ExperimentationStorage can be turned on with spark config.