Skip to main content

Entity Properties

Entity Properties

Entity Properties are categorical details about an entity (e.g. a user) in an experiment, which you can use across all experiments to filter or group experiment results in the Explore section. Create these at Data -> Entity Properties.

You can either provide additional detail about an entity that doesn't typically change (e.g. a user's home country), or a property that may change as part of an experiment (e.g. Subscriber Status : True/False). For the latter, you provide a timestamp field which will be used to identify most recent value prior to the user's exposure. This prevents imbalanced groups and biased results from when an experimental treatment impacts the property, for example if it increased the subscription rate.

image

image

Example Data

For property sources, Statsig only needs a user_id and property fields. Property sources can define fixed properties (e.g. a users Country of origin), but can also define dynamic in which case you need to provide a timestamp for Statsig to identify the most recent pre-exposure record.

Column TypeDescriptionFormat/Rules
timestampOptional an identifier of when the property was defined. Required for dynamic propertiesCastable to Timestamp/Date
unit identifierRequired At least one entity to which this metric belongsGenerally a user ID or similar
property columnsRequired Fields which can be used to group by and filter results in exploratory queries

For example, a static property source could just be:

user_idcompany_idcountry
my_user_17503c_22235455US
my_user_18821c_22235455CA

Which could be used to filter and group by any experiment that was exposed one either user_id or company_id.

For a dynamic property, it might look like this:

user_idtimestampcompany_idintent_segmentspend_segment
my_user_175032023-10-10c_22235455high_intenthigh
my_user_175032023-10-11c_22235455high_intenthigh
my_user_175032023-10-12c_22235455mid_intenthigh
my_user_188212023-10-10c_22235455low_intentlow
my_user_188212023-10-11c_22235455low_intentmid
my_user_188212023-10-12c_22235455low_intentmid

The first user in this example has their intent_segment property change on 2023-10-12; based on what the intent_segment was prior to their exposure, they might have different intent_segment values for different experiment analyses.