Experiments

An experiment variant has a name, code, organization and effective dates. It can be enabled or disabled. A measured property points to its definition, so the team can understand which variant context applied to the session facts and observed conversion values. That relationship is especially useful when a campaign changes its treatment while the property remains the same.

Source mappings describe how an incoming field relates to a target field, including type, transform rule and effective date. Metric definitions supply a grain, numerator and denominator expressions, unit, aggregation method and version. Their Draft, Approved and Retired states help the team distinguish a current definition from work still under review or an older measure retained for explanation.

• Variant record. The code and effective window identify the configuration used by a property.
• Field mapping. Source and target fields explain what enters the measurement model.
• Metric definition. Grain, expressions and aggregation method describe how a result should be read.

Agree how the organization names variants and when each definition takes effect. A code should remain understandable to the Analyst who revisits an earlier property. Decide who reviews a metric definition before it reaches Approved, and record that reviewer. The team should also specify whether a later mapping version applies only to new collection or requires a separate review of existing property readings.

Do not collapse unlike units into one display label. A count of sessions and a conversion ratio answer different questions. Write the numerator and denominator expressions plainly enough that another reviewer can compare the persisted result with the underlying facts. Choose an aggregation method that matches the unit and the reporting window.

A growth team tests a second landing page treatment. Admin creates a variant with its own code and effective window for the organization. The Property Owner assigns the definition to the measured property. The Analyst checks the source mapping before collecting consented facts, then compares the exposure and conversion values for the reviewed property. If a metric definition is used to read a result, the Analytics Lead checks its version and Approved status before interpreting the value.

When a campaign changes the variant after the first analysis, retain the earlier definition and its effective dates. That way a reviewer can tell which treatment the earlier property result described. The mapping and definition records provide context; the actual facts, run history and review decision still determine whether a published outcome is ready.

• Experiment Definitions — Review variant codes, effective dates and organization before assigning a property.
• Property Lifecycle — Find the property that uses a selected definition.

Start with configure measurement, then compare cohorts using the reviewed source.