Operations

A run records the operation key, source revision, start and completion times, input hash and result reference. Queued, Running, Succeeded, Failed and reconciliation required tell the team whether a collection or comparison reached a dependable conclusion. A run marked Succeeded is meaningful only when the workflow has confirmed the intended effects. For cohort comparison, that means reading back the updated result fields.

Numbered steps provide a closer view. Each step names its action, input and output references, timing and status. When an external receipt exists, its identifier, source, date and evidence can be kept with the step. This helps an Analyst distinguish an operation that stopped before writing from one whose external action is uncertain.

• Run identity. A run number and operation key distinguish one attempt from another.
• Step sequence. The numbered actions show where an input became an output or stopped.
• Source history. A linked property and state history place the run in its lifecycle.

Agree which operation key describes collection and which describes cohort comparison. Keep the run number unique enough for a person to use in a handoff. Decide what external receipt evidence is required when an action leaves the app. The failure reason should tell an Analyst where to resume rather than merely repeat the word Failed.

A failed or uncertain step deserves investigation before a second attempt. The input hash and result reference help identify whether the second request repeats accepted work. A reconciliation required run makes the uncertainty visible. The team can then inspect the original source and output before deciding whether another action is safe.

An Analyst starts collection for a property after confirming consent. The run records the source revision and its numbered steps. One step reports Unknown because its receipt is not yet available. The Analyst opens that step, checks the external source and receipt evidence, and leaves the run as an exception until the result is understood. A second collection without this check could create another fact for the same event.

For a cohort comparison, the Analyst checks the resulting session, conversion, goal and exposure values on the property. Only after the read-back confirms those effects should the operation be treated as Succeeded. The Analytics Lead can use the same run to explain a changed overview trend to the Reviewer.

• Run Exceptions — Investigate failed and reconciliation-needed operations by property.
• Property Lifecycle — Place the run against the current source state.

Follow collect sessions for the first run and compare cohorts for the verified result-field update.