Records and Metrics
You may think records are a report card. They are not. A record is material for following a single execution. Whether it went well as a business matter cannot be told from the record alone.
This page explains how to read an execution record and what the numbers do not represent. The procedure for reading is in the Record Guide.
An execution record is material for following one execution
The execution record screen lets you follow one execution from its start to its end under five headings: “Request”, “Evidence”, “What was done”, “Confirmation”, and “Result”. The request shows what started it (a sweep, a webhook, the external API, a manual start, and so on).
An execution record holds the details of one execution. You can follow which registered agent ran with which settings. An execution record shows the “execution environment image”: the container image of the registered agent that ran. For workflows it shows the image digest the administrator certified.
While an execution is running, the “Request” section of its execution record shows the elapsed time and the number of operations (“N operations”). A breakdown of operations via connectors is not shown.
You can also follow a workflow execution from the case screen. See the versions of settings and artifacts on the case screen (Case Guide).
A result shows how the execution ended
What the result field shows is how that execution ended. It is not whether the business goal was achieved.
“Completed” means the agent finished it as completed (it did not declare “could not complete”, and there was no refusal by a run rule, no overrun of the execution limit, and no abnormal exit). Check by its contents whether the artifact is what you intended.
“Run ended (completion not measured)” is neither success nor failure
Even when a cycle's registered agent finishes processing and exits normally, the platform cannot judge whether it did its job. The result of such an execution is “Run ended (completion not measured)”. It counts as neither success nor failure and is not part of the denominator of metrics such as the completion rate. If the agent is built to report whether it completed, it is recorded with the usual results (Record Guide).
Metrics are an aggregate of execution records
Metrics are counts of records. So the cautions for reading records show up directly in the numbers. What counts in the denominator is the executions of the registered agents placed on steps in a workflow (excluding review steps and steps passed through), and, in a cycle, the executions of registered agents (both periodic sweeps and runs started by webhook). “Run ended (completion not measured)” is not in the denominator.
Even when the completion rate is high, you cannot say the business went well. “Completed” only means the agent finished it as completed.
Judge business results by records and artifacts
All of this comes together in one point. An execution record is not a report card. If you want to know whether it went well, looking at the numbers is not enough. Judge by the execution record together with the actual artifacts. The record answers only “what happened”; deciding “was that good enough” is up to you.
What to read next
- Record Guide — How to read execution history, action needed, and reasons it could not launch
- Human Review — The mechanism behind review figures