State which comparison can isolate a condition
Synthetic trials use the same rows and export action. D-71 is v4 with setting X on and fails; D-72 is v4 with X off and passes. D-73 changes both build to v5 and X to off, so it cannot isolate which change explains the difference from D-71. Even the controlled comparison supports an association in this fixture, not a general causal guarantee across customer environments.
| Trial | Build | Setting X | Outcome |
|---|---|---|---|
| D-71 | v4 | On | Fail |
| D-72 | v4 | Off | Pass |
| D-73 | v5 | Off | Pass |
Choose a controlled pair
- List changed conditions
- Keep other inputs fixed
- Compare a one-change pair
- Limit the resulting explanation
If the action or dataset also differs, record that difference before interpreting the output.
Which pair isolates X within v4?
Show the worked answer
D-71 and D-72. Their build, input and action match and only X differs. D-71 versus D-73 changes two conditions.
Retain all controlled conditions with the observation. Additional environments need their own trials before extending the explanation.