An operating error can be corrected quickly and still leave a cost. Correction takes time, materials or a new contact with the customer. If this work is not recorded, the company can underestimate the necessary resources and repeat the same problem.
I propose to detect the type of error and the point at which it becomes visible. The distinction helps to understand whether an early verification would be useful. Recording events does not mean automatically attributing faults: incomplete information and ambiguous steps can affect people's behavior.
A hypothetical example concerns orders prepared with an outdated version of the request. The fix absorbs hours of stock, but the cause may be in the way a change is communicated. To intervene only on the preparation leaves intact the initial problem.
The review should choose a few recurring causes and check the effects of the change. It also counts the cost of the introduced control. A process improves when it prevents major errors with proportionate verifications, keeping work understandable and without adding unnecessary steps to each operation.