Workflow Performance Baseline
Definition. The measured performance of a workflow before a change, including throughput, latency, cost and quality, used to determine whether the change improved anything. Without a baseline, a change's effect is a matter of opinion.
How it works. Measure the baseline over a representative period, including peaks. Record throughput, cycle time, cost per item, exception rate and quality. Re-measure after the change over a comparable period, and use a control group where possible. Compare per segment, since a change often helps one segment and harms another.
Example. A team changes the extraction model to improve accuracy. Without a baseline, they report the new model as better because it looks better on samples. With a baseline, the comparison shows accuracy improved 3 points but cost per item rose 22 percent and latency doubled, which breaches the commitment for one document type. The change is adopted only for the document types where the trade-off is favourable.
Use case. Baseline measurement is required for any workflow change, and it is the discipline that prevents changes from being adopted on impression. It is also the basis for a business case.
Warnings. A baseline measured over an unrepresentative period produces misleading comparisons. Never compare before and after without accounting for seasonality. And never report an aggregate improvement without checking per-segment effects.
Limitations. Baselines require measurement effort before the change, which is often skipped under pressure. They also become invalid if other factors change at the same time.