Track workflow quality with tension-paired metrics…
Track workflow quality using multiple metrics that create tension with each other—at minimum cycle time, error rate, and energy cost—to prevent optimization of any single metric from corrupting the system.
When This Fires
- When setting up measurement for a workflow you want to improve over time
- When optimizing a workflow and noticing that one metric improves while others degrade
- When a workflow feels "optimized" but produces burnout or quality problems
- Complements Measure actual elapsed time over 3-4 cycles before… (measure before optimizing) with guidance on what to measure
Common Failure Mode
Speed-only optimization: "We reduced cycle time by 40%!" — but error rate doubled and the executor is exhausted. The cycle time metric looks great in isolation, but the workflow is worse overall. Without counter-metrics, this degradation is invisible in the measurement system.
The Protocol
(1) For each workflow, track at minimum three metrics: Cycle time (wall-clock start to finish), Error rate (defects, revisions, or rework per execution), and Energy cost (subjective 1-5 rating of how depleting the execution was). (2) After each improvement, check all three metrics. A genuine improvement either improves one without degrading others, or improves two while slightly degrading the third. (3) Reject changes that dramatically improve one metric while degrading another — these are metric games, not real improvements. (4) If you must prioritize, use context: cycle time matters most under deadline pressure, error rate matters most for high-stakes outputs, energy cost matters most for daily-frequency workflows. (5) Add domain-specific counter-metrics as needed: output quality score, handoff satisfaction, learning rate. The more metrics that must be satisfied simultaneously, the harder it is to game the system.