Monitoring, Review and Continuous Improvement
Measure, compare, correct, learn
This domain closes the loop. It asks whether the work is on track, whether the outcome is being realised, and whether the organisation is learning. Three separate questions with three separate instruments.
Concepts to hold
- Metric versus KPI. All KPIs are metrics; only those tied to a strategic objective are key. A dashboard of forty metrics usually has no KPIs.
- Leading and lagging. Leading indicators are actionable but noisy; lagging are reliable but late. Pair them.
- Baseline and variance. Variance % = (actual − baseline) ÷ baseline. State the sign convention; otherwise a favourable variance reads as a problem.
- Tolerances. Monitoring triggers action only outside agreed tolerance; inside tolerance, report and continue. Reacting to every fluctuation is over-correction.
- Benefits realisation. Measured post-handover against the business case, with a named owner and a review date. Otherwise nobody owns the outcome.
- Review types. Lessons learned look backward at what happened; retrospectives look at how the team worked; both feed the next cycle via PDCA.
Traps
Vanity metrics that always rise (page views, tickets closed) without linking to value. Goodhart's law: once a measure becomes a target, it stops being a good measure. Treating a review as a status meeting rather than a decision point. Closing a project without confirming benefit ownership.
Rule of thumb: the improvement loop
Measure against baseline, investigate variance outside tolerance, correct the cause (not the symptom), then re-baseline deliberately and record the lesson. Repeat.
Worked illustration
Baseline processing time 40 hours, actual 46. Variance = (46 − 40) ÷ 40 = +15%. Tolerance is ±10%, so the breach triggers root cause analysis. If tolerance were ±20%, the correct action is to record and continue, not to launch a corrective project.