Module 6 — Metrics That Matter · Lesson 6.1
A Framework for Measuring Management Quality
What is worth measuring, what is worth ignoring, and how to tell which is which
~11 min
What you'll learn
- Distinguish input, throughput, and outcome metrics
- Recognize the failure mode of each layer when over- or under-weighted
- Use the framework to evaluate any metric a stakeholder asks for
- Set up the four-metric per-layer dashboard the rest of the module fills in
Every management book published in the last twenty years includes a chapter on metrics. Most of those chapters are wrong in the same way: they tell you what to measure without telling you how to think about whether a measurement is worth the cost of collecting it. This lesson is the framework that the rest of the module fills in.
Three layers, in increasing distance from the work
Input metrics are about what is going into the system. Skill declarations, charter completeness, hours of attention spent. They are the easiest to measure because they are observable directly from the team's behavior, and they are the most game-able because the team can produce them without producing any real outcome.
Throughput metrics are about how the system is moving. Tasks closed per week, cycle time, estimate accuracy, dismissal rate. They are harder to measure because they require the system to be running, but they are also more honest — you cannot produce healthy throughput numbers without actually moving work.
Outcome metrics are about whether the work mattered. Customer retention, revenue per seat, NPS, the project's stated V being met. They are the hardest to measure because outcomes lag and are influenced by factors outside the team's control, but they are the only metrics that, in the end, matter.
All three layers have a place. The pathology is over-weighting one and under-weighting the others.
The three failure modes
Input-only management is the most common form of bureaucratic decay. The team tracks compliance metrics — how many forms got filled, how many trainings got attended, how many charter fields got populated — and confuses the production of inputs with the production of value. The visible artifact is a healthy dashboard and a slowly dying business. Government agencies excel at this.
Throughput-only management is the operator's failure mode. The team obsesses over cycle time, on-time delivery, and task throughput, and forgets to check whether the work being delivered actually moves a customer outcome. This is the failure mode of teams that have great quarterly metrics and quietly losing market share.
Outcome-only management is the founder's failure mode. The team only looks at revenue and growth, and is blind to the input and throughput signals that would predict whether the outcomes are going to continue. By the time the outcome metrics turn, the underlying causes have been compounding for two quarters.
A healthy management dashboard has a thin layer of each. Two or three input metrics that are explicitly tied to a throughput consequence. Three or four throughput metrics that are explicitly tied to an outcome. Two or three outcome metrics that are the things you are actually trying to produce.
The 'so what' test
Before adding any metric to the dashboard, apply the so-what test: if this number changes, what changes about how the team operates?
A metric that passes the test produces a concrete action when it moves. Capture coverage drops? You add another integration. Estimate band widens? You retag and tighten descriptions. Negation hits drop on a specific N? You investigate whether it is irrelevant or invisible.
A metric that fails the test is decoration. It might be true, it might be informative, it might even be interesting — but it does not produce action. Decoration metrics dilute the dashboard, make the real signals harder to spot, and eventually train the team to ignore the dashboard entirely.
The ruthless cut is: every metric you cannot answer 'so what' for, remove. The remaining metrics — usually fewer than ten — are the dashboard.
The shape of the rest of this module
The next three lessons walk the three classes of metrics for a project-managed-in-Kavanah team.
People metrics (6.2). Whether the team's capability model is accurate, whether assignment is producing growth, whether load is sustainable.
Work metrics (6.3). Whether the work is moving, whether the estimator is calibrated, whether the KVN is doing its job.
Operating metrics (6.4). Whether the system as a whole is producing outcomes — capture latency, AI Employee contribution, customer-facing outcome signals.
Each lesson lists four to six metrics per class with definitions, healthy ranges, and the action each is meant to produce. Together they form the dashboard. You will not use all of them; you will use the ones that match your team's stage and stakes. The framework tells you how to choose.
Audit your existing dashboard
- 1
List the metrics currently on the team's dashboard.
- 2
Tag each as input, throughput, or outcome
If the dashboard is all input, you are managing compliance. All throughput, you are managing tempo. All outcome, you are managing in the rearview.
- 3
Apply the so-what test to each metric
If a number changes, what changes about how the team operates? Cut decoration metrics aggressively.
- 4
Aim for thin coverage across all three layers
Two or three of each. The remaining six to nine become the dashboard.
Meta-metrics — about the dashboard itself
- Dashboard layer balance
- Distribution of metrics across input/throughput/outcome.
- Healthy signal: Each layer has at least one and at most five metrics. Skew toward any one layer is a failure mode.
- So-what coverage
- Fraction of metrics for which the team can name the concrete action a change would produce.
- Healthy signal: 100%. Anything below means there is decoration on the board.
- Dashboard read rate
- How often the team's leaders open /reports per week.
- Healthy signal: Multiple times per week. Dashboards that are not read are not influencing decisions.
- Metric review cadence
- Whether the dashboard's metrics are revisited each quarter.
- Healthy signal: Quarterly. Dashboards drift; the right metric this quarter is rarely the right metric next year.
Key takeaways
- ·Three layers: input, throughput, outcome.
- ·Each layer has a characteristic failure mode when over-weighted.
- ·The 'so what' test: every metric needs a concrete action it would produce.
- ·Aim for thin coverage across all three layers.
With the framework set, the next three lessons walk the actual metrics for each layer in turn — starting with the people side.