When to Raise a Signal — and When to Act on It
The signal framework has two distinct moments that are easy to conflate: the moment a signal is raised, and the moment it demands a capital response. Conflating them produces one of two failure modes — raising signals too late because the threshold has not been confirmed, or waiting for capital authority before starting the diagnostic conversation.
The threshold in the Signal Response Summary governs when the quarterly cadence is obligated to fund a response. It does not govern when the signal is first raised. A signal enters the monthly cadence the moment it appears — not after two quarters of confirmation, not after the cause has been diagnosed. The monthly cadence is where the diagnosis happens. It cannot diagnose what it has not yet seen.
The Marginal Cost Signal — A Worked Example
Marginal cost is the signal most often raised too late, because the two-quarter confirmation threshold is misread as a two-quarter waiting period. It is not.
Week 4 — first appearance in weekly cadence data: Marginal cost has risen noticeably in the most recent sprint cycle. One data point is not a signal — it is an observation. The CTO notes it in the weekly cadence record and watches for recurrence. No escalation.
Month 1 — first monthly cadence: The weekly cadence data from the past four weeks shows a consistent marginal cost increase across two teams. The monthly cadence asks the diagnostic question: is this structural or local? Local causes — a temporarily complex sprint, a one-off integration, a team that lost a senior engineer — produce a spike that resolves itself. Structural causes — technical debt accumulation, platform fragmentation, rising integration complexity — produce a trend that does not. At month 1, the cause may not yet be clear. The monthly output is: signal noted, cause under investigation, watch next month.
Month 2 — second monthly cadence: The increase has continued. The monthly cadence now has two data points and can begin to distinguish trend from variation. If the cause has been identified as local and is resolving, the signal is downgraded. If the cause is structural or still unclear, the signal is confirmed as a pattern. The monthly output is: structural pattern confirmed, capital implication being assessed.
Month 3 — third monthly cadence, preparing for quarterly escalation: The cause is confirmed structural. The capital implication is specific: this architectural fragmentation requires X weeks of Y engineers at Z cost to address, and the cost of not addressing it over four quarters is estimated at W. This is the package that enters the quarterly cadence — not a concern, but a decision ready to be funded.
What the quarterly cadence receives: A diagnosed structural problem with a specific capital implication and a cost-of-delay estimate. Not a fresh observation. Not an undiagnosed concern.
What the quarterly cadence never receives: A signal that was not visible in at least one prior monthly cadence. If the quarterly is the first time a signal is raised, the monthly cadence has failed — either the signal was not being tracked, or it was tracked and not escalated.
The General Principle Across All Six Signals
Each signal has a different escalation pace — but the same two-moment structure applies to all of them.
Delivery velocity decline is the fastest-moving signal and the most visible operationally. A single-quarter decline above 10% enters the monthly cadence immediately — not because it is confirmed structural, but because it requires investigation before a second quarter of decline locks in a trend. The weekly cadence observes. The monthly cadence diagnoses. The quarterly cadence funds the response if the cause is structural.
Technical debt ratio is slower-moving and more predictable. It rarely spikes — it accumulates. The monthly cadence tracks its direction. Once it crosses 25% of engineering capacity, the monthly cadence is obligated to diagnose the rate of change and project the trajectory. The quarterly cadence receives it when the projection indicates a delivery velocity impact within two quarters.
Platform reuse rate moves slowly but compounds silently. A single quarter of decline may reflect a specific team’s decision rather than a systemic pattern. Two consecutive quarters of decline indicates something structural. The monthly cadence raises it after the first quarter of decline — not to trigger a capital response, but to begin the diagnostic conversation about whether the cause is platform capability, governance enforcement, or team behavior. The quarterly cadence receives it when the cause is confirmed and the response requires investment.
Integration complexity index escalates faster than marginal cost because dependency accumulation compounds operational friction before delivery economics visibly deteriorate. By the time integration complexity appears in cost or velocity metrics, the structural cause is already several quarters old. The monthly cadence raises it the first time it moves materially. Three consecutive months of sustained rise — not quarterly cycles — indicate a structural architecture problem requiring escalation, because waiting for quarterly confirmation means waiting for the downstream cost signals to confirm what the complexity index already showed.
Marginal cost trend — as detailed above — enters the monthly cadence at first appearance and reaches the quarterly cadence when the cause is confirmed structural and the capital implication is specific.
AI Performance moves faster than the other signals and surfaces structural conditions sooner — which is precisely why it must not be the last signal the organization reads. Poor AI performance appears in weeks, not quarters. It enters the weekly cadence the moment AI outputs become unreliable, unused, or stop improving in production. The monthly cadence diagnoses whether the cause is the model, the data, the workflow, or the feedback loop — following the diagnostic sequence in the AI Performance chapter, in order. The quarterly cadence receives it only when the cause is confirmed structural and the capital implication is specific: data foundation remediation, workflow redesign, platform scaling, or feedback-loop investment. The discipline here is inverted from the other signals. The risk is not raising it too late. The risk is misdiagnosing a system-readiness problem as a model problem, and funding model improvement that cannot produce return because the system around the model is not ready to support it.
The Cost of Waiting
The most common version of this failure is the CTO who observes rising marginal cost for two or three quarters, notes it in the weekly cadence records, but does not raise it in the monthly cadence because the threshold has not been confirmed. By the time the quarterly cadence receives it, the structural cost has already compounded significantly.
The two-quarter threshold is not a waiting room. It is the boundary at which the quarterly cadence is obligated to fund a response whether or not it is convenient. The monthly cadence conversation starts earlier — at the moment the signal appears — precisely so that the quarterly cadence is never receiving a surprise.
A CTO who raises a structural signal for the first time in a quarterly session has been observing it for months in the weekly cadence without acting. That is not patience. It is the signal being absorbed into the operating rhythm rather than acted on.
The discipline is not in the threshold. It is in the instinct to raise the signal the moment it appears — before it is confirmed, before the cause is known, before the capital implication is calculated — so that the monthly cadence has the time to do its work before the quarterly cadence is required to make a decision.
Reading the Thresholds
The thresholds above are starting points, not fixed rules. Every organization has a different baseline, and a signal that is alarming in one context may be acceptable in another. What matters is not the absolute number but the direction and duration of change.
Two principles govern threshold interpretation.
A signal that is stable at a level that would concern most organizations may not require immediate action if the system is coherent and improving. A stable technical debt ratio of 30% in an organization actively remediating is a different signal than a stable ratio of 30% in an organization not tracking it.
A signal that appears healthy in isolation may be alarming in combination. A delivery velocity decline of 8% over one quarter is within the range of normal variation. The same decline alongside a rising technical debt ratio, a falling platform reuse rate, and a rising integration complexity index is a structural emergency — four signals pointing in the same direction simultaneously. The individual thresholds do not capture this. Reading Signals Together, in the next section, does.
Signal Escalation Sequence
Most signals follow a predictable escalation path. Understanding the sequence prevents both premature escalation — treating a single week’s observation as a capital decision — and delayed response — treating a confirmed structural pattern as something still requiring more data.
Weekly cadence: Observe and record. Flag anomalies. Note potential causes. Do not diagnose from a single week’s data.
Monthly cadence (after two consecutive signals): Diagnose. Confirm whether the signal reflects a structural pattern or a local variation. Determine whether a capital response is required.
Quarterly cadence (after monthly confirmation): Fund the response. Allocate engineering capacity. Set the remediation timeline. Make the decision.
Annual cadence: Assess whether the signals of the past year indicate a system that is becoming more or less coherent over time. Not the individual signals — the pattern of how the system responds to them.
A signal that has not been acted on within two quarterly cycles after monthly confirmation has become an accepted condition. At that point, it is no longer a signal. It is the operating model.