How the scores work

On-Call Readiness Assessment

Method version 1.0.0 · 20 questions · 10 categories

Answer scoring

• Yes — 1.0
• Partly — 0.5
• No — 0.0
• Not applicable — removed from its category’s denominator rather than counted as a failure.

Category weights

Not-applicable normalisation

Within a category, a question marked not applicable is removed from the denominator, so the remaining questions carry the full category weight. If every question in a category is marked not applicable, the whole category is excluded and its weight is redistributed proportionally across the remaining categories. The maximum achievable score is always 100.

Bands

90–100 Resilient

Your system has strong coverage and operating discipline. Focus on drift, evidence, and continuous improvement.

75–89 Strong

The fundamentals are in place, but a small number of gaps could still create risk under pressure.

60–74 Workable

The system can function, but important parts still depend on memory, heroics, or informal coordination.

40–59 Fragile

Multiple failure points may turn a manageable incident into a missed or delayed response.

0–39 Reactive

Start with ownership, coverage, escalation, and one tested delivery path before expanding the program.

Recommendation logic

The three lowest-scoring applicable categories become the 30-day plan. Each produces one action for this week, one artifact to create, one test to run, and one explicit success condition. A category is only reported as a strength when it scores 75% or above, so the result never congratulates a team for a weak area.

Alert Fatigue Score

Method version 1.0.0 · higher score means higher risk

Contributor weights

Volume pressure heuristic

There is no universal safe number of pages, so the curve is published rather than implied. Pages per responder per week map to a contribution between 0 and 1:

• 0–2 pages: low contribution, rising linearly to 0.25
• 2–5 pages: rising from 0.25 to 0.60
• 5–10 pages: rising from 0.60 to 0.95
• Over 10 pages: maximum contribution of 1.00

Overlapping categories

Some inputs describe populations that can overlap — a page can be both a duplicate and a repeat of an unresolved root cause. The true union of two overlapping shares lies between the larger of the two and their sum. The model uses the midpoint of that range for scoring, and the result page reports the full range rather than presenting a single confident number. The same rule produces the flagged-interruption range.

Risk bands

0–19 Low noise risk

The current paging load appears mostly actionable and supported by context. Watch for drift.

20–39 Watch

One or two sources of avoidable interruption are beginning to affect the system.

40–59 Strained

Responders are likely spending meaningful attention on pages that do not improve a decision.

60–79 High risk

Noise, weak context, routing errors, or after-hours burden may be undermining response quality.

80–100 Critical noise

Pause alert expansion. Fix actionability, ownership, and duplicate sources before adding more pages.

What is never asked

The tool asks for aggregate counts only. It does not ask for, and has nowhere to put, raw alert payloads, alert names, integration identifiers, responder names, or a company name.

Escalation Policy Builder

Method version 1.0.0

The builder produces no score. It computes timing and applies validation rules. A step marked sequential fires after the previous step’s wait expires; a step marked parallel fires at the same moment as the step before it and adds nothing to elapsed time. Time to the final target is the sum of the waits that precede it; one full pass adds the final step’s own wait.

The builder raises a critical issue when the first step has no target or the policy has no stop condition, and warnings for a zero-minute wait that is not marked parallel, a full pass longer than the maximum duration you set, the same target appearing in three consecutive steps, and a fallback that resolves to the same target as the first step.

Shared limitations

• Both scores are self-reported. They measure what you believe about your system, which is useful for finding gaps and useless as an audit.
• The weights are a considered starting point, not an empirical result. No benchmark dataset stands behind them, and we do not present one.
• A score cannot see your specific architecture, contractual obligations, team history, or the constraints people are carrying.
• Neither score should be used to rate an individual’s performance. They describe a system.
• When a method version changes, previously exported reports keep showing the version that produced them.