| name | signal-vs-noise-filter |
| description | Use when a situation contains many facts, feelings, events, rumors, or weak cues and the real signal must be separated from noise. Use when applying signal detection, signal-versus-noise filtering, pattern relevance review, or decision-signal extraction. |
| license | Apache-2.0 |
Signal vs Noise Filter
Use this skill to identify which observations should change the diagnosis or action plan.
Method Notes
- A signal is information that changes the likely explanation, priority, risk, timing, or next action.
- Noise may be true but not decision-relevant.
- The method works best after an initial evidence map or fact clarification.
Required Inputs
Collect or infer these inputs before execution:
- decision question or action choice
- list of observations, events, facts, cues, and concerns
- baseline expectation before the event
- timing of key changes
- available evidence strength
If an input is missing, mark it as missing, state the assumption used, and add a validation action.
When Not To Use
Do not use to repair unreliable evidence or fill basic situation gaps. Use evidence-map when facts, claims, sources, and confidence are mixed; use 5w1h-analysis when who, what, when, where, why, or how is missing. Use this method only after the input is good enough to judge decision relevance.
Step-by-Step Execution
| Step | Required input | How to execute | Output |
|---|
| State the baseline | Prior expectation, normal pattern, status quo. | Clarify what would have happened if nothing material changed. | Baseline expectation. |
| List candidate signals | Observations, events, user concerns, timing. | Convert each item into a candidate signal. Remove duplicates. | Candidate signal list. |
| Rate decision relevance | Decision question, action options. | Ask whether this item changes explanation, priority, risk, timing, stakeholder stance, or validation need. | Relevance rating. |
| Rate evidence strength | Source, directness, recency, corroboration. | Mark each candidate as strong, medium, weak, or missing. | Evidence-strength rating. |
| Separate noise | Low-relevance or low-evidence items. | Keep true-but-not-actionable items in a noise list so they do not dominate the plan. | Noise list. |
| Name the core signal | High-relevance and sufficiently evidenced items. | Select the 1-3 signals most likely to change action. | Core signal diagnosis. |
Output Template
### 1. Baseline
- What we expected before the change:
- What changed:
- Why this matters:
### 2. Candidate Signal Table
| Candidate signal | Evidence strength | Decision relevance | Changes what? | Signal / noise / watchlist |
|---|---|---|---|---|
| | strong / medium / weak / missing | high / medium / low | explanation / priority / risk / timing / stakeholder / validation | |
### 3. Core Signals
1.
2.
3.
### 4. Noise and Watchlist
- Noise to avoid overreacting to:
- Watchlist item that needs more evidence:
### 5. Action Implication
- What the signal means:
- What it does not prove:
- Next validation:
Quality Gate
- A signal must change a decision, not only feel interesting.
- Do not downgrade a real signal because it is politically uncomfortable.
- Do not upgrade a weak cue because it fits the preferred story.
- Name what the signal does not prove.
- Keep the final signal list short enough to guide action.