Signal Synthesiser Data

signal-synthesiser-data

Agent: Data Analyst

L2 data analyst (Nx) responsible for data modelling, instrumentation implementation, metrics dashboards, funnel analysis, and signal synthesis.

Department ethos: ideal-data-growth.md

Skill Description

The signal synthesiser combines quantitative signals from product analytics, revenue data, support metrics, and infrastructure logs into a unified product health view that surfaces trends, anomalies, and correlations invisible when each source is reviewed in isolation.

When to Use

• When the weekly or monthly product review requires a consolidated health report spanning multiple data sources.
• When a metric moves unexpectedly and the team needs to correlate it with signals from other systems (support tickets, error rates, campaign launches).
• When leadership requests a single-page view of product health rather than navigating multiple dashboards.
• When siloed teams are making conflicting decisions because they each see only their own data.

Workflow

1. Identify signal sources: List all data sources contributing to product health — analytics platform, data warehouse, support ticketing system, error monitoring, revenue system, NPS/CSAT surveys.
2. Extract key signals: For each source, pull the 3-5 most important metrics for the reporting period. Normalize time windows and cohort definitions across sources.
3. Detect anomalies: Flag metrics that deviate more than 2 standard deviations from their trailing 30-day average. Cross-reference anomalies across sources to identify correlated movements.
4. Synthesise narrative: Write a narrative connecting the signals into a coherent story. Distinguish between correlated movements (likely same root cause) and coincidental timing.
5. Highlight action items: Extract 2-3 actionable insights from the synthesis. Each insight should identify the signal, the hypothesis, and the recommended next step.
6. Distribute report: Publish the synthesis report to stakeholders via the agreed channel (Slack, email, wiki) at the defined cadence.

Anti-Patterns

• Dashboard forwarding: Sending a link to five dashboards and calling it "signal synthesis" forces the reader to do the synthesis work. Why: the value of this skill is the narrative connecting disparate signals, not the raw data.
• Correlation without caveats: Stating that two metrics moved together without noting that correlation does not imply causation misleads decision-makers. Why: spurious correlations lead to wasted experiments targeting non-causal relationships.
• Stale cadence: Producing a weekly synthesis when the business reviews data monthly creates reports no one reads. Why: the synthesis cadence must match the decision cadence.

Output

Success:

• A product health synthesis report combining signals from 3+ data sources, with anomaly flags, cross-source correlations, a narrative summary, and 2-3 actionable insights with recommended next steps.

Failure:

• A data source is unavailable or delayed, producing an incomplete synthesis. Report the missing source, the impact on the narrative, and when the complete synthesis will be available.

Related Skills

• signal-synthesiser-data-p -- the post-launch variant focuses on a specific release window; this skill provides ongoing health synthesis.
• metrics-dashboard-builder -- dashboards provide the raw metrics that this skill synthesises into narrative.
• funnel-analyser -- funnel analysis is one of the key signal sources feeding into synthesis.