Prompt Cards

From the past 48 hours of breaking news, separate 'signal' (persistent stories) from 'noise' (transient hype). Explain why some stories fade while others persist, and estimate which signals will grow stronger.

Gather today's top 20 global news items from multiple sources (Perplexity, Reuters, Bloomberg, NHK) and build a matrix of topics, frequency, and sentiment. Identify which 3 trends are accelerating fastest and predict their likely outcome within 30 days.

Goal: deploy real-time anomaly detection and alerting. Data: streaming events and KPIs. Steps: 1) Forecast baselines with ETS/Prophet; 2) Residual monitoring with robust z-scores; 3) Alert playbooks; 4) Postmortem template. Output: alert catalog and on-call runbook.

Goal: stitch cross-device journeys. Data: login IDs, hashed emails, device graph. Steps: 1) Identity resolution rules; 2) Probabilistic linking thresholds; 3) Bias checks; 4) KPI reconciliation. Output: unified journey map and data contract.

Goal: track feature adoption lifecycle. Data: feature flags + events. Steps: 1) Adoption curve by cohort; 2) Time-to-first-use and repeat; 3) Impact on retention/LTV; 4) Sunset or double-down decisions. Output: adoption dashboard and product roadmap inputs.

Goal: compare mobile vs desktop behavior. Data: GA4 device category segmentation. Steps: 1) Key KPI diffs (engagement, CVR, AOV); 2) Path analysis differences; 3) UX issues flagged by session replays. Output: prioritized mobile UX backlog.

Goal: analyze pricing page sensitivity and elasticity. Data: pricing page events, trials, conversions, order values. Steps: 1) Build exposure cohorts by price/test; 2) Demand curve fit; 3) Elasticity by segment; 4) Recommend packaging experiments. Output: elasticity curves and pricing test plan.

Goal: detect robots/scrapers from logs. Data: server logs (CDN/WAF), GA4 anomalies. Steps: 1) UA/IP reputation, JA3/TLS fingerprints; 2) Request-rate and path entropy anomalies; 3) Honeypot endpoints; 4) Impact on KPIs. Output: filter rules and mitigation recommendations.

Goal: design segmentation and CDP activation. Data: events, traits, CRM. Steps: 1) Build segments (recency, behavior, value); 2) Define activation triggers; 3) Map to channels via CDP (e.g., Segment); 4) Measure incremental lift with holdouts. Output: segment catalog and activation plan.

Goal: quantify Core Web Vitals impact on conversion. Data: CrUX or RUM data for LCP/INP/CLS + conversions. Steps: 1) Bin users by CWV percentiles; 2) Model conversion vs CWV with semi-parametric fit; 3) Estimate ROI from improving to 'Good'. Output: CWV-improvement business case and roadmap.

Goal: generate CRO hypotheses backed by evidence. Data: funnels, heatmaps, surveys, NPS, session replays. Steps: 1) Aggregate pain points; 2) Map to heuristics (clarity, friction, motivation); 3) Prioritize via PIE/ICE; 4) Draft A/B specs. Output: ranked hypothesis bank with expected lift.

Goal: predict churn propensity and design save flows. Data: user activity, support tickets, billing events. Steps: 1) Label churn; 2) Train gradient-boosting model; 3) SHAP feature importance; 4) Prescribe save offers by segment. Output: playbook + targets for triggered interventions.