A recurring four-step Workflow that runs each time a new Event Insight Report is uploaded. It produces a standardized health check by analyzing event volume trends against prior periods, reconstructing the conversion funnel with stage-by-stage drop-offs, computing engagement depth ratios to reveal…
Consumes the latest Event Insight Report plus all prior reports. Compares event volumes, normalizes by period length, flags >20% changes, identifies new/disappeared events, decomposes changes into user-count vs. frequency drivers. Produces a structured trend report with sections: PERIOD METADATA, VOLUME CHANGES, NEW EVENTS, DISAPPEARED EVENTS, PLATFORM SHIFTS, CHANGE DECOMPOSITION.
Consumes the same Event Insight Reports plus Agent 1's structured trend report. Maps credit and browsing events into a conversion funnel (Visit → Browse → Interest → Action). Calculates conversion rates between stages per period, identifies biggest leakage points. Produces a structured funnel analysis with sections: FUNNEL PER PERIOD, STEEPEST DROP-OFF, CROSS-PERIOD COMPARISON, KEY FINDINGS.
Consumes the same Event Insight Reports. Computes events-per-user ratios for every event type. Classifies into engagement tiers: one-time (~1.0), occasional (1.5–3.0×), habitual (>3.0×). Flags extreme concentrations. Produces a structured engagement report with sections: ENGAGEMENT RATIOS, TIER CLASSIFICATION, TIER SHIFTS, EXTREME CONCENTRATIONS, OVERALL TREND.
Consumes structured outputs from Agents 1, 2, and 3. Consolidates all three analyses into a single executive health report. Does not add independent analysis — strictly consolidates, prioritizes, and formats upstream findings. Produces: TOP-LINE METRICS, EVENTS REQUIRING ATTENTION, FUNNEL STATUS, ENGAGEMENT SHIFTS, PRIORITIZED RECOMMENDATIONS, DATA QUALITY NOTES.
A recurring four-step Workflow that runs each time a new Event Insight Report is uploaded. It produces a standardized health check by analyzing event volume trends against prior periods, reconstructing the conversion funnel with stage-by-stage drop-offs, computing engagement depth ratios to reveal sticky behavior versus one-and-done visits, and synthesizing all three perspectives into a single executive-ready report with a prioritized action list. Designed for periodic monitoring rather than ad-hoc investigation.
Runs entirely on what you paste in — no accounts to link, no data access to grant.