AcmeSaaS
June–July 2026
Decision Log
Autonomous Actions (SENTINEL acted on its own, no human triggered this)
Loading autonomous actions from SENTINEL...
Data Flags
Decision Impact Trace
Trace what happened after a decision: signals, timeline, and pattern analysis
Days of signals before outcome
0
Loading...
Loading...
Loading...
Pearson r measures linear correlation between decision timing and the outcome metric trajectory, computed from real weekly time-series data via scipy. r > 0.8 is a strong pattern signal. Correlation is not causation: see the Causal Analysis tab for the full Bradford Hill breakdown.
Event Timeline ℹChronological sequence of signals from the decision date to the outcome. Orange line = decision date. Hover nodes for full details.
Metric Trajectory: Before vs After the Decision ℹRed solid = Churn Rate (left axis). Blue dashed = NPS Score (right axis). Orange line = decision date. Live data from BigQuery via Fivetran.
NPS (dashed, right axis) and Churn Rate (solid, left axis): orange line = decision date.
Live data from BigQuery via Fivetran.
The Data That Existed, and Was Ignored ℹLeft: metrics at decision time. Right: what actually happened. This is the counterfactual gap SENTINEL would have flagged.
At time of decision
What happened next
Industry Benchmark Comparison (your metrics vs. published SaaS benchmarks) ℹYour metrics ranked against published research. NPS: Medallia B2B SaaS 2024 (n=2,847). Churn: ChurnZero 2023 (n=1,200). Bottom quartile = immediate action needed.
Loading benchmark data...
Signals That Predicted This ℹResearch-backed leading indicators that existed before the outcome. SENTINEL uses these to fire early warnings.
Alternative Explanations (correlation is not causation) ℹGemini 3 generates plausible confounding factors a skeptic would raise. Intellectual honesty before attributing causation.
SENTINEL generated these alternative explanations using Gemini. A rigorous analysis considers confounding factors before attributing causation.
Bradford Hill Causation Criteria (nine-criterion causal inference framework, Hill 1965) ℹHill (1965): 9 criteria for causal inference from observational data. Each scored 0–1. Score ≥ 0.7 = criterion met. Originally from epidemiology; applied here to business decisions.
Loading Bradford Hill analysis...
All Decisions in the Lookback Window (ranked by causal signal, no cherry-picking) ℹEvery decision in the 14–90 days before the outcome, ranked by causal signal strength. SENTINEL shows every candidate. You make the final call.
What To Do Differently Next Time ℹSpecific, actionable recommendations tailored to the metrics and decision context. Not generic advice.
Decision Log
ℹEvery business decision logged here with a live Fivetran data snapshot frozen at the moment of logging. Click "Details" to see exactly what metrics existed when the decision was made.
Transcript Decisions
ℹPaste any unstructured text: meeting notes, Slack threads, emails, or strategy docs. Gemini 3 reads it and extracts every business decision with type, rationale, and participants. Each extracted decision is automatically logged with a live Fivetran metrics snapshot.
Paste any meeting transcript, Slack thread, or doc, and Gemini 3 extracts every decision automatically
Gemini 3 powered extraction
SENTINEL reads unstructured text and extracts every business decision, complete with type, rationale, and participants. Each extracted decision is automatically logged with a live Fivetran metrics snapshot. No manual entry required.
Ask SENTINEL
ℹAsk natural language questions about your decision history. SENTINEL uses Gemini 3 with full scenario context: it cites specific metrics, dates, and decision IDs in every answer.
Ask anything about your decision history, and Gemini 3 answers with real data citations
Why did we raise prices?
What caused the churn spike?
What data existed at decision time?
What should we have done differently?
What are the active early warnings?
Explain the Pearson r score
Your Data: Custom Analysis
ℹEnter any real decision you made and your before/after metrics. SENTINEL runs the same three-method causal inference battery (Granger causality, ITS, Mann-Whitney) used on demo scenarios. Gemini 3 generates a narrative regardless of statistical outcome.
Enter any decision and its metrics. SENTINEL runs the full causal inference battery on your data, no pre-built scenarios.
How this works: Enter a real decision you made, plus before and after metrics. SENTINEL runs Granger causality, Interrupted Time Series, and Mann-Whitney tests, the same pipeline used on the demo scenarios. No hardcoded answers.
METRICS AT DECISION TIME
ℹEnter your metrics on the day the decision was made, not after. This is the "data that existed and was ignored." Even one metric (e.g. NPS) is enough to run the analysis.
CURRENT / OUTCOME METRICS
ℹYour metrics now (or at the time the outcome became clear). SENTINEL measures the delta between these values and the decision-time values to compute the effect size.
Five or more weekly values enables Granger causality. More data points means stronger statistical tests.
Fivetran Platform
ℹThe full Fivetran integration surface SENTINEL drives via the fivetran-mcp server (stdio JSON-RPC) with REST API fallback. Every tile here is a live MCP tool call: account, groups, connectors, destinations and webhooks. This is SENTINEL's "superpower" layer.
Live control panel: account, groups, connectors, destinations, and webhooks, all via the Fivetran MCP server