Causal Ledger
The Causal Ledger is Apex's timeline intelligence system. It records every significant change — configuration updates, experiment launches, connector events, anomalies — and overlays them on your business metrics so you can see what changed when and how it affected outcomes.
Why It Matters
Growth teams make dozens of changes: launching experiments, adjusting scoring models, connecting new data sources, changing budget allocations. Without a timeline, it's impossible to know which change caused a metric to move. The Causal Ledger gives you that visibility.
What Gets Recorded
The ledger automatically captures events across several categories:
Configuration Changes
| Event | Recorded when |
|---|---|
| Scoring model activated | A new scoring model version becomes active |
| Scoring rule changed | A scoring model is created, edited, or deleted |
| Vertical changed | The project's industry vertical is updated |
| Funnel changed | The conversion model is switched |
| Goal changed | A conversion goal is created, updated, or deleted |
| Budget changed | A marketing budget is created or updated |
| Integration changed | A snippet, SDK, or MCP server is installed or verified |
Experiment Events
| Event | Recorded when |
|---|---|
| Experiment started | An experiment begins collecting data |
| Experiment ended | An experiment is stopped or reaches its sample size |
| Experiment winner | A variant is declared the winner and promoted |
Connector Events
| Event | Recorded when |
|---|---|
| Connector connected | A sensor (Google Ads, Stripe, HubSpot, etc.) is linked |
| Connector disconnected | A sensor loses its connection or is removed |
Detections
| Event | Recorded when |
|---|---|
| Visitor milestone | Cumulative visitor count crosses a threshold (1, 10, 100, 1K, 10K, 100K) |
| Anomaly | A metric falls outside its normal range (2 standard deviations from the 30-day average) |
| Identity stitch | An anonymous visitor is linked to a known contact |
Annotations
You can also add manual notes to the timeline — useful for recording external events like product launches, press coverage, or marketing campaigns that happen outside Apex.
Daily Metric Snapshots
The Causal Ledger captures a daily snapshot of key metrics:
- Total visitors and total events
- Form submissions and conversion rate
- Average lead score and hot lead percentage
- Active experiments
These snapshots form the baseline that the anomaly detector compares against and that the correlation engine uses to measure before/after impact.
Detected Correlations
When enough snapshot data exists around an event, Apex computes before vs. after correlations: average metric values in the 7 days before an event compared to the 7 days after. Each correlation includes:
- The metric affected (visitors, conversion rate, lead score, etc.)
- The change (percentage and direction)
- A confidence level based on the data available
Info
Correlations show association, not definitive causation. They're a starting point for investigation — "conversion rate jumped 15% after we launched experiment X" is a signal worth exploring, not a proof.
Anomaly Detection
The anomaly detector runs daily after each metric snapshot. It compares today's values against the trailing 30-day average (requiring at least 7 days of history). If any metric falls outside 2 standard deviations from the mean, an anomaly event is recorded with the expected range and actual value.
This catches sudden shifts — a traffic spike from a viral post, a conversion rate drop from a broken form, or a scoring change that reclassified your pipeline.
Using the Timeline
The timeline page (/dashboard/timeline) shows:
- Causal Ledger chart — An area chart of your selected metric with vertical markers on dates that have events
- Event feed — Chronological list of all recorded events, filterable by category (Configuration, Experiments, Detection, Milestones, Notes)
- Correlation table — Before/after metric comparisons for each event
- Capture Snapshot — Manually trigger a metric snapshot at any time
- Add Note — Record an annotation for context that Apex can't detect automatically
Next Steps
- Set up scoring models — model changes are tracked automatically
- Run experiments — experiment lifecycle events appear on the timeline
- Connect sensors — connection changes are recorded