Portfolio Intelligence
SYNTHETIC CASE STUDY

Portfolio analytics,
open to inspection.

Explore acquisition, ordered journeys and returning cohorts from an entirely synthetic event stream. These results are not from real portfolio visitors. Your private real-traffic dashboard is a separate view.

Fixed historical window; not a live feed.

Eligible users

—
Distinct simulated users in selected sessions.

Eligible sessions

—
New session at exactly 30 minutes of inactivity.

Engagement rate

—
60 active seconds, 2 views or a high-intent action.

High-intent rate

—
Résumé, LinkedIn or contact action; once per session.

Sessions across the selected window

Zero-activity days are included. Daily users must not be summed to obtain period users.

View daily counts
Sessions and distinct users by UTC day
Day Sessions Users

An ordered journey, not four counters

Same session, strictly increasing event positions. A contact action before a project does not become a downstream conversion.

    Rates use all eligible sessions. Visitors can take other paths; this sequence describes one journey and does not explain causation.

    Where sessions begin

    Acquisition and landing page use the first event in time, with event ID as the tie-breaker. Later direct events cannot overwrite that source.

      Project attention and later actions

      One session-project pair per row. Repeated project views do not inflate the denominator. A later action is associated with a project; it is not attributed lift.

      Selected window; deep means at least 45 summed project-event seconds
      Project Sessions Deep sessions Later-action sessions

      Return cohorts with an observation boundary

      Full 90-day fixture history, independent of the analysis selector. Monday-based cohorts are defined by first observed activity. Each cell shows returning users / cohort users. Incomplete weeks remain unavailable, even if activity has already arrived.

      Observed weekly return, not cumulative retention or real evaluator behavior
      First observed week Users W0 W1 W2 W3 W4

      A zero is an observed absence; an em dash means the full week has not elapsed. The fixture cannot establish real-world retention or business lift.

      Trust is an accounting identity

      Loading validation evidence…

      Every delivered row has one disposition. Duplicate IDs, invalid contracts, bot, internal, test-marker and consent-denied events stay out of eligible sessions.

      raw_events→stg_event_audit→stg_events→int_sessions→session / journey / cohort marts→verified JSON
      Inspect provenance and metric contracts

      Model contracts · Definitions · Independent Python oracle · Architecture & recovery

      What this case does not claim

      REFERENCE ADAPTERS · NOT ACTIVATED

      GA4, PostHog and BigQuery references are not execution evidence. Real p75 Web Vitals, visitor-level funnels and real retention are unavailable. The live owner view uses cookie-free GoatCounter aggregates; it cannot reconstruct individual journeys. This page sends no analytics events and stores no visitor identifiers.