FFI Journey Analyser

📦Self-Hosting

💰

Costs

Monthly:Free

Annual:Free

🔑

Access

CLIENT ACCESSINTERNAL

ℹ️

AI-powered GA4 journey analysis for FE Fund Info with HubSpot and Salesforce cross-referencing

Purpose: AI-powered journey analytics for FE Fund Info (FFI). Looks up users by email or GA4 ID, cross-references HubSpot contacts, matches Salesforce leads to GA4 journeys, and generates Gemini summaries of behaviour and conversion paths.

Where it runs: Google Cloud Run (europe-west1, project ffijourney) behind Cloudflare Access at ffijourney.pacing.agency. A Cloudflare Worker proxies the custom domain to the *.run.app URL.

Last verified: July 2026

Quick Reference

Feature	Capability
Journey analyser	Email or GA4 `_c_ffi` lookup with AI 5-section journey summary
Leads intelligence	Salesforce leads enriched with GA4 match status, filters, and export
Audience builder	Identified GA4 users with saved filter sets and snapshot cache
Global data chat	Gemini Flash/Pro with aggregate BigQuery tools
API + MCP	17 REST endpoints at `/api/v1`, 12 MCP tools
Reports	Six auto-discovered BigQuery analytics reports
Data health	15 parallel metric jobs per `/health/data` refresh

Overview

FFI Journey Analyser is a client-specific deployment forked from Pacing-Agency/user-journey-analyzer. It replaces Supabase auth with Cloudflare Access and adds HubSpot mirror queries, Salesforce lead matching, and a full leads intelligence workspace.

Key capabilities:

Dual lookup — email (HubSpot cross-reference + multi-method GA4 matching) or direct GA4 user_id / user_pseudo_id
HubSpot → GA4 matching — gclid, msclkid, li_fat_id, conversion-time window, and SF-date window
Leads table — Salesforce CRM leads with inline rematch, CSV import, and HubSpot change badges
Journey cache — merge-based enrichment pipeline with BigQuery cache (ffijourney.journey_cache)
API keys — self-service rotation at /profile; scope-gated /api/v1 for automation and MCP

A reference re-implementation of FFI also exists in the AQL Journey monorepo (clients/ffi/) but the live FFI tool remains this separate codebase and deployment.

Account Access

Production URL: ffijourney.pacing.agency
Authentication: Cloudflare Access (FFI staff Google accounts via Zero Trust policy)
API docs: Swagger UI at /api/v1/docs
MCP setup: /mcp-setup (Cursor, Claude Desktop, Zed)
Account type: CLIENT_ACCESS, INTERNAL

Technology Stack

Layer	Technology
Backend	Python Flask + FastAPI (`/api/v1/` via `DispatcherMiddleware`)
AI	Gemini 2.5 Flash / 2.5 Pro via `google-genai` (Vertex, `europe-west1`)
GA4 data	BigQuery `ga4-streaming-data.analytics_297687303`
HubSpot data	BigQuery `ga4-streaming-data.hubspotdata1.ingest1`
Salesforce leads	BigQuery `ga4-streaming-data.leads1`
Auth (web)	Cloudflare Access JWT
Auth (API)	`X-API-Key` (SHA-256 hashed keys in `ffijourney.ffi_app.api_keys`)
MCP	`mcp_server.py` (12 stdio tools)
Deployment	Google Cloud Run + Cloudflare Worker proxy

Repository

Source: Pacing-Agency/ffi-journey-analyser
Base fork: Pacing-Agency/user-journey-analyzer
Implementation plan: PLAN.md in the repository

Data Architecture

Dataset	GCP project	Purpose
`analytics_297687303`	`ga4-streaming-data`	GA4 events (hostname filter: `www.fefundinfo.com` required)
`hubspotdata1`	`ga4-streaming-data`	HubSpot contact mirror
`leads1`	`ga4-streaming-data`	Salesforce leads, matching, chat sessions, HubSpot snapshots
`journey_cache`	`ffijourney`	Enriched per-user journey cache
`audience_cache`	`ffijourney`	Audience Builder snapshots
`ffi_app`	`ffijourney`	API keys, admin allow-list, import log

Primary GA4 identifier: user_id in _c_ffi format (99%+ coverage post-April 2026 GTM publish). Use user_pseudo_id for historical data before April 2026.

Local Development

git clone https://github.com/Pacing-Agency/ffi-journey-analyser.git
cd ffi-journey-analyser

python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt

gcloud auth application-default login
gcloud config set project ffijourney

cp run_local.example.py run_local.py   # set CF_DEV_EMAIL; CF_DEV_MODE=true
python3 run_local.py

Runs at http://localhost:8080. Local dev bypasses Cloudflare JWT when CF_DEV_MODE=true.

Deployment

Push to main triggers the Deploy to Cloud Run workflow (deploy.yml):

Authenticates to GCP as ffi-deploy-github@ffijourney.iam.gserviceaccount.com
Deploys via gcloud run deploy --source .
Purges Cloudflare edge cache for ffijourney.pacing.agency

Manual deploy: ./deploy.sh

CI / CD

Workflow	Trigger	Purpose
`deploy.yml`	Push to `main`	Cloud Run build, deploy, CF cache purge
`hubspot_sync.yml`	Daily 07:00 UTC	HubSpot delta → BQ upsert → lead cascade
`warm-cache.yml`	Daily + weekly	Audience snapshot and journey cache warm-up

User Journey Tool — base platform this tool was forked from
AQL Journey — ecommerce sibling on the shared CORE engine (separate live deployment)
Google Analytics — GA4 property and BigQuery export
Google Cloud — ffijourney and ga4-streaming-data projects
Cloudflare — Access policy and Worker proxy for ffijourney.pacing.agency
n8n — workflow automation (can call /api/v1 with scoped API keys)

Last Updated: July 2026
Owner: Ben Power
Status: Active (production)

Costs

Access

Quick Reference​

Overview​

Account Access​

Technology Stack​

Repository​

Data Architecture​

Local Development​

Deployment​

CI / CD​

Related Tools​