Cypher-first analytics
ROAS, LTV, and uplift computed in Neo4j with channel, cohort, and experiment dimensions.
This portal replaces the restaurant landing page with the new analytics case study: Neo4j-backed KPIs, BI storytelling, and geospatial targeting. It is built for LinkedIn, clients, and recruiters to explore the project live on ghuman.online.
ROAS, LTV, and uplift computed in Neo4j with channel, cohort, and experiment dimensions.
GeoCell catchments and cannibalization overlays mapped to stores and campaigns.
Referrals and communities surfaced with graph algorithms for advocacy targeting.
Power BI pages for performance, cohorts, geo, experiments, and influence.
Saved queries cover ROAS, CAC, CTR/CVR, cohorts, LTV, churn, uplift tests, referral influence, geo catchments, and cannibalization. Exports feed Power BI and the map overlays.
// Channel performance and ROAS
MATCH (camp:Campaign)-[:USES]->(ch:Channel)
OPTIONAL MATCH (cust:Customer)-[o:ORDERED]->(ord:Order)-[:FROM]->(camp)
WITH ch.name AS channel, camp.name AS campaign,
sum(o.spend) AS spend, sum(ord.value) AS revenue
RETURN channel, campaign, spend, revenue,
CASE WHEN spend = 0 THEN 0 ELSE revenue / spend END AS roas
ORDER BY roas DESCPython ETL loads CSVs to Neo4j (customers, campaigns, impressions, clicks, orders, stores, GeoCells).
Constraints + projections; derive NEAR, LOCATED_IN, REFERRED, and ABTest variant relationships.
Cypher/graph algorithms for ROAS, LTV, churn, influence, and cannibalization heatmaps.
Exported tables to Power BI and GeoJSON overlays surfaced through the Next.js Insights Portal.
Set environment variables in .env.local and run npm run test:neo4j to validate the Aura/Neo4j connection.
The shared driver utility (src/lib/neo4j.ts) is ready for API routes or server components to execute Cypher and stream results into the portal.