Enrichment (Geo, Device, IP)

💼 Business Context

• "Users in Mumbai on iOS" style personalization and analytics constantly need geo/device context that raw events rarely include on their own
• Doing enrichment once, centrally, means every downstream team gets consistent geo/device fields instead of reimplementing their own lookup
• Owned by Platform Engineering

🔌 Technical Overview

Enrichment runs as a .NET Core middleware step in the Ingestion API pipeline: IP address resolves to geo (country/city) via a MaxMind-style lookup service cached in Azure Cache for Redis; device/browser fields are parsed from the User-Agent header (client-supplied device fields on mobile are trusted directly); and the raw IP is then discarded rather than persisted, keeping the stored event privacy-minimized while still geo-enriched.

💾 Enriched Context Block

"context": {
  "geo": { "country": "IN", "city": "Pune" },
  "device": { "type": "mobile", "os": "iOS 18", "browser": null }
}

🔗 Integration Points

• Azure Cache for Redis — cached IP-to-geo lookup table for low-latency enrichment
• .NET Core Ingestion API middleware — enrichment runs inline before publishing to Event Hubs
• User-Agent parsing library — device/browser classification
• Identity & Profile Service — enriched geo/device fields feed the unified profile's device graph

🧰 Services Consumed

• Owning microservice — Cxos.Ingestion.Application (see the Full Application Service Map)
• Database — Azure Cache for Redis (cache only, no system-of-record database)

⚠️ Non-Functional Considerations

• Scale: enrichment lookups are cache-backed to keep per-event latency low even at high throughput
• Latency: adds low-single-digit milliseconds via the Redis cache; a cache miss falls back to a slower lookup without blocking the request
• Reliability: enrichment failures never block ingestion — an event proceeds with partial context rather than being rejected
• Security/Privacy: raw IP addresses are discarded immediately after geo resolution — only the derived country/city is persisted

🎯 Enterprise Example

A retailer's Analytics API can answer 'which cities are driving mobile conversion this week' without any team having written custom geo-lookup code — every event already arrived enriched, consistently, from day one.