Pineapple on pizza

No data available

Feature ownership
Members

Goals

Q2 2026 objectives

Isolate ingestion Kafka topics

Kafka infrastructure is shared across ingestion and other teams, making it much harder to maintain ingestion reliability and scale independently.

  • Separate Kafka infrastructure for ingestion from other teams
  • Evaluate WarpStream as a full MSK replacement
  • Fall back to multi-MSK cluster setup if WarpStream doesn't pan out

PersonHog

Person processing should be stable and fully handled by PersonHog.

  • Route all reads and writes through PersonHog — no direct DB calls from other services
  • Define and guarantee SLAs for read/write latency and availability
  • Ship the read/write split architecture to production
  • Roll out safely with strong validation of data

Redesign the capture contract

Capture lacks the structure needed for strong rate limiting and efficient downstream processing.

  • Define a new capture contract with a strict schema for all endpoints
  • Include critical metadata in headers so downstream consumers don't need to parse payloads
  • Standardize response and error codes for SDKs
  • Establish the SDK harness as the canonical way to stay on contract
  • Migrate all SDKs to the new contract
  • Enable rate limiting

Define and commit to ingestion SLAs

The primary ingestion lane needs consistently low latency, not just high availability.

  • Publish metrics and SLAs for all pipelines, especially the general lane
  • Automate monitoring to protect 99.99% of traffic from the 0.01% causing lag

Improve per-partition throughput

Current per-partition processing speed won't support 10x growth.

  • Benchmark and identify per-partition bottlenecks
  • Evaluate alternative languages or architectures
  • Reduce the number of running ingestion pods

Make ingestion pipelines easy to manage

The ingestion team will own more pipelines over time and needs better tooling to manage them efficiently.

  • Make Helm charts easy to configure for deploying new pipelines
  • Provision metrics and alerting automatically with each pipeline
  • Build MCP tooling for interactive pipeline support

Handbook

Features

  • Client data capture
  • Session recording capture
  • Identity resolution
  • Property definitions

Slack channel

#team-ingestion

Questions about this page? or post a community question.