Integration Engineering Blog

Your SaaS sends events to external tools, but the gap between "sends events" and "guarantees delivery" is wider than most teams realize. This post breaks down the five delivery gaps — retries, idempotency, dead letters, isolation, and visibility — and what it takes to close them.

Hand-rolled integrations look cheap until retries, dead letters, credential storage, replay tooling, and on-call load show up. A practical build-vs-buy look for SaaS teams.

OAuth tokens do not fail on schedule. They expire during jobs, race during refresh, and get revoked without warning. This guide covers the multi-tenant patterns that keep SaaS integrations running.

Duplicate webhook deliveries are normal in at-least-once systems. Learn idempotency keys, dedup strategies, and Node.js patterns that prevent double-processing.

One event in, many deliveries out — each with its own retries and failure isolation. A practical guide to fan-out architecture for SaaS integration teams.

There are three common ways to send product events like signups, upgrades, and cancellations from your SaaS to HubSpot: build the integration directly against HubSpot's API, send events through Zapier, or use an event routing layer like Meshes. Each approach has different tradeoffs in engineering effort, reliability, and long-term flexibility.

Production webhook retry logic requires exponential backoff, jitter, idempotency tracking, dead letter queues, and per-destination rate limiting — typically 1,500–3,000 lines of code before you add any destination-specific logic. This post walks through building it yourself, then explains why most SaaS teams shouldn't.

When webhooks fail and retries are exhausted, events disappear unless you have a dead letter queue. This post explains what DLQs are, how to implement them for webhooks, and why most teams get this wrong.

Syncing customer data between HubSpot and Salesforce is a rite of passage for growing SaaS teams. This post covers the common approaches, where they break, and how an event-driven integration layer avoids the mess.