Moat in High-Fidelity Customer Syncs
Zachary Kirby, co-founder of Vessel, on building the Vercel for integrations
The real moat is shifting away from having lots of connectors and toward running customer-facing syncs with higher fidelity, better freshness, and better control. That separates native integration infrastructure from CDP, ETL, and reverse ETL tools, which are built mainly for centralizing data for analytics and then pushing modeled outputs back out. In practice, a product team needs per-tenant permissions, rate limit handling, webhooks, and object-level debugging, not just a generic pipe between apps.
-
ETL and reverse ETL vendors are optimized for analytics workflows. Fivetran pulls data into a warehouse, dbt models it, and Census pushes warehouse outputs into tools like Salesforce. That stack serves data, ops, and marketing teams, not engineers shipping in-product integrations.
-
Native integration platforms are built around transactional product workloads. They handle read, write, and subscribe flows inside the product itself, where a SaaS vendor may need hourly or near real-time CRM syncs, custom objects, tenant-specific configs, and logs that show exactly which field or permission broke.
-
This creates overlap at the feature level, but not at the system-design level. An integration company can add ETL, pass-through APIs, or embedded flows, but the underlying architecture still reflects its core job. Breadth-first platforms smooth over the common case, while depth-first platforms win when sync quality directly affects product quality and deal conversion.
Going forward, more of the integration budget will move from internal data plumbing toward product infrastructure that helps software vendors win and retain customers. CDPs, ETL, and reverse ETL will remain the backbone for warehouse-centric analytics, but the fastest-growing edge will be platforms that make customer-facing integrations feel like a core product surface, not a back-office data job.