The Official Census Blog
Check our product
Get the best data & ops content (not just our post!) delivered straight to your inbox

Integration platforms are an old idea

Feb 25, 2022
Brad is a co-founder of Census who doubles as a software developer in San Francisco. Prior to founding Census, he co-founded Meldium, a YC-backed startup helping teams & companies manage accounts & passwords for cloud apps.

Integration platforms have gained popularity since Web 2.0 and, thanks to REST APIs and webhooks, the ecosystem has developed over the last decade so much to the point that Zapier is a household name. Unfortunately, most integrations evolved without consideration for other evolutions happening in parallel—namely the accessibility and power of a data warehouse as a central source of truth.

At Census, we didn’t break any new ground by saying System A should be able to talk to System B, even if they aren’t designed to do so. So what makes Census different?

3 key tenets make Census different from legacy integration platforms:

  1. All data should be marshaled through a central source of truth
  2. The best central source of truth is a data warehouse
  3. Integrate data, not events

Let's break these down:

Marshal data through a central source of truth: Data teams achieve scale through simplicity and manageability. We talk to a lot of companies who did the easy thing–started connecting SaaS apps directly to each other via point to point–and are now tangled up in a web of undocumented and unmanageable integrations.

The best central source of truth is a data warehouse: The SQL/relational ecosystem is arguably the richest in all of computing (right up there with TCP/IP), and while SQL has its warts, it's pragmatic, widely used, and going stronger than ever.

SQL is standards-based while still being extensible and dynamic, widely supported, easy to hire for, and inspectable. And there are a tremendous number of tools you can bring to bear on things that speak SQL.

Integrate data, not events: Legacy iPaaS are designed to capture changes that originate in System A and forward them as actions to System B. This model is contained by what A and B decide to generate and consume. Operational Analytics is different in that it’s data-oriented, not action-oriented.

The ELT approach nails this. You read high-fidelity data out of your source systems and into your warehouse before you perform any transformations, even trivial ones.

Census, and Operational Analytics by way of reverse ETL integration, is the logical extension of this—instead of trying to actuate the destination system to get it to some desired state and being constrained by that system's levers and buttons, just load the data you want into that system's tables directly.

Salesforce is successful not because it’s a great CRM, but because it’s a good-enough CRM that's programmable. Most SaaS applications are much less programmable, and nobody wants to learn dozens of different, proprietary SaaS programming languages.

The good news is you can reprogram your SaaS today, and you can do it using a tool you already have (your warehouse) combined with Census to operationalize your data and to put exactly the bits you want into your SaaS tools directly.

This was the aha! moment our team had way back when we were writing our first lines of code, and our conviction in this idea has only grown stronger since.

Census is a deeply transformative technology and our customers are just beginning to tap that power. Want to be a part of this transformation? I'm hiring engineers of all levels - come join me and let's be part of this together.

Related articles

No items found.
Related integrations
No items found.
Get the best data & ops content (not just our post!) delivered straight to your inbox