Discuss your projectContact
Real estate · USA

n8n automation layer for a real estate brokerage

Sales, operations and marketing workflows moved from Zapier to self-hosted n8n, with one team responsible for all of them

Industry
Real estate
Geography
USA
Type
Automation
Engagement
Ongoing, 15+ months
Architecture
Main, worker and monitoring servers
Support
Tier 2 and Tier 3
Technologies
  • n8n
  • Zapier
  • Supabase
  • TypeScript
  • Python
  • Docker
  • Grafana
n8n automation layer for a real estate brokerage

Project overview

The client is a large residential real estate brokerage in the United States. Its sales, operations and marketing processes run on automations that connect the CRM, the phone system, web forms and project boards, with Supabase as the main data store.

These automations were spread across Zapier, n8n and custom code. The brokerage wanted one team to own the whole layer and report to its CTO, and to move the Zapier flows to self-hosted n8n to keep costs under control. New automation requests had to be picked up in short cycles.

What we did

We took over the existing workflows, documented them and moved them to self-hosted n8n, where the client pays for servers instead of paying for every task run. The CRM, telephony and project tools are connected through their APIs.

  1. Audit of the existing Zapier and n8n flows and the systems each one touches

  2. Migration to self-hosted n8n, one group of flows at a time

  3. Custom TypeScript and Python nodes, webhooks and small services where a no-code step was too slow or too limited

  4. Monitoring with error tracking and Slack alerts for overloads and unavailable services

  5. Weekly planning and demos with the CTO, plus second and third line support

Key results

One team owns the brokerage's automation layer and demos the work to the CTO every week

Zapier flows moved to self-hosted n8n, so the automation bill no longer grows with every task run

Separate main, worker and monitoring servers instead of a single machine

No webhook timeouts after a 2026 load spike: webhooks answer at once, two workers process the queue, database connections are pooled

Workflows documented for non-technical staff

In summer 2026 a burst of CRM webhooks overloaded n8n. We changed the webhook flows to acknowledge each request at once and process it in the background, added a second worker and tuned the database connection pools. The timeouts stopped. The memory alert now fires at 90% usage, so the team is no longer paged for normal peaks.

Around 30 Zapier flows now run in n8n on the client's own servers, and the brokerage spends roughly 40 to 50% less on automation than it would with per-task pricing. We have worked with the client without a break for more than a year.

Frequently asked questions

Why move from Zapier to n8n?

Zapier charges per task, so a busy brokerage pays more every month as it automates more. Self-hosted n8n costs the same server fee at any volume and lets engineers add custom code where a flow needs it.

Can n8n handle high webhook volume?

Yes, in queue mode. Webhooks should answer immediately and pass the work to separate workers, and the database needs connection pooling. This client runs on that setup.

Find out what a project like this would cost

Discuss your project

Tell us about your goals — we’ll outline the team, scope and timeline.

More case studiesAll case studies

Read the full case study Discuss your project