Built with: ClickHouse, and Postgres with PostGIS on Supabase.
At a glance
- From timeouts to under a second: On Postgres, 25-mile comps searches were slow when they finished and often timed out. The new comps engine on ClickHouse returns them in under one second.
- About 1–2 seconds in the app: Agents see comps roughly 20x faster than before, with photos loading separately.
- No source-data rebuild: ClickHouse reads from the same data feeds Agreeable Data built for launch.
Results reported by Breezy’s engineering team, July 2026.
About Breezy
Breezy provides a mobile app for residential real estate agents. Comparable-property searches, or comps, help agents evaluate a property against similar nearby sales. Agents use those results during conversations with clients, so comps have to come back quickly.
Breezy wanted to bring nationwide property coverage into that workflow. To do that, the team needed a way to ingest multiple listing service (MLS) and property records, make them searchable, and keep the app responsive as the data grew.
Nationwide property data in, nearby comps out, with photos loading separately. Conceptual illustration.
Comps outgrew Postgres
Agreeable Data built the data feeds into Postgres on Supabase so Breezy could launch quickly. Postgres and its PostGIS extension handled the property data and map-based searches well enough to get the product live.
As coverage grew, wider searches slowed down. 25-mile comps searches, and searches across an area drawn on the map, would often time out, which no agent can wait through with a client. Breezy and Agreeable Data both worked on speeding up the Postgres queries, but the gains weren’t enough.
ClickHouse was a better fit: it’s built to scan and filter very large tables quickly, which is exactly what a wide comps search does. The question became how to make the switch without redoing the data work behind the launch.
The data foundation carried forward
Because Agreeable Data had built Breezy’s data feeds for launch, the switch didn’t start from scratch. Agreeable Data connected those same feeds to ClickHouse, so none of the data intake had to be rebuilt.
“Agreeable Data made it easy to switch over, since they were able to use the data pipelines they had already created and connect those to ClickHouse.”
— Justin DeWind, Head of Engineering, Breezy
With the data in place, Breezy’s engineers built the new comps engine on ClickHouse. Photos load separately, so the comparisons appear before the property photos finish loading.
A sub-second comps engine
By July 2026, Breezy’s engineers were reporting these results from the new comps engine:
| Measure | On Postgres | On ClickHouse |
|---|---|---|
| Core comps report, 25-mile radius | Slow when it finished; often timed out | Under one second |
| Comps results in the app | Slow when it finished; often timed out | About 1–2 seconds, roughly 20x faster |
“The wall time for the new comps engine is blazing fast. We are talking about sub seconds.”
— Justin DeWind, Head of Engineering, Breezy
For real estate agents, comps are now fast enough to run while the client is still in the room.
Agreeable Data is a fractional data team for SaaS and eCommerce companies. If your app’s search is outgrowing its first database, we can help you move it without starting over on your data.