Findigs replaced Fivetran with PeerDB to break growth shackles

Customer story

Findigs replaced Fivetran with PeerDB to break growth shackles

Findigs currently uses RDS Postgres as its primary database for handling rental properties, applications, decision-making criteria, and billing information. This data is ingested into their Snowflake data warehouse, which supports embedded product analytics, business dashboards, external customer data feeds, and internal operational tools. The accuracy and freshness of the data in Snowflake are crucial for their property management clients, as it helps them make informed decisions to improve the tenant experience.

Findigs couldn't sync large tables with Fivetran due to potential 5x cost increase

Findigs were using Fivetran to replicate data from Postgres to Snowflake. One of their major hurdles with Fivetran was the inability to replicate their large tables that had over a billion rows. These tables were append-only tables and were growing fast. Using Fivetran for these tables was impractical because of its pricing model. Fivetran charges based on the number of new rows moved, which could lead to skyrocketing costs by orders of magnitude.

Not being able to replicate these tables was affecting their business because it caused gaps in their data analysis. They needed historical data for their customers’ measures of success on the platform, including vacancy rates and unit days on market.

"We were unable to move our historical tables using Fivetran due to increasing costs and data volumes. This limitation created gaps in data analysis, impacting our customers' ability to get a full view of the business data." - Jason Jho, CTO, Findigs

Findigs also wanted more control over replication speed and latency

Additionally, Findigs wanted better control over replication speed and latency for both initial load and change data capture (CDC). There were a few large tables for which they wanted to improve initial load and resync times, as Fivetran was very slow and did not provide knobs to tune performance. Additionally, there were a few tables with latency requirements less than what Fivetran supported (i.e. 15 mins).

Findigs replaced Fivetran with PeerDB to achieve more with less

Findigs first came across PeerDB through PeerDB’s hacker news launch. They resonated with PeerDB's focus on Postgres replication. The value proposition of 10x faster replication from Postgres to Snowflake, along with PeerDB's predictable and cost-effective pricing model, aligned with their requirements. Let us understand how PeerDB was able to deliver value to Findigs.

Predictable Pricing Model ensured that costs don't increase by 5x

PeerDB charges based on provisioned compute (vCPU/Memory) rather than the amount of data transferred. Findigs had a 3TB Postgres database, replicating around 500 million rows per month with an average replication latency of 1 hour. Based on these parameters and some testing, the PeerDB team recommended a suitable tier that could handle this load.

This allowed Findigs to know exactly what their monthly costs would be, with no surprises in the future. It gave them the freedom to move their large historical tables from Postgres to Snowflake with no worry, thereby enabling their customers to obtain a full view of the business data!

PeerDB provided fine tuned control over pipeline performance

PeerDB provides a simple yet advanced data movement experience for users. There are multiple knobs that users can tune to control performance of both initial load and change data capture (CDC). For example, for the initial load, users can specify parallelism, and for CDC, they can specify the desired latency, which starts from 10 seconds.

Findigs found this configurability very useful. Using the parallelism feature, they were able to move their large tables with over a billion rows within a day instead of weeks. They were also able to configure lower replication latency of a few minutes for some of their tables. This level of customization and efficiency wasn't achievable with Fivetran.

In today's market, numerous data ingestion tools exist. However, PeerDB stands out by uniquely leveraging PostgreSQL's native capabilities. This approach not only reduces costs but provides significant performance advantages, enabling seamless synchronization of billions of active rows monthly. Moreover, PeerDB offers the right level of visibility and customization, ensuring our data reaches our customers accurately and on schedule.

Easy to configure and monitor with confidence through PeerDB Cloud

PeerDB Cloud provided a simple and a low touch experience to setup and manage replication from Postgres to Snowflake. Findigs was able to spin up a PeerDB Cloud instance and kick off replication within a few minutes. Through PeerDB Cloud's sleek UI, they were able to gain full operational visibility, including progress of the Initial Load, throughput per table, replication latency, logs, alerts, Postgres specific metrics, and so on.

Technology Stack and Architecture

Findigs replicates data from Postgres to Snowflake using PeerDB. Over 90% of the data in Snowflake comes from Postgres. Hence PeerDB is a critical component of their Data Stack. On Snowflake, they perform periodic transformations/aggregations using dbt and expose processed data to their customers through Thoughtspot Analytics and other downstream clients.

Try out PeerDB

We hope you enjoyed reading this customer story. If you have a use case similar to Findigs and want to use PeerDB to replicate data from Postgres to Snowflake:

About

Findigs makes renting simple and accessible for everyone by streamlining the application and decision process, reducing tenant fraud, and offering 1-click services for all renters.

Use case

Property Tech (PropTech), Fintech, SaaS Data Analytics and BI

Solutions

Postgres to Snowflake CDC

Authors

Jason Jho, CTO, Findigs
‍Sai Srirampur, CEO, PeerDB