How a fast-growing AI startup tamed spiralling Snowflake costs by 60%

Snowflake costs spiralled 10X as a hypergrowth AI startup scaled rapidly

Synthesia operates a modern data stack with Snowflake as their central data warehouse, supporting over 300 total users. The platform serves 80 direct Snowflake users plus 250+ additional users through Omni for analytics and dashboards.

Their data infrastructure follows a layered Kimball approach using dbt for transformations, Fivetran for ingestion, and Segment for event data. Most stakeholders access only the cleaned gold layer, while dbt developers and admins have full platform access.

As a fast-growing AI company, Synthesia's Snowflake spending increased dramatically - growing 10X over two years as they scaled their platform and user base.

The situation became critical when they realized their annual credit purchase would run out halfway through their contract, forcing immediate action on cost optimization.

Basic cost monitoring tools couldn't provide needed visibility

Before SELECT, Synthesia relied primarily on Snowflake's built-in cost management interface, which proved inadequate for their complex infrastructure needs.

The challenge was compounded by the fact that non-technical team members had initially managed their Snowflake setup, with their original admin being a VP of Product. This meant the team lacked the technical foundation needed for sophisticated cost monitoring and optimization.

Adopting SELECT for immediate cost visibility and control

With costs spiralling and credits running out faster than expected, Synthesia needed immediate visibility into their Snowflake spending patterns. Rory reached out to SELECT for help with their cost management challenges.

The adoption process was remarkably fast. While IT approvals took about a week, the actual SELECT setup took under one hour once permissions were in place.

The team immediately activated SELECT's automated savings functionality, providing instant cost reduction benefits from day one.

SELECT delivered 60% cost savings through targeted optimizations

Within the first month, SELECT enabled Synthesia to identify and address their primary cost drivers, resulting in dramatic savings.

After enabling SELECT's Automated Savings feature, Synthesia shifted towards the second part of their optimization process which involved multiple targeted interventions identified through SELECT's visibility & insights features.

SELECT's immediately highlighted specific areas for improvement, including dbt models that needed incrementalization and warehouses that were significantly oversized for their workloads.

One major discovery was warehouses running three sizes larger than necessary, with SELECT's utilization metrics providing the confidence needed to right-size them immediately.

The team also identified inefficient Census jobs syncing data to Salesforce that could be refactored or turned off entirely after consulting with business stakeholders.

Streamlined monitoring enables proactive cost management

SELECT transformed Synthesia's cost management from reactive firefighting to proactive monitoring. Rory established a routine of checking three to four customized saved views every Monday morning.

His weekly monitoring includes warehouse costs at weekly and monthly horizons, long running queries (customized for their specific use cases), unused assets, and monitoring recently introduced workloads.

This streamlined approach takes just minutes but provides comprehensive visibility into their entire Snowflake environment, enabling early detection of cost spikes or new inefficient workloads.

Beyond cost savings, the optimizations also delivered performance improvements on their dbt pipelines, creating additional value from their SELECT investment.

Looking ahead: Centralized cost reporting and department allocation

Synthesia plans to leverage SELECT as their central hub for all Snowflake cost reporting and analysis. The team is exploring using dbt meta keys from SELECT's dbt integration feature to allocate spending by department, enabling more granular cost attribution across their organization.

As a cost-conscious company, this capability will be essential for maintaining their disciplined approach to spending as they continue scaling their AI platform.

SELECT's upcoming Omni integration will provide additional visibility into dashboard costs and lineage, while partition monitors will enable granular cost anomaly detection at the dbt model level.