.webp)
Together AI scales outbound and unlocks an untapped stream of 500+ new prospects with Unify
Goldbelly is a curated online marketplace for gourmet food and food gifts from across the US. With an ever-expanding, curated selection of over 10,000 restaurants and food makers, Goldbelly offers a platform for small local businesses to reach new customers outside of their local area.
Goldbelly uses Snowflake as its primary data warehouse, aggregating data from various sources including its core customer-facing application.As is often the case, as Goldbelly's growth skyrocketed, so did its Snowflake costs. The data team sought to improve the efficiency of their Snowflake environment and reduce operational costs, but were cautious about disrupting critical or customer-facing workloads.
Espresso AI’s Analysis
To understand the optimization opportunity,, Espresso AI followed a three-step process:
- Establishing a baseline: Analyzed the customer’s metadata in a proprietary simulation environment to replicate daily workloads and pinpoint inefficiencies.
- Modeling agent impact: Simulated how Espresso’s autoscaling and scheduling agents would optimize usage by reducing idle time and improving warehouse efficiency.
- Synthesizing the data: Delivered a detailed per-warehouse breakdown of potential savings from these optimizations.
Espresso’s autoscaler was projected to reduce Goldbelly’s Snowflake costs by 19%.
Implementation
Espresso AI’s onboarding process was fast, secure, and required minimal effort from Goldbelly:
- User setup: Goldbelly ran a script to create an Espresso AI user with restricted Snowflake permissions.
- Risk-free trial: A two-week trial with a money-back guarantee demonstrated savings in a live environment.
- Full control: Goldbelly retained complete control, with logs of all actions and a guarantee that no queries would be altered.
To support monitoring and validation:
- Espresso AI provided daily comparisons of credit usage before and after optimization.
- All automated changes, such as cluster scaling, were fully transparent and reversible.
Results
During the trial, Goldbelly saw immediate value from Espresso AI:
- Cost savings: The projected 19% reduction was validated in production, delivering over $40K in annualized savings.
- Efficiency gains: Dynamic workload management reduced idle time and optimized resource usage. The time saved optimizing Snowflake was able to be reallocated to higher value work.
- Full transparency: Goldbelly’s data team verified savings via detailed logs and dashboards, including controlled tests comparing optimized and non-optimized warehouses.
Following the successful trial, Goldbelly expanded Espresso AI’s solution to more Snowflake warehouses.
