How Top Companies Save Millions in Snowflake Costs

A case study of Instacart, Opendoor, HelloFresh, and Hightouch

Sahil Singla

Co-founder of Baselit

In today’s data-driven world, managing cloud computing costs effectively is a critical challenge for many top companies. Snowflake, a leading cloud data platform, offers powerful capabilities but also poses significant cost implications if not managed wisely. This blog provides an in-depth look at how prominent companies like Instacart, Opendoor, HelloFresh, and Hightouch have successfully navigated the challenges of Snowflake to achieve significant cost savings.

From Instacart slashing their $28 million annual Snowflake bill by 56% through detailed cost monitoring and budget allocation, to Opendoor’s sophisticated method for query cost calculation leading to a 15% cost reduction, these stories are not just case studies but beacons guiding towards efficient cloud cost management. We’ll explore the strategies these companies employed, such as warehouse optimization, query analysis, and the use of tools like Query Profile and Query Acceleration Service, offering valuable lessons for businesses aiming to enhance their cloud computing efficiency.

How Instacart cut their $28 million Snowflake bill by 56%

In 2021 and 2022, Instacart faced a hefty Snowflake bill of $28 million each year. They took immediate action to optimize their costs [1]. Their strategy included:

Fine-grained Cost Monitoring: Instacart started monitoring costs at the query level, using query tags to pinpoint areas needing cost reduction.
Budget Allocation: They assigned budgets to each department, considering past usage and growth projections.
Warehouse Utilization: The company optimized warehouse use by tracking metrics like query runtime and spillage, focusing on expensive queries.
Query Analysis: Using Query Profile, Instacart identified costly steps in query execution and mismatched warehouse capacities.
Handling Spillage: For queries still causing remote spillage after optimization, they shifted them to larger warehouses.

Their efforts revealed that while Snowflake was not necessarily more expensive than its peers, but keeping costs from creeping up required careful monitoring, controls, and best practices.

How Opendoor reduced their Snowflake costs by 15%

When Opendoor launched their project to cut their data warehouse expenses, their analytics team faced a hurdle: they had some rudimentary cost data and estimates but lacked specific details on the costs of individual query executions [2]. To tackle this, they created a method to calculate the cost per query. This led to comprehensive reports that tracked expenses for each warehouse and query, offering insights into high-cost areas and usage patterns.

With this in-depth cost analysis, Opendoor successfully executed various optimizations, leading to a notable 15% decrease in Snowflake spend. They used these reports to pinpoint and rectify inefficiencies, by employing tactics like:

Adjusting warehouse sizes for certain queries
Optimizing expensive SQL queries
Changing query schedules

How HelloFresh saved 30% costs using an ideal warehouse size configuration

HelloFresh, the top meal kit provider in the US, achieved a 30% cost reduction in their Snowflake warehouse expenses [3]. Their approach involved analyzing different warehouse configurations using a custom-built, multi-threaded Python application. This tool tested various workloads on Snowflake, helping to identify the most efficient warehouse size for their requirements.

To further cut costs, HelloFresh integrated Query Acceleration Service into their Snowflake framework. This reduced the average query’s performance impact, even with a smaller warehouse. The result was a 30% decrease in warehouse credit usage for their largest warehouse in 2023, with only a minor increase in median query time.

The above strategy not only offered immediate savings but also set a template for optimizing all their Snowflake warehouses, effectively answering their key question about the ideal warehouse size for their workload.

How Hightouch found 5-figure savings with two simple methods

Hightouch, the leading reverse ETL platform, significantly reduced its Snowflake costs through two key strategies [4]. First, they downsized an oversized warehouse used for dbt Cloud jobs from Medium to Small, a change prompted by their findings that the warehouse was underutilized. Although this adjustment slightly increased job durations, it greatly reduced credit consumption, leading to cost savings.

The second strategy involved optimizing a specific model that was bottlenecking their job processes. By implementing incremental materialization, they reduced the model’s runtime by about 91%, further decreasing credit usage. This optimization was particularly effective as the model was run frequently. The combined effect of resizing the warehouse and streamlining the bottleneck model resulted in substantial savings.

The case studies of Instacart, Opendoor, HelloFresh, and Hightouch highlight a fundamental truth in cloud computing: significant cost savings are achievable with the right strategies. By employing fine-grained cost monitoring, optimizing warehouse utilization, and adjusting to ideal warehouse sizes, these companies have demonstrated that managing Snowflake costs is not just about cutting expenses but also about smart resource management. These stories serve as a template for other organizations seeking to optimize their cloud data platform costs.