Snowflake: A Data Platform that Does More and Costs Less

A number of blog posts, mainly from competitors, call out Snowflake as expensive. But in our conversations with customers, we frequently hear that Snowflake provides the best value for the money of any enterprise data platform option available today.

So what gives? The fact is that data platforms are not created equal, and comparing the total cost and the value derived from one with another is difficult. Usage patterns vary widely and requirements change constantly, so an accurate comparison requires more nuance than simply comparing cost per core or price performance against a canned benchmark. 

At the same time, an accurate comparison is incredibly important. The value of an effective data platform is incalculable: it allows enterprise users to connect with customers, optimize efficiency, improve competitive advantage, and generally thrive in today’s complex and connected markets. Conversely, the downsides of an ineffective data platform are also big and dramatic: lost customers, lower revenue, damaged reputation, decreased ability to compete, heavy fines for bungling regulatory compliance, and increasingly the potential to go out of business because of uninformed decisions leading to bad business outcomes. 

So, it’s essential but challenging to get the value-for-money analysis right. Let’s spend some time looking at the factors in this analysis and why we think Snowflake might be more cost efficient than you previously believed.

Historically, it’s been a huge challenge to manage costs of complex technology solutions. Back in the on-premises days, you had to add up bills from many vendors to know how much a given solution actually cost; think servers, storage arrays, network gear, operating systems, virtualization software, app software, and everything else you may have bought along the way to make it all work. That’s in addition to the costs for the people it took to shovel coal into the infrastructure fires, the data center itself, implementation, migration, planning, and maintenance. Getting a clear view of how much a transaction actually cost the organization was essentially impossible.

Cloud usage has made it easier to know how much it costs to deliver technology solutions, potentially acting as a de facto chargeback model by giving the user a single bill for their usage of services. But all cloud solutions aren’t created equal when it comes to cost manageability. Many options on the market today suffer from some combination of the following issues that make it hard for users to measure and control costs:

• Limited visibility: The idea of one bill for the whole solution is great, but in reality users often have to combine multiple separate services, each with separate bills or many separate lines on one extremely complex bill. What’s more, hidden costs and price variability can make it hard to know what’s going on or how much it costs.
• Low elasticity: Not every cloud solution is fully elastic, with long spin-up/spin-down times, and limited ability to increase, decrease, or pause and resume instantly. Some solutions are elastic themselves, but the way they are billed makes it more like on-premises usage where you have to predict and pay for usage before it happens, often leading to waste.
• High complexity: Solutions can initially seem cheap to use, but are hard to manage and maintain, often costing far more in the long run.

The Snowflake Data Cloud addresses these three issues as core design points of the service, giving customers a balance of automation, control, and technical benefits that in most cases lead to better economic outcomes than alternative options. Here’s how Snowflake addresses these issues and enables customers to derive maximum value from their data while minimizing total cost of ownership. 

You can’t manage what you can’t measure, leading many organizations to be unaware of the true cost and effectiveness of their solutions. Snowflake has far better cost visibility than other solutions, making it easy to understand and control costs, thanks to the following key capabilities:

• One bill for the whole solution: When you build your own data platform from components, or use a platform that you have to manage, you often get separate bills, especially for infrastructure versus software. With Snowflake you get one bill that shows you all the costs of the solution in one place, which makes it much easier to understand cost impacts of usage.
• Easy attribution of costs: Snowflake customers use object tags to group usage of resources by departments, projects, or individuals. This provides a clearer picture of who’s using what, and makes it much easier to prioritize and restrict usage depending on cost and business value. 
• Simple pricing model: Some data platforms have incredibly complex pricing models, with a dizzying list of different rates depending on factors such as which cloud you run on, which instances or engines you choose, programming languages used, high availability options, support plans, or payment models. Snowflake has a simple pricing model for compute consumption:
  [usage time, per second] multiplied by [pricing rate for the selected Snowflake Edition]

The simplicity of this model, regardless of deployment details, makes it easier for customers to predict and plan their usage, and work within their budgets to get the work they need done.

• No hidden charges: Other data platforms have lots of upcharges for accelerators, infrastructure choices, increased quotas, enterprise support, feature-specific charges, and other charges that mean if you’re not an expert who’s constantly vigilant, you might pay far more than you expect. Snowflake eliminates these hidden charges, sticking to a simple pricing model and making it easier for customers to know what they are getting and what they are paying for.
• Powerful real-time cost controls: Organizations moving from on-premises or inelastic cloud data solutions go from compute scarcity to abundance, and from rigid usage models to full flexibility. Snowflake’s flexible deployment model comes with a great deal of power as well as responsibility to control costs; the alerts and limits built into the Snowflake Data Cloud allows organizations to closely monitor usage down to the individual virtual warehouse level, and to optionally set hard caps on usage at a granular level. Other data platforms may have rough-grained attribution or after-the-fact reporting, but Snowflake users have the power to prioritize work, limit spending on less critical workloads, and accelerate the jobs that really matter.

Data workloads are often highly variable—busy times like month end, holiday seasons, and product launches lead to spikes, while slack times see far less usage. Many cloud data platforms suffer from limited elasticity of the solution itself, or from rigid pricing models that mean even if the customer uses less, they pay the same amount. Snowflake addresses this with:

• Instant pause/resume, scale up/down: Elasticity doesn’t work when there’s a delay in increasing scale or resuming paused instances. The delay leads to queued queries and frustrated users, and generally forces customers to provision more than they need to avoid these delays. Snowflake pays for standby instances that are ready to instantly join a customer’s cluster in every cloud region Snowflake operates in, meaning that customers can carry the exact amount of capacity they need at any given moment, and be confident that additional resources will be there “just in time,” enabling true elasticity.
• Autoscaling to increase the benefit of instant elasticity: A truly elastic solution is a great thing, but users don’t always know when more capacity is needed. With Snowflake multi-cluster warehouses, customers can set minimum and maximum levels of scaling, and allow the system to increase and decrease capacity as needed based on usage levels, further increasing the benefits of an elastic platform.
• Elastic, usage-based billing: Some cloud data platforms are technically elastic, but use a rigid billing model where customers have to predict and pay for estimated usage, regardless of whether they use that much or less, or require customers to pay for a hot pool of available capacity to gain elasticity of usage. Who cares if a system is elastic if you have to pay for more than you use? Snowflake charges for actual usage to the second, which combines with the technical elasticity features to support highly efficient usage.

Data is any company’s most valuable asset, and getting value from it requires a powerful platform. Harnessing these capabilities is nearly impossible with a highly complex solution that requires the user to effectively build and manage it. The Snowflake Data Cloud is built to reduce complexity at every turn, making it easier for customers to get good value from their platform. Some of the key aspects of simplicity delivered by Snowflake include:

• Fully managed service: Unlike a build-it-yourself or customer-managed solution, Snowflake is a fully managed service. This means that customers don’t have to worry about upgrades, maintenance, network configuration, storage management, security configurations, penetration and vulnerability testing, high availability, and all the other tasks that need to happen for a data platform to stay healthy and productive. This dramatically reduces the staff burden, and therefore the cost of the Snowflake Data Cloud.
• Elimination of silos: Scalability limits and functional divisions often mean that data lives in separate repositories that are hard to integrate. Even if you have a single platform today, the inevitable growth of data may well push the boundaries of platforms with scale limitations. Snowflake can support an essentially infinite-sized data repository, with an essentially unlimited number of compute clusters that can attach to it. This massive scalability in a single platform allows customers to avoid and eliminate silos, putting all their data at their fingertips and making it much easier to derive value from their data.
• Ubiquitous SQL for administration: Snowflake offers a near-zero administration model, with high levels of automation and few knobs to turn. What work there is to do to keep the system working optimally is all handled in SQL, which is a ubiquitous language and skill set, making it easier to find employees and potentially avoid the need to hire high-cost experts who know alternative languages and methodologies that are less widely known.
• It just works: Data platforms vary tremendously in terms of how they work and how easy they are to manage. Regardless of the features, capabilities, and performance of a given system, if it’s too complicated to both do the job now, and evolve over time to meet changing requirements, then what good is it? A common thread in the feedback we hear from many of our thousands of happy customers is that “Snowflake just works.” This intangible may be one of the biggest factors when it comes to economic outcomes—the elegance of the highly automated Snowflake Data Cloud makes it work optimally to solve customer problems and deliver value.
• Easy to get started: Snowflake customers are consistently amazed by how quickly they can get to work on their data. Setup, customization, and data migration often take months on more complex platforms, but because Snowflake is so simple and automated, getting started is easy and fast. Customers are often up and running in a matter of days, allowing them to launch services faster, experiment more, and start to get value from their data right away when they would still be working on configuring alternative solutions.
• Continuous Innovation: Because Snowflake is a fully managed service, all of our customers run on the same platform, and all of them benefit when we improve that platform. We continuously evaluate the impact of and implement behind-the-scenes new instance types, new chips, new features, and other improvements to the Snowflake Data Cloud, empowering customers to focus on their data—while we focus on improving the performance and usability of the platform, which translates to lower costs for customers with no effort on their part. Our continuous innovation has yielded a more than 20% reduction in the average cost of warehouse queries for customers over the last 3 years.

The Snowflake Data Cloud is a comprehensive platform for enterprise data that gives powerful tools to users. But this power does not come with a higher price tag. Because of the visibility, elasticity, and simplicity we’ve built into the platform, customers get these powerful capabilities at a lower total cost than they would with other platforms. We encourage you to give it a try for whatever data use cases you have, and would be happy to work with you to understand how this compares to what you do today in terms of operational and cost efficiency.