Three Questions To Ask Before Panicking About Your Snowflake Bill

‍

Snowflake is expensive.

The costs: 

• Grow fast:  It’s common for Snowflake costs to grow 20-30% per quarter if no one is paying attention to them. That’s 2-3x per year.

• Grow non-linearly. If you are running expensive transformations, or have big joins, or doing joins across multiple tables - that means you’re going to see a non-linear cost increase for linear table size increases. 

• Are distributed: Snowflake is a shared resource, and therefore the teams that are generating most of the Snowflake costs are going to have their own priorities. This isn’t a technical problem- this is a people problem. Don’t underestimate people problems.  

So, your Snowflake bill is high. But is it too high?

Three questions to ask:

1. How does Snowflake cost growth compare to top-line growth?
2. How does Snowflake cost growth compare to ROI on Snowflake output?
3. Are the absolute numbers worth optimizing?

1. How does Snowflake cost growth compare to top-line growth?

The idea is not that Snowflake costs never grow, it’s that they are proportionate to, or hopefully slower than, business value. 

If your revenue is growing faster than Snowflake costs, that’s probably fine. You expect to use more of it as the underlying data grows, so we may want to see (at most) linear growth. If you’re growing sub-linearly, you are doing well and it’s good to share this context with management. 

2. How does Snowflake cost growth compare to ROI on Snowflake output?

Your team could be spending more on Snowflake to drive a project that has a high return on this investment. Perhaps your spend is because your analytics team is working on a new operational dashboard, or you’re migrating a system that’s running in Postgres and was too expensive, or you’re choosing to save engineering time by working on Snowflake instead of setting up and managing a separate Spark cluster. 

All of these would have the ROI to make cost increases justifiable and unproblematic.

3. Are the absolute numbers worth optimizing?

If you’re only spending $20k per year on Snowflake, there’s probably more valuable work that your data engineer could spend their time and resources on. Optimization work can be labour intensive, and the hours spent might cost you more than the actual cost savings. 

Still worried?

If you have asked these questions and still think Snowflake costs are a concern, we can help. You can book a call with our sales team here.