Data warehouse

Snowflake — contract review

Snowflake contract review: capacity vs on-demand pricing, the cost overrun risk, and the negotiation points that matter.

Analytics dashboard on a laptop screen
Photo by Lukas on Pexels

Snowflake is the dominant cloud data-warehouse platform, with usage-based pricing measured in credits consumed per second of compute. The standard commercial model is a Capacity Commitment — pre-purchased credits at a tiered discount over On-Demand pricing — with a 1- or 3-year term.

Snowflake contracts are notorious for credit-burn-rate uncertainty: poorly-tuned queries or runaway dashboards can consume hundreds of credits per day, leading to large unbudgeted overage charges. The standard MSA contains specific clauses to scrutinise: usage-alert commitments, the price per credit at overage, and the rollover policy for unused committed credits.

POCsheet's AI evaluates Snowflake proposals against typical SaaS benchmarks and the specific risk patterns of usage-based pricing.

Typical contract terms

Standard contract term
1, 2 or 3 years
Pricing model
Pre-purchased credits + on-demand overage
SLA
99.9% monthly uptime for the warehouse layer
Credit rollover
Typically NOT rollover — use-it-or-lose-it; negotiate this aggressively
Overage pricing
List price per credit (significantly higher than committed price)

Common red flags

  • Blocker

    No credit rollover

    Unused credits at the end of the contract period are forfeit. This punishes conservative usage and rewards burn-fast behaviour.

  • Strong

    Overage pricing not capped

    Once committed credits are exhausted, list pricing applies — often 30–50% higher than the committed rate. Negotiate a cap on overage rate.

  • Minor

    Resource Monitors not contractually required

    Snowflake provides usage governance tools but doesn't contractually require alerts before overage. Make it a contractual SLA.

Negotiation levers

  • Negotiate credit rollover (or partial rollover, e.g. 25% of unused commit) to the next term.
  • Cap the overage rate at no more than 1.25× the committed credit rate.
  • Require usage-alert SLAs (notify finance / engineering when crossing 75% / 90% / 100% of monthly commit).
  • Lock in price-per-credit for the full multi-year term, not just year one.

Alternatives to consider

  • BigQuery

    Per-query pricing — more predictable for ad-hoc analytics, less so for high-volume pipelines.

  • Databricks

    Stronger for ML/streaming workloads; comparable for pure SQL analytics.

  • Amazon Redshift

    More predictable provisioned pricing; less elastic at high concurrency.

Get YOUR custom Snowflake analysis

Drop your real Snowflake proposal — get red flags + negotiation language in 60 seconds

POCsheet reads your specific contract (not a generic one) and surfaces the deviations from your standard playbook, with verbatim source citations.

Try free analyzerSign up to compare 2–3 vendors side-by-side

More vendor reviews