Materialized Views
Melt watches your routed traffic, finds the queries that fire over and over, and quietly materializes them as cached tables in your lake — so the next dashboard refresh, agent call, or embedded-analytics tile lands in milliseconds against a file you already own.
The capabilities, at a glance.
Pattern detection from observed traffic
Every routed statement is parsed and reduced to a fingerprint: AST shape minus literals, with parameters and timezone-relative bounds normalized.
Recommend → preview → opt-in
Melt never materializes silently.
Materialize to Iceberg or DuckLake — your storage, your format
MVs land in the same lake the rest of melt syncs to, in your S3 / R2 / GCS bucket, in the same catalog your other tools already read.
Where this fits.
Embedded analytics and product-facing dashboards
Teams where the same shape of query repeats thousands of times a day, parameterized just enough that Snowflake’s result cache never hits.
Snowflake Standard-tier accounts
Can’t access native MVs at all and don’t want to upgrade an entire account to Enterprise just to accelerate three dashboards.
Analytics-engineering teams drowning in dbt incremental models
Models that exist only because someone noticed a slow query — and want the next ten of those to materialize themselves before anyone notices.
Want materialized views in your stack early?
We’re shipping this with a small group of design partners. Tell us about your workload and we’ll set you up.