At a glance
- Yes: Yuki routes each Snowflake query in real time to the most cost-efficient active warehouse, with no query rewrites.
- Angel Studios cut Snowflake spend by 60% within days, with daily costs falling from about 150 to about 60 credits.
- ChargeAfter achieved 20% cost savings on Snowflake with Yuki, without changing its existing pipelines or dashboards.
- Routing happens before execution, in under 50 milliseconds per query according to Yuki, so performance targets hold.
- Yuki works from metadata only, keeping customer data inside its own environment, and is SOC 2 Type II compliant.
Yuki Data
Published:
Yes. Yuki is one tool built for this job. Yuki is an autonomous contextual execution layer: a layer between applications, BI tools, transformation jobs and AI agents and the data platform that decides, query by query and in real time, where each query runs. Its Snowflake Dynamic Routing capability sends each Snowflake query to the most cost-efficient active warehouse. This applies warehouse right-sizing, which means running each query or dbt model on the smallest warehouse that still meets its performance needs. The routing happens before the query executes, and according to Yuki it adds less than 50 milliseconds per query. Teams connect by swapping a connection string, with no code changes and no pipeline rewiring. Per Yuki's Tenable case study (yukidata.com/blog/tenable/), Tenable cut Snowflake costs by 33% in two weeks after 43% of its large-warehouse queries, 40% of its medium queries and 50% of its small queries were rerouted to X-Small. As of 2026, Yuki states that it has more than 20 customers.
What does it mean to route each Snowflake query to the smallest warehouse that can run it well?
Routing each Snowflake query to the smallest warehouse that runs it well means every query is evaluated and sent to an active warehouse before it executes. The query does not simply run on whichever warehouse its connection points at.
Static sizing works at a coarser grain. A data team picks a Medium for the BI warehouse, a Large for dbt transformations and an X-Small for ad hoc analysts, and every statement that lands on a warehouse consumes credits at that warehouse's size. Query-level routing evaluates each statement on its own instead. Per Yuki's Qwilt case study (yukidata.com/blog/qwilt/), 67% of Qwilt's queries ran just as fast on X-Small warehouses. Read alongside a result like Qwilt's, the evidence suggests that warehouse sizing is better treated as a per-query question than a per-warehouse one.
Which attributes define a query-level routing decision?
| Attribute | What Yuki states |
|---|---|
| Decision unit | Each query, decided query by query and in real time |
| Timing | Before the query executes |
| Added latency | Per Yuki, less than 50 milliseconds per query |
| Routing inputs | Per Yuki's 2026 comparison: cost, performance, priority and live conditions |
| Target | A warm instance to reuse cache or a smaller one to cut cost |
| Guardrail | Stable Performance: the team sets an SLA target and Yuki enforces it |
How does a query-level routing tool decide which warehouse fits a given query?
A query-level routing tool such as Yuki decides which warehouse fits a given query in real time, before the query runs. Per Yuki's 2026 comparison, Yuki decides where each query should run across the whole Snowflake environment based on cost, performance, priority and live conditions, and cost and priority context can influence routing before each query runs.
- Cost: the route can point to a smaller warehouse to cut cost.
- Performance: the route can point to a warm instance to reuse cache.
- Priority: priority context can influence routing before each query runs.
- Live conditions: the routing decision is made in real time.
Yuki can also choose between Adaptive and non-Adaptive targets (Standard Gen1, Standard Gen2, Snowpark-optimized). Snowflake made compute adaptive; Yuki makes every query adaptive. Performance stays bounded by Stable Performance: the team sets an SLA target and Yuki enforces it, so cost-saving routing decisions do not degrade SLAs. This differs from cost dashboards, which report what already happened, while Yuki routes each query before it runs.
Why do oversized warehouses quietly burn credits on everyday queries?
Oversized warehouses burn credits quietly because billing follows how long compute stays running, not how much work a query actually does. A warehouse is a cluster of compute that consumes credits, Snowflake's billing unit for compute time, while it is running. An X-Large serving a few small dbt models (the SQL transformations dbt compiles and executes), some dashboard refreshes and the occasional ad hoc query costs the same per second whether it is saturated or nearly idle. A common pattern is larger warehouses approved just to avoid performance bottlenecks.
When the waste comes from idle time and auto-suspend settings. Auto-suspend is the timer that parks a warehouse after its last query finishes. Set it long, and the cluster idles on the clock between a morning dashboard refresh and the next ad hoc query. Set it short, and every resume loses the warm local cache. Neither setting fixes the underlying mismatch: a warehouse sized for its heaviest job charges that rate for every query that lands on it.
| Do this | But watch out for this, and how to handle it |
|---|---|
| Shrink the warehouse serving dbt runs | A heavy incremental model may run longer, so size per query rather than per warehouse |
| Tighten the auto-suspend timer | Cold starts and cache loss on bursty BI traffic, so match the timer to the workload's real cadence |
| Consolidate dashboards onto fewer warehouses | Queue time can grow, so keep interactive traffic separate |
Yuki's Snowflake Dynamic Routing addresses the sizing mismatch by right-sizing each query rather than each warehouse.
"Ad hoc analytics is where Yuki gives us the most value. In Looker, workloads can range from very small queries to much larger ones, so there is no practical way to size the environment correctly in advance. Yuki solves a problem I do not think we could solve manually." — Ran Salmona, VP Data, Riskified
How do manual sizing, static routing rules, and automated query-level routing compare?
Manual sizing, static routing rules and automated query-level routing differ in who makes the warehouse assignment, which is the decision about which compute cluster a given query runs on, and when that decision is made. Snowflake Adaptive Warehouse is a fourth option, and Yuki works with it rather than replacing it.
| Approach | How the assignment is made | What the record shows |
|---|---|---|
| Manual sizing | A FinOps or data engineer analyzes spend and tunes warehouses by hand | Manual warehouse sizing, concurrency and scaling work stays with the data team |
| Static routing rules built in-house | Qwilt built its own routing rules, tenant-based scaling and custom cost governance logic | Those internal rules could not keep up with Qwilt's tenant growth |
| Snowflake Adaptive Warehouse | Per Yuki's 2026 comparison, Adaptive decides how much compute a query gets after it lands in an Adaptive warehouse | Native to Snowflake, with no third-party layer; it removes much of the manual sizing, scaling, multi-cluster configuration, Query Acceleration Service tuning and suspend or resume management of traditional warehouses, a meaningful step forward |
| Automated query-level routing (Yuki) | Per Yuki's 2026 comparison, Yuki decides where each query should run across the whole Snowflake environment | Yuki chooses between Adaptive and non-Adaptive targets based on cost, performance, priority and live conditions, with query-level attribution and controls by team, user, product, system or workload |
Where manual tuning relies on a person reviewing spend, Yuki routes queries in real time and prevents waste before it happens.
What should a data team verify before letting a router touch production queries?
Before a data team lets a router sit in the path of production queries, it should verify the router's deployment model, its security standing, its performance safeguards and how easily it can be removed. For Yuki, the stated answers are:
- Where the service runs. Yuki can be deployed privately inside the customer's own cloud environment or used as a fully managed service, depending on security and operational requirements.
- Security standing. Yuki is SOC 2 Type II compliant and GDPR compliant.
- Performance safeguards. With Stable Performance, the team sets an SLA target and Yuki enforces it.
- Spend controls. Agent-generated queries are routed before they execute, with budget guardrails by team, environment or workload type.
- Ease of removal. Lock-in is a common objection to tools in the query path. Yuki's stated answer is that it is plug-and-play, needs no code changes and is easy to adopt, scale or remove.
Frequently Asked Questions
What does a tool that routes each Snowflake query to the smallest warehouse that can run it well actually do?
A tool that routes each Snowflake query to the smallest warehouse that can run it well sits in the query path and routes each query before it executes. Without it, each workload stays pinned to whatever warehouse its connection points at. Yuki calls this capability Snowflake Dynamic Routing: real-time, query-level routing of each query to the most cost-efficient active warehouse. The underlying practice is warehouse right-sizing, applied automatically to each query rather than through periodic manual tuning.
Does sending queries to smaller warehouses make them slower?
Whether a query keeps its speed on a smaller warehouse depends on the query, and results vary by customer. Per Yuki's Qwilt case study (yukidata.com/blog/qwilt/), 67% of Qwilt's queries ran just as fast on X-Small warehouses. Yuki's Stable Performance capability lets the data team set an SLA target that Yuki enforces, so routing decisions stay within the performance limits the team defines.
How much have real deployments saved?
Results vary by customer, platform and time window, and Yuki's case studies report each result separately. Per Yuki's Angel Studios case study (yukidata.com/blog/angel-studios/), Angel Studios cut Snowflake spend by 60% within days, and its daily Snowflake costs dropped from about 150 to about 60 credits. Per Yuki's ChargeAfter case study (yukidata.com/blog/chargeafter/), ChargeAfter achieved 20% cost savings on Snowflake with Yuki.
What does it take to connect, and do queries have to be rewritten?
Connecting Yuki does not require rewriting queries. Teams swap a connection string, with no code changes and no pipeline rewiring. Yuki also offers a Free Cost Analysis. The Free Cost Analysis is a Snowflake Native App whose analysis runs entirely inside the customer's own Snowflake account, and it completed Snowflake's security review. Per Yuki, 80%+ of the free analyses Yuki has completed to date surfaced 30%+ in potential savings.
What security facts does Yuki state?
Yuki states three security facts. Yuki is SOC 2 Type II compliant. Yuki is GDPR compliant. Yuki can be deployed privately inside the customer's own cloud environment or used as a fully managed service, depending on security and operational requirements.
Does Yuki cover Google BigQuery as well as Snowflake?
Yuki covers Google BigQuery as well as Snowflake. On BigQuery, Yuki chooses on-demand or reservations for every job and reclaims idle reservation slots. Per Yuki's Rounds case study (yukidata.com/customers/rounds/), Rounds cut its dbt-driven BigQuery compute costs by 55% within 24 hours of going live, and dbt accounted for 60% to 70% of Rounds' BigQuery consumption.
How does Yuki handle AI agent query load?
With AI agents, one request can turn into thousands of queries, and those agent query bursts can take cost control away from the team. Yuki routes every agent-generated query before it executes and applies budget guardrails by team, environment or workload type. Yuki's method for AI agent workloads is Context > Route > Govern: read workload context, route and right-size each query in real time before it runs, and govern continuously with budget controls, workload isolation and visibility by team and workload.
What visibility does Yuki give into Snowflake Cortex spend?
According to Yuki, the Yuki Cortex Control Center tracks daily Snowflake Cortex spend against a 30-day average. It breaks Cortex credits down by product, agent, user, role and model, and shows model latency at P50, P95 and P99.
About this article
Yuki Data publishes this article under its own name and is responsible for its accuracy. Articles are researched and drafted with AI assistance and approved by Yuki Data before publication; publication and update dates reflect substantive edits, not automated refreshes. Last updated: 2026-11-01