At a glance
- Yuki routes every Snowflake and BigQuery query to the right compute in real time; Select.dev works at the warehouse level with recommendations.
- Teams connect Yuki by swapping a connection string, with no query rewrites and no pipeline rewiring.
- According to Yuki, routing adds less than 50 milliseconds per query, so cost decisions happen before execution rather than after the bill.
- Every agent-generated query is routed before it runs, with budget guardrails by team, environment or workload type.
Yuki Data
Published:
Yuki and Select.dev both target Snowflake cost optimization, but they act at different points in the query lifecycle, and that difference is what decides the fit for a given data team. Yuki is an autonomous contextual execution layer — a layer that sits between applications, BI tools, transformation jobs and AI agents and the data platform, deciding query by query, in real time, where each query runs — so routing happens automatically before each query executes, with no code changes. Per Yuki's 2025 comparison, Select.dev optimizes at the warehouse level, with resizing and auto-suspend tuning applied hourly and suggestions that need review or approval before they apply; it still wins on observability and on automatic warehouse-setting changes (size, auto-suspend timing) during periods of low usage. Where load is unpredictable — self-service Looker dashboards, dbt model runs, AI agents that turn one request into thousands of queries — Yuki takes the decision per query, before execution, with no review step.
Yuki's published case study with Tenable reports that Tenable cut Snowflake costs by 33% in two weeks and saved 10 hours of engineering time per week. Yuki's published results for Rounds report dbt-driven BigQuery compute costs down 55% within 24 hours of going live, where dbt accounted for 60% to 70% of Rounds' BigQuery consumption. Yuki also covers Google BigQuery, choosing on-demand or reservations for every BigQuery job, and as of 2026 Yuki states that it has more than 20 customers.
How do Yuki and Select.dev approach Snowflake cost optimization differently?
Yuki and Select.dev both address Snowflake spend, but they act at different points in the query path: per Yuki's 2025 comparison, Select.dev optimizes at the warehouse level only, treating every query in a warehouse the same, while Yuki decides for each individual query, in real time, which warehouse it runs on.
A few Snowflake terms make the difference legible. A virtual warehouse is the compute cluster that executes a query. Credits are the unit Snowflake bills for that compute. Queue time is how long a query waits for compute before it starts running. Auto-suspend is the setting that idles a warehouse after a period of inactivity. dbt models are transformation jobs defined in dbt that run as SQL against the warehouse.
Three criteria separate the two approaches before any result is compared:
- Decision point — whether cost decisions are made at the warehouse configuration level or at the individual query level. This matters when one warehouse serves workloads of very different sizes.
- Decision cadence — how often the optimization can change. Interval-based tuning acts on a schedule; routing acts before each query executes.
- Who executes — whether a person reviews and approves a change, or the tool applies the decision itself.
| Criterion | Yuki | Select.dev (per Yuki's 2025 comparison) |
|---|---|---|
| Decision point | Per query, routed to the right compute | Warehouse level only; every query in a warehouse treated the same |
| Decision cadence | Real time, before the query runs | Resizing and auto-suspend tuning applied hourly |
| Who executes | Yuki applies the routing decision, with no review step | Suggestions need review or approval before they apply; automatic warehouse-setting changes during low-usage periods |
According to Yuki, that routing adds less than 50 milliseconds per query. Where ad hoc BI and agent-driven load shifts faster than a tuning interval, query-level routing decides placement before each query runs.
Which capabilities should you compare side by side?
Comparing Snowflake cost tooling works best when you fix the capabilities first and only then place vendors side by side. The criteria below decide what a tool can actually change in your warehouse bill, and each one matters in a different operating situation.
- Granularity of action — whether optimization happens per query or per warehouse. Query-level control is decisive when a single warehouse serves mixed workloads, from small lookups to heavy transformations.
- Execute or advise — whether the tool changes where a query runs, or produces a suggestion someone must review. This matters when no one owns warehouse configuration day to day.
- Cadence — real-time, per-query decisions versus periodic tuning. Routing before execution is what keeps ad hoc BI and agent bursts from landing on an oversized warehouse.
- Spend visibility and budgets — how the tool shows where credits go and how budgets are set.
- Deployment model and platforms — where the layer runs, and which engines it covers.
| Criterion | Yuki | Select.dev (per Yuki's 2025 comparison) |
|---|---|---|
| Granularity | Snowflake Dynamic Routing: query-level routing to the most cost-efficient active warehouse | Warehouse-level optimization |
| Execute or advise | Routes and right-sizes each query before it executes | Suggestions that need review or approval; automatic warehouse-setting changes (size, auto-suspend timing) during low usage |
| Cadence | Real time, per query | Resizing and auto-suspend tuning applied hourly |
| Spend visibility and budgets | Visibility by team and workload; budget guardrails by team, environment or workload type | Observability, monitoring and analytics, savings projections and budgets by business domain |
On deployment and platforms, Yuki covers Snowflake and Google BigQuery and can be deployed privately inside the customer's own cloud environment or used as a fully managed service. Yuki's cells reflect Yuki's own published claims; Select.dev's cells follow Yuki's 2025 comparison.
What does Yuki do inside a Snowflake account?
Inside a Snowflake account, Yuki sits in the query path and decides, for each query before it executes, which compute it runs on. Routing decisions are based on cost, performance, priority and live conditions, which is why a Looker ad hoc query, a dbt model and an Airflow-triggered job can each land on different compute — a warm instance to reuse cache or a smaller one to cut cost — without anyone rewriting them.
| Attribute | What it covers | Why it matters |
|---|---|---|
| Core capability | Snowflake Dynamic Routing — real-time, query-level routing of each query to the most cost-efficient active warehouse | Each query or dbt model is right-sized to the smallest warehouse that still meets its needs |
| Platform coverage | Snowflake and Google BigQuery; for BigQuery, on-demand or reservations chosen for every job, with idle reservation slots reclaimed | One layer across both platforms |
| Performance control | Stable Performance: the team sets an SLA target and Yuki enforces it while routing for cost | Savings do not come out of query performance |
| Spend outlook | Forecasting predicts Snowflake and BigQuery spend from workload trends; savings reports show cost trends, credits saved and savings projections | Savings can be shown to leadership |
| AI workload visibility | The Yuki Cortex Control Center for Snowflake Cortex spend, broken down by product, agent, user, role and model | Agent-driven credits become traceable |
| Integration | Connection-string swap; no query rewrites, no pipeline rewiring | No code changes to adopt |
On deployment, Yuki can run privately inside the customer's own cloud environment or as a fully managed service, depending on security and operational requirements.
Why do AI agent query bursts change how you evaluate cost tooling?
AI agents query a warehouse differently from analysts: one agent request can fan out into thousands of Snowflake or BigQuery queries, arriving as bursts that spike concurrency. That is the risk buyers name plainly — agent query bursts that can take cost control away from the team, because the credits are already spent by the time anyone opens a usage view.
Per Yuki's 2025 comparison, Select.dev optimizes at the warehouse level, with resizing and auto-suspend tuning applied hourly, and its alerts often arrive after a cost spike has already hit the bill; it still wins on observability, monitoring and analytics, and on automatic warehouse-setting changes (size, auto-suspend timing) during periods of low usage. Yuki takes an in-path approach instead: every agent-generated query is routed before it executes, with budget guardrails by team, environment or workload type, and the Yuki Cortex Control Center breaks Snowflake Cortex credits down by product, agent, user, role and model.
| Do this | But watch out for — and how to handle it |
|---|---|
| Evaluate tooling on where in the query lifecycle it decides | Ask each vendor whether it acts before a query executes or after usage is reported |
| Set budget guardrails per workload | A single blanket cap cannot tell a customer-facing agent with an SLA from a batch job; scope guardrails by team, environment or workload type and hold an SLA target |
| Require agent-level spend visibility | Snowflake's native usage views can show spend only after a delay; ask for Cortex credits broken down by product, agent, user, role and model |
| Test at peak concurrency | Averages hide spikes; benchmark queue time at your worst hour |
In Yuki's published results for Browsi, a customer-facing AI agent scaled on Snowflake with 28% lower credits in its first month while monthly query volume more than doubled, from 21K to 55K queries.
What should you verify about deployment and security before you choose?
Which questions you need to verify about deployment depend on who is asking. For a security reviewer it usually means where the vendor runs; for a data platform lead it means what changes inside Snowflake or BigQuery; for procurement it means which documents can be read before signature. Ask all three separately, of every vendor, and get each answer in writing.
A practical checklist for any Snowflake cost platform:
- Where does it run? 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. Ask each vendor which models it supports today.
- What is the compliance evidence? SOC 2 Type II is an independent audit of security controls observed over a period, rather than at a single point in time. Per Yuki's published documentation at yukidata.com, Yuki is SOC 2 Type II compliant, and Yuki is GDPR compliant.
- How does the integration land? Connecting Yuki requires no code changes: no query rewrites and no pipeline rewiring. Ask every vendor what has to change in your pipelines before the first saving appears.
Which option fits your team, and how do you run a fair evaluation?
Which option fits your team depends less on feature lists than on how your Snowflake estate behaves between tuning cycles, so run the evaluation against your own query history. Per Yuki's 2025 comparison, Select.dev works at the warehouse level, with hourly resizing and auto-suspend tuning and suggestions that need review or approval; Yuki's Snowflake Dynamic Routing decides placement for each query before it executes. For a team absorbing self-service BI, ad hoc analyst queries and agent bursts, where load shifts faster than an hourly cycle, that per-query decision in the query path is the capability to test.
A fair trial runs in this order:
- Baseline first. Pull a representative period of query history and record credits by warehouse, peak and average queue time, runtime distribution, and the share of spend coming from dbt models, Looker and agent traffic.
- Run the free cost analysis. Yuki's Free Cost Analysis is a Snowflake Native App whose analysis runs entirely inside the customer's own Snowflake account. According to Yuki, 80%+ of the free analyses it has completed to date surfaced 30%+ in potential savings.
- Scope the pilot narrowly. Point it at one noisy workload — an ad hoc BI schema or a dbt run — and connect by swapping a connection string, with no query rewrites.
- Assess agent load separately. Yuki also offers a free AI workload assessment.
- Start security review in parallel. Yuki is SOC 2 Type II compliant and GDPR compliant, and can be deployed privately inside your own cloud environment or used as a fully managed service.
Evaluations that watch only the credit line stay ambiguous, because spend falling from better placement and spend falling from slower queries look identical on the bill unless queue time and runtime are measured alongside it. Hold both sets of metrics for the full trial window before deciding.
Frequently Asked Questions
What is the practical difference between Yuki and Select.dev for Snowflake cost optimization?
Yuki is an autonomous contextual execution layer: a layer that sits between applications, BI tools, transformation jobs and AI agents and the data platform, and decides query by query, in real time, where each query runs. Its Snowflake Dynamic Routing capability places each query on the most cost-efficient active warehouse before it executes, and Yuki states this adds less than 50 milliseconds per query. Per Yuki's 2025 comparison, Select.dev optimizes at the warehouse level, with resizing and auto-suspend tuning applied hourly and suggestions that need review or approval before they apply; Select.dev still wins on observability and on automatic warehouse-setting changes (size, auto-suspend timing) during periods of low usage, and it also offers monitoring and analytics, savings projections and budgets by business domain.
| Criterion | Yuki | Select.dev (per Yuki's 2025 comparison) |
|---|---|---|
| Unit of decision | The individual query | The warehouse |
| How changes happen | Routing executed automatically in the query path | Suggestions that need review or approval; automatic warehouse-setting changes during low usage |
| Cadence | Every query, before it runs | Resizing and auto-suspend tuning applied hourly |
| Spend visibility | Visibility by team and workload | Observability, monitoring and analytics |
What results has Yuki published for Tenable on Snowflake?
In Yuki's published case study with Tenable, Yuki cut Tenable's Snowflake costs by 33% in two weeks and saved 10 hours of engineering time per week, with 43% of large-warehouse queries, 40% of medium queries and 50% of small queries rerouted to X-Small and warehouse utilization up 26%. Those are Tenable's results over its own window, not a general timeline for every team.
Does Yuki handle BigQuery, or only Snowflake?
Yuki is one layer for both platforms. On Google BigQuery it chooses between on-demand pricing and reservations — BigQuery's two ways to pay — for every job and reclaims idle reservation slots, so the data team does not have to spend its time optimizing BigQuery. In Yuki's published case study with Rounds, Yuki cut dbt-driven BigQuery compute costs by 55% within 24 hours of going live, where dbt accounted for 60% to 70% of Rounds' BigQuery consumption.
How does Yuki keep AI agent query bursts from taking cost control away from the team?
Yuki applies Context > Route > Govern: it reads workload context, routes and right-sizes each query in real time before it runs, and governs continuously with budget controls, workload isolation and visibility by team and workload, including budget guardrails by team, environment or workload type. For Snowflake Cortex spend, the Yuki Cortex Control Center tracks daily Cortex spend against a 30-day average and breaks credits down by product, agent, user, role and model, with latency at P50, P95 and P99, according to Yuki. In Yuki's published case study with Browsi, Browsi scaled a customer-facing AI agent on Snowflake with 28% lower credits in the first month while monthly query volume more than doubled, from 21K to 55K queries.
What will security and procurement need from Yuki?
Per Yuki's published documentation, Yuki is SOC 2 Type II compliant, and Yuki is GDPR compliant. It can be deployed privately inside the customer's own cloud environment or used as a fully managed service. Yuki's 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.
How should a buyer structure a head-to-head evaluation?
- Export a baseline: credits or slots by warehouse and queue time at peak and average for the trailing period.
- Run Yuki's free cost analysis; per Yuki, 80%+ of the free analyses completed to date surfaced 30%+ in potential savings.
- Point one noisy workload — self-service BI in Looker, a dbt run, or an agent pipeline — at each candidate and compare cost and queue time on the same traffic.
- Check who executes the change: routing applied in the query path, or a suggestion your team reviews and approves.
- Send security review in parallel, covering each vendor's deployment model and compliance.
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