How Riskified Cut Looker Compute Spend Approximately 50% With Yuki

By Perry Tapiero
October 8, 2026 | 5 min read
Riskified logo on a green desk banner

Riskified’s data team could size a warehouse for the average query or for the biggest one.

Neither option worked.

Most Looker queries needed very little compute. A smaller share needed a lot more. Size for the heavy ones, and you overspend on almost everything. Size for the average, and you slow down the requests analysts care about most.

So the default warehouse was oversized by design.

That is not a configuration mistake. It is what happens when one warehouse has to serve self-service analytics that change from one query to the next.

With Yuki, Riskified achieved approximately 50% savings across Looker workloads and approximately 20% savings across the broader mix of workloads routed through Yuki.

This is how they got there, and why the biggest gains showed up exactly where manual tuning was impossible.

About Riskified

Riskified empowers ecommerce growth by outsmarting risk. Fraud, policy abuse, and false declines don’t just threaten security. They also cut into revenue, stretch teams, and erode margins. As fraud grows faster and more sophisticated, merchants need more than a tool. They need a partner accountable for the outcome.

That’s the difference Riskified brings. Its AI-powered fraud and risk intelligence platform draws on machine learning trained across a global network which based on data available to us, includes approximately 50 publicly held companies worldwide. This gives merchants the accuracy to approve more good orders, block fraud, and curb abuse without adding friction for legitimate customers.

Riskified also stands behind its decisions. Merchants can shift chargeback liability and calibrate performance to match their own risk tolerance and business goals. The result is less uncertainty and a clearer path to profitable growth.

For merchants under pressure to grow revenue while managing risk, Riskified turns fraud prevention from a cost center into a driver of confident, sustainable growth.

Data Stack

  • Snowflake
  • Looker
  • dbt, Airflow
  • Databricks, Apache Iceberg

The Challenge: Self-Service Analytics Breaks Static Warehouse Sizing

Scheduled pipelines are the easy part of Snowflake cost control.

They run on a pattern. You can profile them, learn what they need, and match them to the right warehouse size.

Analysts do not work on a pattern.

For Riskified, three problems came together.

1. One warehouse size could not fit every query

Most queries in Looker required relatively little compute. A smaller share needed significantly more.

That spread creates a no-win choice. Sizing the warehouse for the largest queries led to overspending on most workloads. Sizing it for the average query risked slowing down the requests analysts cared about most.

So the team made the safe call, and the default warehouse was built oversized by design.

Safe, but expensive.

2. Ad hoc workloads were impossible to predict

Riskified could profile scheduled ETL jobs and match them to the right warehouse sizes.

Analyst workloads were different.

The data team had no reliable way to know which query would run next, how much data it would scan, or how much compute it would need. At that pace, manual resizing was not a realistic option.

You cannot hand-tune a workload you cannot see coming.

3. Analysts should not have to think about infrastructure

Riskified’s analysts focus on fraud prevention and risk signals, as well as related, time-sensitive business questions.

Warehouse sizing is not their job.

The team needed a way to control Snowflake costs without asking analysts to choose compute resources, change connection settings, or understand the infrastructure behind every query.

The Real Problem: You Cannot Right-Size the Average Query

Most Snowflake cost advice assumes you can study a workload and tune it.

That works for ETL.

It breaks down for self-service BI, where the next query might be tiny or might be heavy, and nobody knows until it arrives.

The requirement was simple to state and hard to execute:

  1. evaluate every query as it arrives
  2. send it to the warehouse size it actually needs
  3. do it without analysts changing anything
  4. do it without the data team reviewing queries or resizing warehouses by hand

That is not a dashboard problem.

It is a decision that has to be made in the query path, one query at a time.

The Solution: A Real-Time Routing Layer for Looker

Riskified deployed Yuki to make warehouse-sizing decisions automatically for Looker and for selected Snowflake workloads.

A real-time routing layer

Yuki sits in the query path and evaluates each workload as it arrives.

Instead of sending every Looker query to the same default warehouse, Yuki routes the query to the warehouse size that matches its actual compute requirements.

Right-size every query, not the average query

Lightweight queries can run on smaller, more efficient warehouses.

Heavier queries still receive the compute they need.

No more paying for the biggest query on every query.

Invisible optimization

Analysts continue using Looker as usual. They do not need to choose warehouse sizes or think about Snowflake infrastructure.

The data team does not need to continuously review queries or manually resize warehouses.

“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

That is the core of the story.

The hardest workload to optimize by hand became the workload where Yuki delivered the most value.

Results: Approximately 50% Savings Across Looker Workloads

Riskified achieved approximately 50% in savings across Looker workloads managed by Yuki.

Looker was the hardest workload to optimize manually, and that’s exactly where automated routing delivers the most value.

Most queries did not need the warehouse that they would have used by default. Yuki identified the lighter ones and routed them to more efficient compute automatically.

The savings were consistent from day one.

Broader Workload Savings: Approximately 20%

Across the broader mix of workloads routed through Yuki, including more predictable ETL traffic, Riskified observed approximately 20% savings.

The gap between the two numbers tells the story.

Predictable ETL can often be profiled and manually optimized.

Ad hoc analytics changes from one query to the next.

The more unpredictable the workload, the more value Riskified received from automated routing.

Hands-Off Optimization

For Ran Salmona, VP Data at Riskified, not needing to check the dashboard was a sign that the system was working.

That is what good infrastructure feels like.

You stop watching it because it keeps doing its job.

No Analyst Training Required

Riskified’s analysts continued to ask business questions while Yuki automatically handled the infrastructure decision.

There was no need to train users on warehouse sizing or ask them to decide how much compute a query might need.

No new tools. No new habits. No change management project.

A Clear Qualification Question

Riskified now uses a simple question when discussing Yuki with other Snowflake customers:

How much of your Snowflake traffic comes from analysts?

When most traffic comes from scheduled ETL, manual tuning may solve part of the problem.

When a large percentage comes from analysts, Looker, and self-service BI, static warehouse sizing becomes difficult to manage efficiently.

That is where Yuki delivers the most value.

“If most of your traffic comes from analysts, you need a solution like Yuki. There is no practical way to manually optimize a workload that changes from one query to the next.”
Ran Salmona, VP Data @ Riskified

As Yuki CEO Ido Arieli Noga put it:

“Riskified’s Looker environment shows why static warehouse sizing breaks down for self-service analytics. Most queries do not need large compute, but the few that do force teams to overprovision everything. Yuki makes that decision query by query, cutting waste without putting infrastructure decisions on analysts.”
Ido Arieli Noga, CEO and Co-Founder @ Yuki

Ido Arieli Noga, CEO and Co-Founder of Yuki

What This Unlocked for Riskified

For Riskified, Yuki helped unlock:

  • Lower Looker compute spend without asking analysts to change how they work
  • An end to sizing the default warehouse for the heaviest query
  • No manual query reviews or warehouse resizing for the data team
  • Savings that showed up where manual tuning could not reach: ad hoc, self-service analytics

Takeaways

1. Static sizing fails when workloads are unpredictable

A warehouse sized for the heaviest query overspends on everything else. A warehouse sized for the average slows down the queries that matter. Self-service BI needs a decision per query, not per warehouse.

2. ETL is the easy part. Analysts are the hard part.

Scheduled pipelines can be profiled and tuned. Ad hoc analytics cannot. That is why Riskified saw approximately 50% savings on Looker and approximately 20% across the broader mix.

3. The best optimization is invisible

Analysts kept using Looker as usual. The data team stopped resizing warehouses. Nobody had to learn anything new.

4. Ask one question first

How much of your Snowflake traffic comes from analysts? The higher the share, the bigger the opportunity for automated routing.

Want to See What You Could Save?

If a large share of your Snowflake traffic comes from Looker, analysts, and self-service BI, static warehouse sizing is costing you more than it should.

Yuki routes each query to the right-sized warehouse in real time, with no code changes and no new habits for your analysts.

Download the full Riskified case study

By Perry Tapiero
Perry Tapiero leads marketing at Yuki, driving demand generation and brand growth for B2B and B2C SaaS companies in FinTech, AdTech, and Cybersecurity. With 15+ years of experience, he specializes in go-to-market strategies, ICP refinement, and managing multi-million-dollar campaigns using HubSpot and Salesforce. Previously at other companies, he led ABM, PBM, and product marketing initiatives that drove ARR growth and helped achieve Gartner Magic Quadrant recognition. Perry was a regular contributor for marketers and now shares his insights on LinkedIn.

Table of Contents

Free cost analysis

Take 5 minutes to learn how much money you can save on your Snowflake account.

By clicking Submit you’re confirming that you agree with our Terms and Conditions.

Follow us on LinkedIn

Related posts

Free cost analysis

Take 5 minutes to learn how much money you can save on your Snowflake account.

By clicking Submit you’re confirming that you agree with our Terms and Conditions.

Skip to content