How Levi’s Achieved 90% Dimension Accuracy with AI–and Why It Changed Their Supply Chain Economics

A Decade of Impact: Levi’s Customer Story

In global apparel supply chains, some of the most expensive problems aren’t visible to customers but they quietly compound at scale. For Levi Strauss & Co., one such challenge was deceptively simple: accurately knowing the physical dimensions of thousands of products before they ever moved through the network.

With new SKUs introduced every season and data coming from a large, distributed vendor ecosystem, manual dimension capture had become inconsistent, slow, and error-prone. Small inaccuracies in length, width, height, and weight translated into inefficient packaging, underutilized containers, higher transit costs, and avoidable penalties downstream.

To solve this, Levi’s partnered with Pluto7 to apply machine learning in a focused, logistics-first way using historical product data and Google Cloud to predict folded product dimensions with high accuracy. The result was a practical AI solution that delivered 90% dimension accuracy in six weeks, enabling better planning decisions long before products reached distribution centers.

This engagement reflects a broader pattern Pluto7 has observed over the past decade: the most impactful AI initiatives rarely start with flashy use cases. They start by fixing the quiet planning gaps that sit between intent and execution.

The Challenge: Manual Dimensions at Global Scale

Levi’s introduces thousands of new products every season across multiple styles, fits, and sizes. Traditionally, the responsibility for providing product dimensions sat with vendors often resulting in incomplete, delayed, or inconsistent data.

This created several downstream challenges:

• Packaging plans were based on estimates rather than facts
  
• Container utilization was suboptimal, wasting space and increasing freight costs
  
• Third-party logistics providers lacked early visibility to pre-book capacity
  
• Errors led to penalty charges and reactive, last-minute adjustments

The problem wasn’t a lack of data it was the inability to use existing historical data to predict what new products would look like operationally.

The AI Approach: Predicting Dimensions Before Products Exist

Pluto7 implemented a Logistics Machine Learning solution on Google Cloud that shifted Levi’s from manual estimation to predictive intelligence.

Instead of waiting for vendors to submit measurements, the model analyzed Levi’s historical product data to predict folded dimensions for new SKUs before they were physically handled.

Key inputs included:

• Historical product dimensions
  
• SKU-level attributes (size, fit, style variations)
  
• Brand and category information
  
• Consumer call-outs and product characteristics
  

The system understood, for example, that a 32×30 jean behaves very differently from a 40×44, even within the same product line allowing predictions at a granular, SKU-specific level.

The Technology Foundation: Google Cloud + Pluto7

The solution was built natively on Google Cloud Platform, leveraging services designed for scalability and rapid iteration, including:

• BigQuery for large-scale data processing
  
• Google Cloud AI Platform for model development
  
• App Engine, Cloud Run, and Datastore for application scalability
  
• Google Data Studio for visualization and operational insight
  

To make the intelligence accessible to business users, Pluto7 developed a custom user interface that allowed Levi’s teams to:

• Refresh data dynamically as new planning activities emerged
  
• Select product attributes via intuitive dropdowns
  
• Instantly generate predicted dimensions for new SKUs
  

What had once taken hours or days of manual effort became an on-demand, repeatable workflow.

Results: 90% Accuracy in Six Weeks

Within six weeks of implementation, Levi’s achieved 90% prediction accuracy for product dimensions. That accuracy unlocked tangible business outcomes:

• Optimized packaging with better carton utilization
  
• Lower transit costs through improved container planning
  
• Reduced penalty costs caused by shipping discrepancies
  
• Elimination of manual data entry errors for upcoming products
  
• Earlier freight booking, giving logistics partners better visibility
  

Critically, these gains scaled seamlessly during seasonal product surges when thousands of new SKUs enter the system at once.

Beyond Dimensions: Smarter Volume and Freight Planning

The engagement extended beyond dimension prediction. Pluto7 also built supplier-level machine learning models using historical purchase order data to forecast product volumes more accurately.

This allowed Levi’s to:

• Move beyond rolling forecasts
  
• Pre-book ocean and air freight capacity with confidence
  
• Reduce last-minute inter-DC transfers driven by poor visibility
  
• Plan remediation actions earlier in the cycle
  

Together, these capabilities improved Levi’s ability to manage volatility especially during peak seasons and demand shifts.

Where Planning in a Box – Pi Agent Fits

This solution reflects the same architectural philosophy behind Planning in a Box – Pi Agent, Pluto7’s agentic planning platform.

At its core, Pi Agent is designed to:

• Unify fragmented data into a single planning view
  
• Apply AI agents to specific decisions (demand, inventory, logistics, finance)
  
• Surface recommendations that planners can trust, validate, and act on
  

In Levi’s case, the logistics ML solution functioned as a focused planning agent augmenting human planners with predictive insight while keeping them in control of decisions. It’s a clear example of how agent-driven planning works best when applied to one high-impact problem at a time.

Why This Matters

For Levi’s, AI wasn’t an experiment , it was a practical tool to remove friction from planning. By replacing manual estimation with predictive intelligence, the team improved accuracy, reduced costs, and gained the confidence to plan logistics earlier in the lifecycle.

More broadly, this story reinforces a lesson Pluto7 has learned repeatedly over the last decade:
precision in planning compounds. When foundational inputs like product dimensions are accurate, every downstream decision improves.

As part of Pluto7’s Decade of Impact, the Levi’s case stands as a reminder that the most durable AI transformations aren’t about replacing people or ripping out systems. They’re about quietly strengthening the decisions that keep global operations moving at scale, and with measurable business value.