Problem
Size analysts spent weeks manually profiling products each season, creating bottlenecks in the go-to-market timeline. Their process involved working in personalized Excel spreadsheets offline before uploading data to SAP-a fragmented workflow that delayed product launches and ultimately impacted Levi's consumers.
How might we accelerate size profiling without sacrificing accuracy or analyst confidence in the outputs?
Context
At Levi Strauss & Co., each Fall and Spring season launches with hundreds of products across global markets. Before any product ships, size analysts determine which sizes to offer and how many units to allocate per size. Using historical sales data, they forecast demand to balance two critical risks: missed sales from underproduction and excess inventory at season's end.
This process directly impacts revenue and brand reputation. Get it wrong, and customers can't find their size - or the company sits on dead stock.
Approach
Discovery
I joined the team to solve this challenge, starting with qualitative interviews with size analysts across markets. I shadowed their workflows, documenting every step from data gathering to final SAP upload. The fragmentation was worse than expected - each analyst maintained their own Excel templates with custom formulas, creating knowledge silos and preventing any standardization.
Design Strategy
Working with Product and Engineering, we defined an MVP focused on one core workflow: profile creation. Rather than rebuild everything analysts did in Excel, we identified which tasks were repetitive and which required human judgment.
A critical early decision was introducing AI-driven profile suggestions. Initial analyst reactions ranged from skeptical to threatened - many feared automation meant job elimination. We reframed AI as a starting point, not a replacement, positioning it as a tool that handled repetitive data analysis so analysts could focus on strategic decisions.
Building Trust
To gain buy-in, we established monthly stakeholder reviews with leadership and weekly working sessions with analysts. This frequent feedback loop helped us:
- Validate that AI suggestions were directionally correct
- Surface edge cases where human judgment was essential
- Demonstrate incremental value before full launch
Design System Integration
Like other internal tools, SizePro initially had no connection to Levi's brand. I introduced design patterns from levi.com - typography, color, and interaction patterns that felt familiar to analysts who shopped the brand. This wasn't cosmetic; it signaled that internal products deserved the same craft as customer-facing ones.
Outcome
SizePro launched to analysts in North America and Europe, delivering immediate impact. Time spent on profile creation dropped by half, allowing analysts to manage larger product assortments without expanding headcount. More importantly, analysts shifted from data entry to strategic analysis-reviewing AI suggestions, adjusting for market nuances, and focusing on high-value products.
The project also established a precedent: incremental delivery works for internal products. Stakeholders who initially resisted an MVP approach saw the value of shipping sooner and iterating based on real usage.
Reflection
This project reinforced that trust is built through transparency. Analysts weren't resistant to AI-they were resistant to being blindsided by it. By involving them early, showing our work, and positioning AI as augmentation rather than replacement, we turned skeptics into advocates.
If I were to do it again, I'd push for even earlier technical validation. We spent time refining designs before confirming the AI model could deliver reliable suggestions, which created downstream rework. Tighter collaboration between design and data science from day one would have accelerated delivery.