Levi's Bulk Edits
Industry
Retail Planning
/
Enterprise Machine Learning
/
OVerview & GOALS
The goal of this project was to remove barriers to Levi's' sizing analysts' adoption of SizePro, an in-house sizing app. My approach was to speed up sizing analysts' workflow while using SizePro, while also boosting their analytic capabilities by enabling them to compare machine learning recommendations in one place. This would potentially >5x analysts’ coverage of styles that made the most impact for Levi’s globally.
User-provided definition of “bulk edits”: The ability to edit the size distribution for multiple apparel styles for one sales channel.
Team & Timeline
As the sole product designer for SizePro, I worked closely with the Sr. PM, as well as an overseas team of 5 full stack developers.
I balanced finding an elegant solution with expediency, as this was a critical piece for cementing the success of this new product.
/
View Live Example
No items found.
I. Problem Discovery & Definition
Before this release, sizing analysts in different regions used their own tools for making bulk edits. Because SizePro requires them to enter edits one product at a time, most analysts were hesitant to move their workflows into SizePro. Some differences between regions and specific analysts include:
- Different regions have their own use cases for bulk edits, ranging from pre-season planning and analysis to applying pre-determined edits from sales teams.
- Analysts also often had their own approach for how they grouped different apparel styles and sales channels when making edits.
My approach was to focus on creating a streamlined yet flexible experience for grouping, analyzing, and editing data.
Below is the final bulk editing flow in this implementation. The iterations below showcase how I arrived at this flow.
No items found.
User Quotes
I would want to use bulk editing as the primary interface for pre-season planning. Today I make bulk edits on a monthly cadence due to generally lower volumes - I'm grouping styles together. I'm conducting analyses on the size distributions while making edits.
- APAC sizing analyst
Today I would not want to use bulk editing for pre-season planning. I only use bulk edits to input exact changes from DTC teams on a seasonal cadence. There is no analysis involved.
- North America sizing analyst
/
II. Solution Ideation & Validation
Iteration
The first challenge I tackled was how to create a system that allowed analysts the flexibility to select custom groupings of products.
- I first tried approaches that allowed maximum flexibility and built into the existing flow for editing products one by one. However, this approach seemed to add cognitive load both to the existing flow and created multiple steps before even arriving at a bulk editing state.
- I then focused on creating a minimal set of filter options for groupings. More can always be added later if needed, but the final set in release was validated by all regions.
No items found.
/
Iteration
One of the trickiest constraints during this project was that size analysts can only edit styles with the same “size grids” (e.g. “XS M L XL" or "M L XL XXL")
- I first tried standard options such as dropdown menus, assuming that size grids could be represented in a standard shorthand, e.g. "XS-XL", "24|32-30|34". However, I quickly learned that this assumption does not hold; size grids do not adhere to a specific range or pattern. I tend to want to show things as visually as possible and avoid strange edge cases.
- I then experimented with visually displaying available size grids and allowing analysts to easily compare and select grids. Through validation we found that this approach was intuitive enough to not hinder bulk edits.
No items found.
III. Impact & What's Next
After this release, SizePro reached feature parity with the suite of tools that it was replacing. Eight new sizing analysts across Asia and Europe were fully adopted, as well as 3 new sizing analysts in North America. This release now enables analysts to view machine learning recommendations for products grouped at various levels, and also saved millions annually in licensing fees by fully onboarding to SizePro. >90% of bulk edits are now made using SizePro. In APAC, >95% of edits are bulk edits. In other regions, >50% of edits are bulk edits and rising.
In the next iteration, the plan is to bring the regional sizing organizations together to standardize how they name and modify their size grids (which sizes are available for a given product), which will further streamline predictions as well as simplify the interface for selecting size grids.
This is an extremely exciting release that marks a turning point in the trajectory of SizePro by making it much faster and more efficient to monitor and adjust machine learning-generated size distributions. Victoria took a potentially lengthy and complicated workflow and turned it into an easy and intuitive experience. It's going to make our sizing analysts very happy.
- Shelby Greeley, Senior Product Manager, Levi Strauss & Co.
