Case Study

Case Study

Case Study

Order Accuracy Optimization Tool

Order Accuracy Optimization Tool

Order Accuracy Optimization Tool

Custom Ink is an online retailer that provides customized apparel and accessories for groups, businesses, and events. With over 4,000 products across more than 100 categories like t-shirts, sweatshirts, activewear, hats, bags, and drinkware, a streamlined workflow was imperative to correctly map supplier product information with the Custom Ink catalog.

Custom Ink is an online retailer that provides customized apparel and accessories for groups, businesses, and events. With over 4,000 products across more than 100 categories like t-shirts, sweatshirts, activewear, hats, bags, and drinkware, a streamlined workflow was imperative to correctly map supplier product information with the Custom Ink catalog.

Mapping Mishaps: Costly Confusion

Challenge

Even with 4,000+ products in Custom Ink's catalog and hundreds of new products being added weekly, "Product missing or not mapped" accounted for 20% of automated business operations (ABO) failures. The previous process to get products added to or updated in Custom Ink's catalog when mismatched or missing was tedious and time consuming, and the lift relied entirely on engineers. Custom Ink would continue to lose out on increased ABO rates, quicker time to service these issues, and poor experiences by not solving for an improvement to the overall process.


User and Business Goals

  • Create an entirely new interface application for the sourcing and logistics team to easily take on the mapping product tasks to save 4 full time employees (FTE)

  • Automatically download new products from all network suppliers on a weekly basis

  • During product import, the application looks at updates to the supplier catalogs and creates new mapping suggestions, product updates, and product removals

  • Have an audit log of any changes made within the application

  • Have ability to override sizes and colors for each supplier

Mapping Mishaps: Costly Confusion

Meet the Team

Meet the Team

Lead UX Designer (Me!)

1 Product Manager

1 Project Manager

2 Software Engineers

Lead UX Designer (Me!)

1 Product Manager

1 Project Manager

2 Software Engineers

Before 👎

Before 👎

The incredibly manual tool engineers built to fulfill a crucial workflow need. Unfortunately, it was a sea of information with little to organize it all.

After 👍

Entirely new tool complete with supplier updates dashboard, mapping results filtered by least accurate to accurate, and admin rights.

After 👍

  • 155+ orders an hour (goal was 65)

  • 99.9% accuracy in content review

  • Successfully saved 5.5 Full Time Employees (FTE)

  • Saved the company $400k in the first 6 months

Why is Everything so Manual?

Why is Everything so Manual?

After conducting three user interviews with the main engineers using this workflow, I discovered all the entirely manual processes that were slowing them down.

After conducting three user interviews with the main engineers using this workflow, I discovered all the entirely manual processes that were slowing them down.

  • Downloading supplier catalogs

  • Checking each individual product for incorrect matchings

  • All information was shown all at once

  • Nothing could be filtered down to make information easier to parse through

  • Downloading supplier catalogs

  • Checking each individual product for incorrect matchings

  • All information was shown all at once

  • Nothing could be filtered down to make information easier to parse through

Cross-Functional Team Synergy

Cross-Functional Team Synergy

I set up weekly working sessions with the product team and engineers that I was solving for, starting with low fidelity sketches and wireframes to narrow direction. Having visuals as talking points during these meetings were incredibly helpful at getting the entire working team on the same page.

I set up weekly working sessions with the product team and engineers that I was solving for, starting with low fidelity sketches and wireframes to narrow direction. Having visuals as talking points during these meetings were incredibly helpful at getting the entire working team on the same page.

In order to keep the project within scope, I designed with the minimum viable product (MVP) in mind. Through weekly calls with engineers and the collaborative relationship we'd built over time, we discussed questions, potential hurdles, and solutions.

Concepting with engineers and product in weekly working sessions

Concepting with engineers and product in weekly working sessions

Working Sessions, Tests, and Iterations

Working Sessions, Tests, and Iterations

After going through four weeks of collaborative sessions, design workshops and more user testing, I created these final screens (for now!) for the Product Mapper. Read below to see what each screen was used for, along with notifications and other essential interactions.

After going through four weeks of collaborative sessions, design workshops and more user testing, I created these final screens (for now!) for the Product Mapper. Read below to see what each screen was used for, along with notifications and other essential interactions.

Final Designs (for now!)

Final Designs
(for now!)

Dashboard

A quick overview of all the suppliers in our network in a dashboard view. New products were automatically downloaded every week, with visible red bubbles above each supplier.

A quick overview of all the suppliers in our network in a dashboard view. New products were automatically downloaded every week, with visible red bubbles above each supplier.

Mapping Results

Mapping Results

Once an individual supplier is selected, team members will see a list view of all the products needing review. Any mismatched data will be highlighted in red text, and the confidence match rating will reflect how accurate the matching is. Any confidence match rating under 70% is usually scrutinized further to ensure information is updated correctly.

Information is sorted by lowest confidence match first, so these can be updated and reviewed immediately.

Once an individual supplier is selected, team members will see a list view of all the products needing review. Any mismatched data will be highlighted in red text, and the confidence match rating will reflect how accurate the matching is. Any confidence match rating under 70% is usually scrutinized further to ensure information is updated correctly.

Information is sorted by lowest confidence match first, so these can be updated and reviewed immediately.

Rejected Styles View

Once a style is rejected, it gets minimized.

Once a style is rejected, it gets minimized.

Multiple Rejected Styles

Example of multiple rejected styles.

Confirmation Modal

Example of a confirmation modal after selecting "Approve All Remaining" call to action (CTA) to give team members a summarized view of all the changes they will be pushing through.

Example of a confirmation modal after selecting "Approve All Remaining" call to action (CTA) to give team members a summarized view of all the changes they will be pushing through.

Filtered View

Once all the correct mappings are updated, only the products requiring attention will remain in the list. Team members have the ability to undo the action, or move forward to the editor page.

Once all the correct mappings are updated, only the products requiring attention will remain in the list. Team members have the ability to undo the action, or move forward to the editor page.

Manual Editor

The Manual Editor page gives team members a quick way to see what the mismatched information is (red error state), and ability to quickly compare and update information in a side-by-side view of Custom Ink's product information.

The Manual Editor page gives team members a quick way to see what the mismatched information is (red error state), and ability to quickly compare and update information in a side-by-side view of Custom Ink's product information.

Empty State

Example of empty state once all mappings are complete.

Example of empty state once all mappings are complete.

Unresolved Issues

Two notifications will be present when there are unresolved issues. The red bubble above the supplier (with number of issues) and a "Pending Attention Required" section will list out the issues needing attention.

Two notifications will be present when there are unresolved issues. The red bubble above the supplier (with number of issues) and a "Pending Attention Required" section will list out the issues needing attention.

Successful Workflow

All supplier products have been mapped and updated. There is a summary of today's mapping results to give a quick overview and confirmation.

All supplier products have been mapped and updated. There is a summary of today's mapping results to give a quick overview and confirmation.

Reflection

Reflection

This was a high-impact initiative driven by engineering, ambitious in scope with just six weeks to deliver. Engineers flagged a major automation opportunity in a highly manual, error-prone workflow that accounted for 20% of order issues. I partnered closely with them and product stakeholders to untangle a deeply technical problem space and rapidly design an intuitive tool that could serve as the foundation for future ML-powered catalog management.

While I wasn’t involved post-deployment, the tool has since expanded in scope and usability, now enabling sourcing, logistics, and operations teams to autonomously correct and update catalog data with minimal engineering support. This positions the business to reduce dependencies, cut down on human error, and scale product accuracy across fulfillment systems.

Feedback from engineering leadership reinforced the strategic value of design in this space:

“This is a new UI that never existed before, and devs didn’t have a set vision for the UX, so it was nice to see the new ideas we embraced together.”


If extended, I would focus on:

  • Embedding lightweight AI or rule-based suggestions to catch mismatches early

  • Continuing weekly working sessions to reduce downstream QA bottlenecks

  • Building internal training to accelerate adoption across non-technical teams

  • Testing workflows with sourcing and logistics to validate end-to-end usability

  • Establishing a feedback loop for continuous iteration and intelligent optimization

© Jimin Ngo 2024

Back to Top

© Jimin Ngo 2025