Feature Discovery Research for MVP for an Enterprise Learning Platform (Amazon Learn)

Role: Lead UX Researcher
Timeline: 4 weeks
Methods: Remote interviews, thematic synthesis, LLM-assisted analysis

Overview

Amazon Learn is the company’s internal platform for employee training and development. One of its key features is checklists, used to validate hands-on skills across different operational roles. These checklists were replacing a legacy system called Knet.

My Process

1. Generative Interviews

I conducted 12 remote interviews with key user groups:

• 4 Validators

• 3 Driver Trainers

• 4 Admins who manage and build checklists

These participants came from delivery stations, data centers, and learning teams.

2. Accelerated Thematic Synthesis

Given the short timeline, I used Amazon’s internal LLM playground to assist in synthesizing insights:

• Created a list of relevant keywords (e.g. print, validation, checklist workflow)

• Used the LLM to summarize each transcript by topic

• Organized all findings in a spreadsheet by theme and user role

• Manually reviewed transcripts to validate and add nuance to the LLM summaries

Insight 1: Printability Is Essential for Safety and Access

The assumption going in was that a fully digital checklist would meet everyone’s needs. But the research showed otherwise. In environments like delivery stations and data centers, printing was not just helpful—it was necessary.

Delivery Stations

Driver Trainers often start with paper. They mark checklists by hand during observations and later enter the same data into a digital PDF for upload. Some had a workaround using fillable PDFs on a Kindle, but even that required extra steps like transferring files manually.

"If it was connected to Knet on a Kindle through a more automated program, that would be a significant improvement." – P8, Driver Trainer

Data Centers

Validators avoid using devices altogether in sensitive zones for safety and privacy. One mentioned copying checklist text into Word or Paint just to print a usable version.

"I don’t want to drop my phone or tablet into the sumps. If the paper gets wet, that’s still better than breaking equipment." – P5, Validator

Although some checklists had PDF attachments in Knet, this wasn’t well known. The need for an obvious and reliable print option came up repeatedly.

Recommendation:

• Add a visible "Print" button for every checklist

• Ensure downloadable, print-friendly formats

• Design for hybrid use cases where digital isn't always possible

Insight 2: Validators Need Location-Based Filtering

Another consistent pain point was the difficulty of finding the right learners to validate. Validators were often shown long, global lists of users, making it hard to identify who needed their attention.

"If I could filter by location and just check off the learners I'm responsible for, that would help a lot." – P9, Validator

"I'd love to just see people on my immediate team so I can help them quickly." – P5, Validator

Some data center teams had even built their own tools to work around this issue. Without proper filtering, both validators and learners experienced delays in checklist validation and certification.

Recommendation:

• Add site or location-based filtering for validators

• Allow bulk selection and validation by group

• Default views should prioritize users based on the validator’s own team or region

Impact

This research shifted the product team’s assumptions about checklist workflows. By highlighting the need for printability and localized filtering, we were able to push for two features that had been deprioritized during early planning:

• A print option was added to all checklist formats

• Location-based filtering was scoped for development in the next release cycle

Both were crucial for safety, accessibility, and usability in high-risk environments.