Hotel Expense Reporting
Reduced task failure from 56% to 11% in usability testing with a guided workflow
Context
Submitting hotel expenses in enterprise systems is often slow and error-prone. This case study focuses on improving how travelers complete hotel expense itemization with a new machine learning–powered experience.
The goal was to reduce errors, improve task completion, and help travelers move through the process with more confidence.
This work is based on usability testing with a prototype, not production data.
the problem
The system relied on users to interpret it.
The existing experience required travelers to compare hotel receipt data with system inputs and decide what needed to be entered or corrected.
Even frequent travelers struggled to complete the task with confidence, having to interpret multiple pieces of information at once, second guess system behavior, and rely on trial and error to move forward.
Completion did not mean success. It often meant pushing through confusion.
the shift
Interpretation was causing errors
The issue was not missing instructions. The system relied on travelers to interpret multiple pieces of information at once.
When travelers had to compare receipts, system inputs, and required fields all at once, errors and hesitation increased.
The problem was not the form itself. It was how much interpretation the system required at a single moment.
What changed
From interpreting the system to guided completion
The redesign shifted from requiring interpretation to guiding travelers step by step.
Broke hotel expense itemization into smaller decisions
Asked questions at the right moment
Structured how receipt information is referenced and verified throughout the task
Helped travelers move forward with more confidence
The system uses machine learning to structure expense data and reduce manual interpretation.
the solution
A guided workflow leads frequent travelers through submission step by step
Instead of exposing a full form, the experience breaks submission into structured steps that request only the information needed at the right time.
Guidance is embedded throughout the flow, helping frequent travelers understand what to do next without relying on instructions or external resources.
Validation is integrated within each step, reducing errors before submission instead of after.
Validation
Tested with frequent travelers using corporate credit cards
9 unmoderated usability sessions (1 per participant)
Participants were frequent business travelers
All had recent experience submitting hotel expenses
All used corporate credit cards
Travelers completed multiple flows:
Current experience
Initial guided experience
Refined guided experience
The second iteration was updated based on feedback from the first round.
Results
Fewer failures and smoother completion
Task failure dropped from 56% to 11% in usability testing
Most participants completed the refined guided flow successfully
Participants described the experience as easier and more intuitive
Even when issues occurred, users were able to recover and continue.
Key insight
Even experienced travelers struggled when they had to interpret everything at once. When structured, they moved faster and with more confidence.
Learnings
What I learned
Design decisions improve when they reflect user behavior, not system structure.
How I grew
I improved my ability to identify when problems are caused by system behavior rather than UI design.
What changed
I moved from improving forms to structuring how systems guide decisions.