Automation Billing System
A billing workflow for maintenance and repair work orders
Building the billing application that helped recover an estimated $75–80M in revenue from a backlog of maintenance and repair work orders, and simplifying it from 23 buttons to 8 as we learned how the finance team actually worked.
Role
C#/SQL Developer (Full-Stack)
Dev Timeline
July 2018 – November 2018
Team
Jeyam Silva & Michael Brutsch

Timeline
Iteration 1
12 weeks · July–Sept 2018
Upload, edit, and download billing details in file formats accepted by their different customers.
Iteration 2
8 weeks · Oct–Nov 2018
Four modules that read left to right and top to bottom. In the download module, buttons became 4 dropdowns.
Added features
Through March 2021
CEDEX codes and AFTRAS features were developed as needs changed.
Ongoing support
Nov 2018 – Mar 2021
New requirements and bug fixes were completed as issues were raised.
Who I worked with
End users (5)
Our client's 3 AP workers, 1 AP manager, and 1 branch manager.
Point of Contact
Our client's employee who was a self-taught developer. He advised us on the Crystal Report requirements
Client’s Customers
Their different billing needs and priorities shaped our requirements.
Development Team
Me, a C#/SQL co-developer, and a senior developer.
Background
The situation
MNR (maintenance and repair) work-order billing covers all the work done for customers that needs to be paid for. Our client’s finance team handled it across several different systems, some covering parts of the billing process. A self-taught developer inside the company had made real progress on the requirements, especially around Crystal Reports, and advised us.
The goal
Iteration 1
Let users upload, edit, and download billing information in the expected format. Users asked us to mimic their old design.
Iteration 2
Reduce cognitive load and optimize usability by consolidating buttons and reorganizing the page to better match the user's mental models of the billing process.
Scope
In scope: database setup, Excel sheets, filters and dropdown lists.
Out of scope: pulling data that had already been billed, and personal user account information.
Constraints
• The first design had to match their retro system: a black, gray, and white palette, with simple shapes and columns
• Each of the client’s customers had different billing requirements and priorities
Research & Process
With five end users, research started with a site visit and continued in a steady email loop. Support conversations showed us where the product helped and where it confused.
5
end users: 3 AP workers, 1 AP manager, 1 branch manager
3
AP workers interviewed on site in Phoenix
2
main design versions, shaped by feedback
23 to 8
buttons, from Iteration 1 to Iteration 2
01 Discover
Learn the finance workflow
Field visit and contextual inquiry
Observed and interviewed all 3 AP workers on site in Phoenix. We spoke with each individually, though others nearby could chime in.
Stakeholder interviews
On-site and off-site conversations to understand the finance world and each customer’s needs.
Requirements and documentation
We gathered requirements and I wrote SRS and SDD documentation.
Stakeholder mapping
Mapped our client’s customers, whose different needs and priorities shaped the requirements and database structure.
02 Evaluate
Learn from real use
Onboarding Walkthrough
Walkthrough with users showed where our original mental models misaligned with the end users.
Support requests as usability evidence
Some feedback pointed to confusion, like a “missing” feature that was already on screen, or verbiage that was inconsistent with financial lingo.
03 Iterate
Keep the loop open
Email feedback loop
Initially we had daily communications, then once or twice a week as requirements were ironed out.
Agile and iterative design
Designed and developed in agile cycles, allowing flexible and quick changes.
Information architecture
Built SQL tables categorizing functions and how they relate, allowing for progressive disclosure within dropdowns.
04 Measure
Quantify the change
Before and after comparison
Reduced clicks per task: iteration 1 had 23 buttons and 5 dropdowns. Iteration 2 had 8 buttons and 4 dropdowns.
From Findings to Design Decisions
Each decision traces back to something users told or showed us.
What I learned
What we changed
User wants vs. user needs: Mimicking their old system UI still caused confusion and delays. User's mental models about the billing process were not optimized.
Iteration 2 dropped the old layout. It reads left to right and top to bottom in order of the billing process flow.
Poor usability: Users sometimes reported a “missing” feature that was already on screen.
Consolidated 23 buttons and 5 dropdowns into 8 buttons and 4 dropdowns, improving information hierarchy and scannability while reducing cognitive load.
Customization: Our client’s customers each had different needs and priorities.
Built a SQL table that categorizes functions and how they relate, so each dropdown autofills the next.
Design vs. backend: Some UI support questions were actually requirements changes from the client’s customers.
Treated them as requirements and bugs to fix, and kept the email loop open as they arrived.
The designed system

Iteration 1: mirrors client's old system layout with black, gray, and white palette

Iteration 2: improved information hierarchy with four modules, with buttons consolidated into dropdowns
Results
$75–80M
revenue recovered, estimated by our client
34.7%
reduction of clickable elements by ver. 2
Beyond the numbers
Users adopted the product quickly. One of our client’s customers wrote to them:
“I have some WONDERFUL news. Yesterday in the M&R system, we were only showing 5 invoices. This morning we are showing 217. 100 are already processed and approved and the other 117 will be done by noon. This is all for the Genset Fuel! We are making progress!”
Dollar figures are client estimates for the whole project, team-wide, including the MNR CEDEX and AFTRAS add-on features. We didn’t collect time-on-task or training-time data.
Reflection
Turn assumptions into questions.
What the project taught me
Product design has to match users’ mental models. As designers and developers, we often build as if we know what’s best for users, but it’s truly collaborative. Sometimes that means looking past what users say they need to what they’re trying to achieve.
What I’d do differently
At the time, design wasn’t prioritized because the focus was billing revenue. With more time, I’d run a competitive analysis of how other companies have optimized their systems for billing needs.
Focus on the facts
Focus on the facts of the matter. We don’t need to push our assumptions aside. We can turn them into clarifying questions for our users.