Process overview
01
Empathize
Interview + workflow observation
02
Define
Problem + design principles
03
Ideate
12 low-fi alternatives
04
Test
Technician review on selected structures
05
Refine
Revised wireframes + hi-fi interface
Ownership and boundaries
I owned the product experience from audit to prototype.
The redesign changed how technicians moved through diagnostic work while keeping the underlying engine and technical evidence intact.
End-to-end ownership
- 01Interface audit
- 02Workflow model
- 03Information architecture
- 0412 low-fi directions
- 05Technician review
- 06Hi-fi prototype
- Collaboration
- 1 network technician
- Fixed constraints
- Engine + raw output
Reconstructed a recent investigation and reviewed the selected interaction structures.
The diagnostic logic and one shared desktop model for Linux and Windows stayed intact.
01 · Empathize
Understand the real diagnostic workflow.
One technician reconstructed a recent investigation through six behavioural questions. The goal was to capture sequence, interpretation, comparison, handoff, and platform differences—not preferences about a mockup.
01
Participant
Network technician
06
Questions
Recent behaviour
03
Needs
Continuity · interpretation · reuse
Interview theme
Investigation sequence
What happened the last time a domain failed?
Which check did you run first—and why?
What emerged
The sequence changed with each result. A DNS response could lead to a server, port, or proxy check.
So I designed
A flexible workflow that preserves the investigation between checks.
Interview theme
Reading and comparison
How do you recognise a suspicious result?
When do you return to an earlier domain?
What emerged
Commands, status, and raw values had to be read together; earlier results were revisited for comparison.
So I designed
Interpretation without hiding technical evidence or previous results.
Interview theme
Handoff and platform
What happens after you understand a result?
What changes between Linux and Windows?
What emerged
Results were copied, shared, or rerun. Commands changed by platform, but the investigation stayed the same.
So I designed
Reusable actions and one task model across both command modes.

- 01
Run
One command becomes the entire screen
- 02
Interpret
Raw rows depend on technician memory
- 03
Continue
Previous domain and command disappear
- 04
Reuse
Copy, compare, and rerun have no clear path
02 · Define
Turn an output screen into an investigation workspace.
The core problem was not inaccurate diagnostics. It was lost working context: every command behaved like a new event, while the technician's real task continued across checks, domains, and actions.
Problem statement
Technicians could run useful checks, but the interface did not preserve the investigation those checks belonged to.
How might we
Keep every result understandable, retrievable, and ready for the next diagnostic action?
Continuity
Preserve continuity
What we learned
A diagnosis spans several commands and domains.
Design response
The interface must remember completed and active tests.
Interpretation
Explain without concealing
What we learned
Raw output is necessary but does not communicate priority alone.
Design response
Interpretation must sit beside inspectable commands and values.
Action
Support the next action
What we learned
Diagnostic work continues after a result is understood.
Design response
Copy, download, reopen, and rerun must stay attached to the result.
Success criteria
How the direction would earn its place.
- 01
Platform and command context stay visible from entry to result.
- 02
A completed test can be restored in one interaction.
- 03
Interpretation sits beside inspectable raw values instead of replacing them.
- 04
Copy, download, and rerun stay attached to the result they affect.
03 · Ideate
Compare structures before choosing a direction.
Twelve low-fidelity alternatives tested structure before visual polish. Each group converged on the option that best satisfied continuity, interpretation, and reuse.
Decision 01
Starting a diagnostic
How much structure should appear before the first command runs?
Evidence used
The technician moved between domains, commands, and operating systems, so all three had to remain visible at entry.
Single command field
Rejected: speed came at the cost of hidden platform context and a greater chance of running the wrong command.
Two-step modal
Rejected: the sequence was clear, but every new test introduced the same interruption.
Workspace header
Chosen: domain, platform, and command stay inspectable without adding another step.
Command library
Rejected: it optimised command discovery, while the primary need was running and continuing known checks.
Final decision
Workspace header
Expected impact
Fewer setup decisions to remember, clearer platform context, and a faster path from domain to first result.
Decision 02
Preserving completed tests
What should become the persistent unit of work?
Evidence used
Returning to a domain required its command, platform, and status, while several results needed to stay comparable.
Accordion history
Rejected: compact when closed, but comparison required repeatedly opening and collapsing domains.
Session dashboard
Rejected: strong overview, but it separated the technician from the active diagnostic thread.
dig +short
HTTP 200Named test tabs
Chosen: each domain becomes a retrievable unit with its command and result context attached.
Vertical timeline
Rejected: chronological order was clear, but parallel checks became a long serial list.
Final decision
Named test tabs
Expected impact
One-click recall of a complete test thread and less dependence on memory or copied output.
Decision 03
Reading and reusing output
How should raw diagnostics become an actionable report?
Evidence used
The technician read command, status, and raw values together, then copied or reran the result.
Raw table
Rejected: preserved every value but continued the original problem of equal visual weight.
Summary cards
Rejected: improved scanning, but separated conclusions from the raw evidence behind them.
Grouped report
Chosen: each diagnostic group combines meaning, raw values, status, and relevant actions.
Guided stepper
Rejected: imposed a linear reading order on an investigation that often branches.
Final decision
Grouped report
Expected impact
Faster scanning by diagnostic purpose while raw values stay inspectable and follow-up actions remain attached.
04 · Test
Test the selected workflow with a technician.
The prototype review focused on three real tasks: resuming a completed domain, switching command mode, and carrying a result into the next troubleshooting step.
Return to a completed domain
Observed
The technician needed the command and status—not only the domain name—to resume confidently.
Refined
Expanded tabs restore platform, command, HTTP status, and port context together.
Switch from Linux to Windows
Observed
The command changed, while the investigation structure remained familiar.
Refined
Platform mode became explicit without creating a second product flow.
Carry a result into the next step
Observed
Understanding the report was not the end; the result still needed to be shared or repeated.
Refined
Copy, download, and rerun actions moved beside the result they affect.
From feedback to structure
Revise the wireframes from observed behaviour.
Each observation changed a specific part of the interaction model. The revised wireframes made that response explicit before the interface moved into high fidelity.
Revision 01
Return to a completed domain
Expanded tabs restore platform, command, HTTP status, and port context together.
Initial wireframe
Refined wireframe
AFTER REVIEWLINUX · dig +short
HTTP 200Revision 02
Switch from Linux to Windows
Platform mode became explicit without creating a second product flow.
Initial wireframe
Refined wireframe
AFTER REVIEWPlatform changes the command, not the workspace.
Revision 03
Carry a result into the next step
Copy, download, and rerun actions moved beside the result they affect.
Initial wireframe
Refined wireframe
AFTER REVIEWDNS resolution
Server response
Port checks
05 · Refine
Build the hi-fi interface from the validated structure.
The visual layer now supports the interaction model already established in the revised wireframes: visible command context, retrievable test threads, explicit platform modes, grouped results, and attached actions.



04 · Shared platform model
Linux and Windows change the command—not the workflow.


Outcome
Keep four checks connected in one reusable workflow.
The redesigned workflow keeps the technician's full investigation visible: what ran, where it ran, what the result means, and what can happen next.
04
Diagnostic checks
DNS, ports, server response, and proxy
02
Platform modes
Linux and Windows in one task model
01
Click to recall
Named tabs restore a previous result
03
Result actions
Copy, download, and rerun
Delivered impact
The interface now carries the cognitive load.
- Less memory work
- The workspace now carries domain, platform, command, status, and prior results between checks.
- Faster return
- Named test tabs restore a completed diagnostic thread in one click.
- Clearer reading
- Results are grouped by diagnostic purpose while raw evidence remains inspectable.
- Stronger follow-through
- Copy, download, reopen, and rerun remain attached to the relevant result.
Before
Each command ended at an isolated output.
After
Every result remains a reusable diagnostic thread.
Accessibility decisions
Meaning remains visible beyond colour and hover.
Written status labels support colour cues instead of replacing them.
Numbered diagnostic groups create a predictable reading order.
Platform, command, and status remain text—not icon-only context.
Result actions stay visibly labelled and do not depend on hover.