03 · Network data intelligence

Making network data visible, testable, and actionable.

A cross-device 3D platform that helped operators, modelers, and AI data scientists turn complex network signals into planning decisions.

Role: Lead Product Designer · UX architect
Scope: 3D network planning and simulation
Partners: Engineering, research, operators, AI teams
Talk track: 7–10 minute case study

Sanitized 3D network planning experience across desktop, tablet, and mobile — Cross-device planningThe experience joined data preparation, spatial modeling, simulation, and analysis across desktop and field contexts. Screens are cropped and sanitized for public presentation.

The challenge

The system had powerful network data and AI-assisted analysis, but experts still needed a usable way to understand the physical world behind the numbers.

Planning required users to assemble map data, model buildings and equipment, run simulations, and interpret dense results. I reframed the product around one spatial decision journey rather than a sequence of specialist tools.

01 · Translation

Data lacked physical meaning

Tables and model output did not clearly explain coverage, obstacles, or downstream impact.

02 · Workflow

Planning took about three hours

Repeated setup and context switching slowed expert decisions.

03 · Scale

Multiple teams shaped one tool

Research, engineering, and regional operators needed shared interaction rules.

Experience architecture

I organized the product around the network data journey.

The architecture connected raw spatial data to a semantic 3D model, simulation inputs, calculated results, and a decision. It gave specialists a shared language while allowing each role to enter at the level of detail they needed.

Sanitized journey from building data to network simulation — From data to decisionBuilding data becomes a semantic environment, then a network model, simulation, and optimization result.

Six-stage decision modelPublic-safe reconstruction

01Acquire
Bring in spatial data

02Model
Create semantic layers

03Place
Set network points

04Simulate
Run signal analysis

05Compare
Read tradeoffs

06Decide
Commit a plan

Foundation for agent supervision

The model analyzed. The planner supervised and decided.

This work predated today's agent and MCP ecosystem, so I do not present it as an agent product. It established the same management principle: make machine analysis observable, connect output to spatial evidence, support scenario comparison, and keep consequential decisions with the expert.

Sanitized 3D city model with generalized network analysis — Spatial explanationColor, position, and layers translated dense network output into inspectable patterns. Navigation and identifiers are excluded.

Supervision model

The model could calculate and recommend. It could not approve a deployment choice. That human-control pattern now informs how I design agent plans, observability, approval, and intervention.

Evidence, trade-offs, reflection

The spatial model evolved through expert use, not visual novelty.

People

Specialists across contexts

Network planners, researchers, modelers, and field specialists needed to inspect the same network evidence at different depths across desktop and field workflows.

Research + testing

Measure the planning task

Workflow observation and usability evaluation connected interaction changes to comprehension, adoption, completion time, and dropout rather than judging the 3D interface on appearance alone.

Trade-off

Accuracy versus cognitive load

Showing every model layer improved technical completeness but obscured decisions. I used semantic layers and progressive detail so experts could reveal complexity when it became relevant.

Reflection

Validate regional variation earlier

I would involve regional operators sooner and test device constraints earlier, reducing the adaptation required as the platform expanded across markets.

System leadership

A three-layer UI system reduced cross-team rework.

I created shared specifications for foundational controls, data visualization, and spatial interactions. Monthly UX sprints connected research, product, engineering, and regional needs; the system made desktop and field experiences feel related without forcing them to be identical.

Foundation

Common controls

Navigation, editing, status, and validation established predictable behavior.

Data

Semantic visualization

Color, layers, legends, and comparison rules made model output interpretable.

Interaction

3D task patterns

Selection, placement, camera behavior, and progressive detail supported expert work.

61→86

Usability score

The redesigned workflow raised measured system usability.

58→84%

Adoption

Usage increased across the first six months after rollout.

3h→1.5h

Planning time

The connected workflow cut a representative planning task roughly in half.

25→10%

Dropout

Fewer users abandoned the workflow before completing a plan.

Global markets

The platform system supported adoption across regional contexts.

~30%

Less UI rework

Shared layers reduced conflict and repeated design decisions across teams.

Return to

Three data platform stories

View selected work ↗