Why revisit DCWP?
Because the usability study I conducted at the beginning of the year had already shown that there were significant issues with the Online Form. I wanted to see if the data itself could tell that same story
The DCWP usability project I ran earlier this year produced a clear finding. The Online Form, despite being one of the agency’s primary intake channels, was the hardest to use. That case study made the argument through user testing. With this project, I took a different approach by using data to reveal potential issues with the intake channels available from the department. I developed four network visualizations, built on top of NYC Open Data’s complaint records, that surface what flows through which channel, and put a number on how much the Online Form is being asked to carry.
Starting where the last project ended
The first viz was a full network of all complaints across all channels. It worked, but it didn’t focus
I built the first version in Gephi using DCWP complaint records from September 2022 through April 2024. Nodes are either intake channels (Phone Call(311), Online Form, Email, Mail, etc.) or complaint topics (Pricing & Billing, Wage & Hour, Harassment, and so on). Edges represent complaints flowing from one to the other, weighted by volume. Color encodes which population a node belongs to: consumers, workers, or shared infrastructure used by both.
The network does its job. You can see Phone Call (311) dominating, the Online Form sitting as a second hub, and the two complaint populations occupying their own regions of the graph. But it shows everything at once, which means the eye has to do all the work. For a case study that’s specifically about the Online Form, I needed to find a way to make the data point at it.
The 3D detour
I wanted to render the network in 3D to make the bridging structure more legible, but Gephi had other ideas
The first thing I tried was a 3D layout using Force Atlas 3D. The plugin successfully computed three-dimensional coordinates for every node, but Gephi’s viewport renders only a flat projection of them. Without orbit controls or any way to navigate the depth axis, the result was a 2D layout with hidden depth values, not the 3D visualization I was hoping for. A real interactive rendering would require exporting the coordinates and rebuilding the network in a different software, which fell outside the scope of this iteration. I’m setting that aside for now, hoping to revisit the concept in the future.
The radial detour
An eye-pleasing layout with the same issues
After the 3D attempt, I tried a radial axis layout. The idea was to group nodes by population (worker, consumer, shared) and fan them out along sparse radiating from a center. Structurally, it worked. Visually, it was a different shape with the same information. Same nodes, same edges, no new findings surfaced. However, it showed potential, thanks to the simpler distribution of nodes. I took the learnings and changed my approach.
Small multiples, finally
A classmate suggested showing the populations separately. That was the unlock.
Instead of one viz trying to show everything, I broke down the network into three visualizations, each showing a different aspect. I built two panels by filtering the network down to a single population: one for worker complaints, with their intake channels, and one for consumer complaints, with theirs. The radial axis layout I’d scrapped earlier finally became useful. With only one population per panel, the spar becomes a clean fan of complaint topics flowing into a small cluster of channels at the base. The shape itself helped highlight the story that workers funnel mostly through Phone Call (311), and Email, consumers funnel almost entirely through 311.
Isolating the Online Form
The fourth panel works specifically as an opener for the case study
The fourth panel filters the network down to a single intake channel, the Online Form, and every complaint topic that flows through it. The result is the visualization I needed to fit within the usability study. The Online Form sits at the bottom as the anchor, with consumer complaints fanning out on one side and worker complaints on the other. The asymmetry highlights the point I make in the study. The Online Form is being asked to handle the entire variety of complaints that come into the agency, from harassment claims to refund disputes to wage theft. It’s the single channel where the system’s full complexity gets routed through one interface. Which is exactly why the usability study comes into play.
Strengthening the original case
The four networks will be added to the DCWP usability case study as supporting evidence. The usability work showed that the Online Form was hard to use. These visualizations show what’s at stake when it doesn’t work, including the volume and variety of complaints that depend on a single interface working well.
What’s next
The 3D version is still in my mind. I want to rebuild it properly in dedicated software. The plan is to showcase it on my portfolio, as a “playground” project, where it can live as a real explorable network instead of a flat projection. The visualization does not need a 3D aspect, however, the data could simplify its message by allowing someone to move through it.