Los Angeles is home to many distinct communities, each with its own story in the data. Photo: Jon Tyson / Unsplash
ACT Data Hub
Designing a path from a community question to evidence a decision-maker will accept.
Context
UCLA, All Children Thrive – California: 14 community coalitions
Role
Research, information architecture, interaction design, build
Team
Sole designer; a student worker prepared map layers
Timeline
Spring 2025 – spring 2026
Methods
Moderated usability testing, information architecture, prototyping
Tools
ArcGIS Hub, Experience Builder, StoryMaps, HTML and CSS with Esri's Calcite components
Context
All Children Thrive – California supported community coalitions working on housing, homelessness, education, environmental conditions, youth spaces, and other conditions affecting children and families. Each coalition chose its own questions and policy goals, usually with a decision date ahead: a council vote, a funding meeting, a report.
I worked in UCLA's data technical-assistance program, helping coalitions turn those questions into evidence. Session after session went to finding individual variables and rebuilding basic maps. Little time was left for what mattered: what the data could support, what it couldn't, and how to make the case.
The resources already existed. They were scattered, named in technical terms, and organized the way data teams organize data. I proposed one hub where teams could explore on their own and bring better questions to our sessions.
First version
I started where most data portals start. In spring and summer 2025 I built the Hub around types of data product: a searchable catalogue of datasets, StoryMap guides organized by data category, and a separate section for mapping tools. Hub pages don't run JavaScript, so I built the pages in HTML and CSS with Esri's Calcite components.
By fall I had added entry points on the homepage for the deliverables teams were working on ("I'm working on a needs assessment"). Underneath, the three sections hadn't changed, and the catalogue was still the most prominent way in.
Testing
In October 2025 I tested the Hub with 5 staff members from 3 funded projects, in facilitated sessions of 45 to 60 minutes. Each brought a question from their own project, and I asked them to find relevant information, interpret it, and decide how they would use it.
0 of 5completed the core tasks without help
5 of 5started in the catalogue and got stuck
4 of 5couldn't work out how to combine indicators
The catalogue set the wrong expectation. It held only products from past ACT projects, but people read it as a library of all community data. Its search matched exact text only: no synonyms, no sense of place. A search for Big Bear didn't surface the San Bernardino County map that covered it.
The homepage asked too much. Six to eight options carried equal weight, with no clear place to start.
Every section was a dead end. Finding a variable led nowhere, and a guide didn't link to its map. People knew which indicators they needed but not how to bring them together.
The guides taught when people wanted to act. Someone who needed a poverty-rate map for a needs assessment didn't want to read a lesson on the data category first.
“There's a lot of tabs… when you see something for the first time, it could be a little bit overwhelming.”
“How would you bring those into one combined map to demonstrate the interplay of those factors?”
Participants, October 2025 testing
But the parts worked once people reached them. Every participant understood the guide to combining indicators as soon as they saw it, the community context maps drew the strongest reactions, and people asked for more step-by-step guidance, not less. The problem wasn't capability. Each part of the Hub was an island.
The Hub's structure before and after testing. Each section was a dead end; the redesign connects them in the order people work.
Redesign
I reorganized the Hub around the work instead of the data: find and explore, understand and apply, communicate and share. Each decision traced back to something I'd watched happen.
Take the catalogue out of the path. I considered a banner explaining what the catalogue held, but testers skipped explanatory text once they had formed an expectation. So the catalogue left the homepage, teams got direct links to their own materials, and a combined exploration map became the way in.
Three paths instead of eight options. The homepage now asks where you are and offers three tasks. Newcomers get one secondary pointer: "New to community data analysis? Start with Understanding Your Data."
Give combining its own place. A new Combining Indicators section, equal in weight to exploring and communicating, is built around the guide every tester understood. Its 4 steps end with validating patterns with data experts before making policy claims.
A next step at every transition. After the maps, a "What's Your Next Step?" panel offers two routes: understand the data or combine it. After combining, a prompt leads to storytelling. The storytelling page opens by asking whether you've combined your indicators yet.
Name things by purpose. "Research Guides" suggested how-to instructions but held background context. It became two sections: Understanding Your Data for context and Combining Indicators for procedure.
Look before download. Map cards traded "Download data" for "Explore map," and context maps moved to the top. People wanted to see the data before they wanted files.
Keep the depth. Testers valued detailed walkthroughs, so I kept the guides substantial and changed how people reached them.
The homepage after the redesign. Three task paths replace six to eight equal options, with one pointer for newcomers.
After exploring the maps
Combining indicators
Before storytelling
Prompts at each transition in the live Hub. Each one points to the next step, or back a step if one was skipped.
An analysis guide in the live Hub: the combined map of school results and neighborhood conditions, then the questions to ask before sharing a claim.
Results
In spring 2026 the project's Equity Advisory Group reviewed the redesign and how the guides were framed for non-technical users; their feedback was minor. Then I ran a final round of 30-minute tests with four project staff who hadn't taken part in October. The tasks matched the first round, and I gave no help beyond validating testers' comments.
4 of 4completed the full path, from exploring to communicating
14coalitions had access to the Hub
5active teams kept using it after testing
The combined exploration map became the most-used component. It was the new way into the Hub, replacing the catalogue.
Assistance sessions changed. Teams arrived having already looked at several conditions and asked about comparison and interpretation instead of where to find a variable. In City Heights, a question about housing costs grew into one about demographic change, community history, and collective efficacy.
I didn't instrument the Hub, so these are observations from sessions rather than measured changes.
The combined exploration map in the live Hub, at census-tract level with one tract open. It became the way into the Hub and its most-used component.
Reflection
Test before building. My first formal test came months after I started building. A short session in spring 2025 would have caught the catalogue problem before I designed around it.
Measure from the start. With usage tracking at launch, the shift in session questions could have been counted, not just observed.
Work within the platform. ArcGIS Hub can't export a single layer for someone to build their own map, so teams could explore and learn but not assemble a finished map inside the Hub.
Support interpretation next. Once teams could find data, interpreting it became the bottleneck. That's where I'd go next, including whether an AI assistant could help a team read a map alongside the guides.
The ACT program closed in 2026. I documented the Hub's information architecture, production routines, and guides so they don't depend on the public site staying up.