ACORN STEED

Work

Lab

Case study · Lighten

As founder and head of design of Lighten, a startup for people leaving incarceration, I designed the structure underneath the product and led the team that built on it.

Everything from Zero

built the team and internal culture, spec, IA, and agency partnerships

Everything from Zero

built the team and internal culture, spec, IA, and agency partnerships

Facilitating Decisions across Relationships

aligned the DOJ, nonprofits, and technologists on one data model

Facilitating Decisions across Relationships

aligned the DOJ, nonprofits, and technologists on one data model

Human-First Structure

built the information architecture from how services were actually described

Human-First Structure

built the information architecture from how services were actually described

Lighten was built to help people leaving incarceration, for whom the first year is when most rearrests happen. In partnership with the Adult Probation Department, within the California DOJ’s Division of Recidivism Reduction and Re-Entry, I led a small pre-funding team to build a modern directory platform that standardized scattered, contradictory service data into something both service providers and help-seekers could act on.

Designing beyond the UI

Social services information was spread across guides that each organization maintained on its own, and that were printed with outdated, wrong, or vague information. Both user groups were non-technical: help-seekers, many without a reliable phone or internet, and providers without the resources to keep their listings current. The solution had to serve both, which meant print as well as web.

I saw that the real UX problem was data management. My solution was a flexible spec, built from the terms people actually used, as the product's foundation. The order came from the user journey for both help-seekers and providers: how they ask, what could waste their time, how they would trust an answer.

My core approach to information architecture

Previous failures set the requirements

Hand-audited the source guide for patterns, successes and failures set the spec.

Engineering built the database from that vocabulary.

→

↓

Separated meaning from order

Defined relational clusters with no hierarchy, applying order per context, web and print.

Designers laid out UI and print designs from these clusters.

→

↓

Structured IA from the user’s frame of reference

Ordered listings by the help-seeker’s questions: offerings, eligibility, contact info.

The printed guide design drew a strongly positive response in testing.

→

↓

Made tacit knowledge explicit and actionable

Turned caseworker know-how into fields: exclusions, what to bring, referral needed, etc.

Partner testing input across several sessions turned this into a key differentiator.

→

↓

Built correction into the model

I added owner fields, last updated dates, and agency accounts for record management.

Engineering built accounts where agencies could manage their own listings.

I used the same thinking to lead the team. Because pay was something I couldn’t immediately offer, I specifically recruited people who cared about the problem. I directed designers across web and print, coached engineers on usability, and carried the stakeholder work so they could build. APD’s input during meetings and testing focused our direction.

Disorganized data isn’t a zombie-ticket item. How information is structured decides who benefits. For our users, it could mean the difference between independence and a return to incarceration.

Disorganized data isn’t a zombie-ticket item. How information is structured decides who benefits. For our users, it could mean the difference between independence and a return to incarceration.

The throughline

Technology and leadership belong in service of people at a hard moment. The model decided what a person could find and trust, and the team made it real. That’s what I think technology is for, and the standard I design to.