Narration Tool
Helping more people access money advice through audio
Overview
People in housing and money trouble often can't get through the advice that is meant to help them. Dense, text-heavy guidance is hardest on the people under the most stress, particularly those with dyslexia or other processing differences.
The Scottish Legal Aid Board funded Shelter Scotland to improve digital access to money and debt advice, and COVID made it urgent as support channels closed and hardship rose. The work had been set up as a service design project, but COVID had made the in-person recruitment and group co-design that service design leans on very hard to run, and the project had spent a long time in discovery without converging on something to build. With the funder expecting delivery against the brief, I was brought into the team to help move it from discovery towards a working, testable solution, and I ran ideation as soon as I joined.
I was the sole UX designer on a team that included a service designer, four content designers and a money and debt advisor.
My Role & Impact
- Brought in to move a long-running project from discovery into delivery, and ran the ideation that set the direction
- Made guerrilla testing the spine of the research when COVID made in-person recruitment and group co-design impractical, securing access in Dundee and Kirkcaldy libraries
- Set the evaluation criteria the team used to choose audio narration from seven concepts
- Used contextual testing with neurodiverse users to choose and validate the tool, landing on Speechify
- Delivered an audio narration tool that, in testing, let neurodiverse users take in advice they couldn't get through text
- Fed the research into the content team's work, which supported cutting the advice from 123 pages to 36
Research and Discovery
COVID had made the usual research hard to run. Recruiting people into sessions wasn't working, and the group co-design that service design leans on was difficult to do safely. Rather than keep waiting for that to ease, I proposed guerrilla testing: instead of recruiting people in, going to where they already were and catching them in context, as restrictions allowed.
That meant setting up access with councils across Scotland and adapting the protocol to each site's rules. I ran sessions in Dundee and Kirkcaldy libraries. Testing in those real contexts surfaced things a lab never would, including how people behaved when they were stressed and short of time.
Key Research Findings
I facilitated workshops using fishbone analysis to turn months of qualitative research into clear design directions. The critical finding was that text itself was the barrier: dense written advice created both cognitive and emotional load for people already under financial stress.
Key insights:
- Dense text overwhelmed people already stressed about money
- Audio and video felt safer than reading about personal problems
- People with dyslexia and ADHD were being shut out of written advice
- Any solution had to update itself as emergency legislation changed week to week
Problem Definition and Strategy
How might we remove the barriers in how advice is delivered, so more people can get support earlier?
As soon as I joined I ran ideation with the team, which generated seven concepts, from interactive tools to visual guides to audio. The team had different priorities, so I set the criteria we would judge against: would it help vulnerable people quickly, could it cope with legislation changing weekly, and could it scale without someone rebuilding it by hand each time. Audio narration was the concept that met all three.
The solution
Audio narration tackled the barrier the research had found: that the text itself was the obstacle for the people who most needed the advice.
Choosing the voice
We didn't build a voice, we tested several off-the-shelf options. The guerrilla sessions were where it became clear which worked. Robotic voices broke comprehension and trust, and we landed on Speechify because people found it natural enough to stay with. Choosing the tool through testing, rather than on a spec sheet, was the point.
Speed control
People could slow the voice down or speed it up to match how they processed information, which mattered most for neurodiverse users.
Text highlighting
Highlighting the line being read helped people follow along and keep their place, a need the contextual sessions surfaced directly.
Streaming, not recordings
The technical decision I pushed hardest for. Pre-recorded audio would have had to be remade every time the emergency rent rules changed, which during COVID was close to weekly. Streaming the text-to-speech meant the audio updated automatically with the content. I set that up early, because the operational reality, not the demo, was what would decide whether it lasted.
Testing and Validation
Concept Validation
Low-fidelity prototypes tested remotely to check the core assumptions before any technical investment. People wanted a natural voice, needed speed control, and switched off at robotic narration.
Contextual Testing
The centre of the research. Guerrilla sessions across Dundee and Kirkcaldy libraries, with neurodiverse participants among them, showed people who found audio far easier to process than text, useful without glasses and while doing other things, and confirmed the need for text highlighting.
Pre-launch Validation
Final testing, with at least half identifying as neurodiverse, confirmed the tool was ready and that the highlighting worked as intended.
It's giving you advice and help and doing it in a way that's really easily understandable.
Launch
We launched on the rent arrears page first, chosen with the money and debt advisor and content team because need was high there and the content changed often, which tested the streaming approach properly. From there it expanded across the money and debt section.
Getting it out mattered. The funder had backed the brief and needed to see delivery, and after a long discovery the priority was a working, tested solution in front of real users, alongside research the wider team could use. Coordinating the launch meant working to the content designers' restructuring timelines, the service designer's view of the whole system, and the advisor's practical constraints.
Live Implementation
The tool launched on the rent arrears page and ran across the money and debt section, with people able to set the playback speed and follow the highlighted text through complex advice.
Outcomes and Impact
The proof was in the testing. Neurodiverse participants who had struggled to get through the written advice could take in the same information through audio. The contextual sessions showed it plainly: people who would have given up on the text stayed with the audio and understood it. For a tool built specifically for the people the written advice was failing, that was the result that counted.
The research had a second life beyond the tool. It fed into the content team's restructuring of the advice, which supported cutting the money and debt guidance from 123 pages to 36, and into accessibility work across other advice sections. In a project under pressure to deliver, getting a tested solution live and turning the research into wider content work was the contribution that mattered most.
I set the success measure as engagement quality rather than reach, since an accessibility tool serves specific needs rather than mass numbers. Early usage bore that out, with sessions averaging two to three minutes.
What I'd take from this
Method when the usual access is gone
When COVID made recruiting people into research impractical, guerrilla testing was what moved the project. Knowing when to stop waiting for ideal conditions and go to where people actually are changed how I approach research under constraint.
Knowing when to stop researching and start making
The project needed to move from discovery to a decision. Running the ideation that forced that choice, and setting the criteria it was made on, mattered as much as any single design.
A technical decision with operational weight
Streaming rather than recording was an accessibility choice that also solved a delivery problem, keeping the audio current as legislation moved. It is the kind of choice that decides whether a tool survives contact with real operations.
Designing for the highest needs lifts everyone
Building for the people the written advice was failing produced research that improved the advice for everyone, and coordinating across councils and a cross-functional team is what made that possible under pressure.