1600×900 — key screen, product shot, or research-in-action photo
Client work · deliveredUX ResearchCROSocial Care
Case Study
Reducing drop-off for users seeking social care
Project walkthrough
Work in progress — this study is still being built
The client had lost confidence in their website and believed a redesign was the next step. The question I was brought in to answer wasn't how should we redesign this — it was does the evidence actually support a redesign, and if so, where?
Research snapshot
RoleSole researcher & designer — end-to-end
Duration6 weeks · 2 research rounds
TeamSolo, reporting to the service manager
Participants10 non-service users, ages 25–45 · 5 per round. Two rounds ran with different participants: the first identified comprehension and navigation issues in the existing experience, and those findings informed the redesign, which was then evaluated with a second group of five.
Research questionsWhy are users with clear intent leaving before making contact? What would make this provider feel trustworthy enough to call?
Constraints & limitationsLive site with no staging environment · small qualitative sample (n=5 per round), directional rather than statistically significant · no access to CRM or referral-source data.
Why this investigation matters
Impact
↓11%
Drop-off reduced
Problem
Weak value proposition
After 30 seconds, most participants could not describe what GoToHealthcare does or who it is for. Trust and Value scored Poor across Services, About and Careers.
Solution
Rewrote hero to lead with emotional benefit
Responds to: Weak value proposition
The tension
A redesign was on the table. Nobody could say why.
Redesign is the most expensive answer to a question nobody had tested yet.
GoToHealthcare was growing, but enquiry volume had plateaued and the site had stopped feeling like an asset. The assumption was that it needed rebuilding.
Why that assumption is expensive
A redesign commits budget and months of attention — and if the problem is behavioural rather than visual, you rebuild the same failure in a nicer typeface.
So the first job wasn't design
It was working out which problems were structural, and which could be solved with clearer language.
Who these users are
People arriving under pressure, often deciding for someone else. Every drop-off is a person who didn't get help.
Discover & Define
What the research pointed to
People were leaving on Services and About — not the contact form. Structural, not cosmetic, and not where a redesign would have started.
3/5
pages scored Poor
Weak value proposition
After 30 seconds, most participants could not describe what GoToHealthcare does or who it is for. Trust and Value scored Poor across Services, About and Careers.
68%
exit rate on Services
Exits happened before Contact
Users were exiting on Services, About and Careers — not on the contact form. Drop-off was structural, not a form problem. The Services page alone carried a 68% exit rate.
4/5
flagged jargon in Round 1
Jargon and no trust signals
Terms like "person-centred care pathways" read as clinical, not human. With no accreditations or testimonials visible, participants questioned legitimacy — "it could be anyone."
The change
Less content, better placed.
Some of this genuinely needed rebuilding. Much of it just needed saying differently.
HOMEPAGE — entry-point friction4 changes
Before — no clear value proposition
After — value, trust and CTA above the fold
What changed & why
01
Decorative hero, no proposition
Users could not state what the service does after 30 seconds — the headline described a feeling, not an offer.
02
Lead with the promise
"Care that feels like home" plus a one-line explanation of who it serves — emotional benefit before service description.
03
Accreditations surfaced
OFSTED Registered and CQC Rated Good moved above the fold — Round 2 participants cited these as the reason they would call.
04
Primary CTA before the scroll
"Book a consultation" sits in the first viewport with a secondary path for browsing services.
ABOUT PAGE — building credibility3 changes
Before — dense founder prose
After — story plus scannable credentials
What changed & why
05
Reader fatigue
Long unbroken prose blocks caused participants to skim past the information most relevant to their search.
06
Story kept, structure added
The founder's lived experience stays — the emotional differentiator — but leads with a headline rather than a wall of text.
07
Credentials as a scannable grid
Years of care, OFSTED, CQC and person-centred commitment surfaced as four tiles readable at a glance.
The decisions
Six changes, each answering something a participant said.
Six changes, sequenced by where the evidence said effort would actually pay back.
3 findings → 6 decisions → 3 measured outcomeseach change answers a specific research finding
D1← Weak value proposition
Rewrote hero to lead with emotional benefit
D2← 68% exit rate on Services
Restructured Services into scannable cards with CTAs
→ Largest single contributor to the drop-off reduction
D3← Trust scored Poor on 3 pages
Accreditations and testimonials above the fold
→ Trust rating 2.4 → 4.1; cited unprompted by 5/5 in Round 2
D4← No CTAs on 4 of 5 pages
Primary CTA on every page, before the scroll threshold
D5← Jargon flagged by 4 of 5
Plain-language rewrite across all pages
→ Comprehension time 68s → 22s
D6← Users could not visualise the service
Contextual photography of real carers
→ No measurable effect isolated — kept on qualitative feedback alone
Trade-offs
Why this solution, and what it cost
Sequencing
Which test runs first, and why
Risks
What could go wrong, and what would explain it
Influencing the decision
The hardest part wasn't the design. The client wanted more content on the homepage to explain every service upfront — the research showed users were already overwhelmed. I used the audit scores and the 68% Services exit rate to argue that less content, better structured, would serve people better. The data won that argument.
Measured Results
What changed — validated in Round 2
Validated through a second round of usability testing. Round 2 repeated the identical tasks with five new participants, shifting from diagnostic to confirmatory questions. Figures below are measured, not projected — but with n=5 per round they are directional, not statistically significant. They show a consistent shift, not a guaranteed effect size.
Measured
↓11%
Drop-off reduced across key user journeys after restructuring content and pathways.
Measured via: Behavioural drop-off data by page and journey, pre vs. post
Measured
68s → 22s
Time for a participant to understand what the service does and who it is for.
Average trust rating out of 5. Trust signals seen went from 0/5 participants to 5/5.
Measured via: Post-task trust rating, identical tasks with new participants
Next
What I'd research next
Longitudinal drop-off tracking to confirm the change holds beyond the initial post-launch window.
Sessions with actual service users and their families, not proxy participants.
Referral-source analysis to connect on-site behaviour to enquiry quality.
Reflection
Design rationale
LOOKING BACK
The real outcome
Not the redesigned pages — a defensible answer to which parts actually needed rebuilding, and which just needed to say something clearer.
What the evidence split
Service pages had a structural problem and needed the rebuild. The trust gap didn't — accreditation existed, just nowhere near the decision. A placement fix moved the number most.
What I nearly missed
It would have been easy to accept the redesign framing; it was the work I was brought in to support. The first solution is rarely the right one.
What I'd do differently
Get moderated sessions with service users earlier, and agree the measurement window before starting.
Every project raises questions it doesn't answer. This page covers the decisions that shaped the outcome — it can't capture the stakeholder conversations, the directions we tested and abandoned, or the trade-offs that never made it into a slide. That's usually where the most useful learning sits.
How I made the case for testing before rebuildingWhich parts I concluded genuinely needed a redesign — and which didn'tWhat we cut that I still think aboutWhy two rounds instead of one bigger studyWhat I'd measure differently with another month
