← All work
Mixed Methods Research Usability Testing Behavioral Analysis IA Design

Why employees don't use mental health benefits — and what changes when they do

10 min read

A research study into trust, privacy, and the hidden barriers that keep employees from accessing workplace wellness support.

What changed because of this research

I shifted the product direction from awareness nudges to privacy-first onboarding, top-level data controls, and warmer support flows.

Product WellNest
My role UX Researcher • Behavioural Research
Methods Screener survey, Semi-structured interviews, Competitive audit, Usability testing
Participants 11 screener, 3 interviews, 2 usability sessions
Timeline 2025
8/11 screener participants felt stressed "sometimes" or more — yet zero had used workplace mental health resources
100% of those who felt the need for support did not use employer-provided resources — awareness was not the problem
3 themes emerged across all interviews: privacy fear, trust failure, and transactional tone

The problem wasn't access. It was trust.

Employee Assistance Programs (EAPs) exist in most mid-to-large organisations. On paper, they solve the problem: employees get mental health support through their employer. In practice, usage rates are consistently low — often below 6% globally.

The common assumption is that the barrier is awareness. The assumption turned out to be wrong.

WellNest is a concept for a workplace wellness platform. Before any interface was designed, I wanted to understand why the existing EAP model was failing people — and what a trustworthy alternative would actually need.

Research questions

  1. Why do employees avoid seeking mental health support through their employer, even when they need it?
  2. What specific concerns prevent engagement with digital wellness platforms in a work context?
  3. What would a digital platform need to offer for employees to trust and use it?

A two-phase approach: broad first, then deep

I chose a mixed-methods approach — starting with a screener survey to identify patterns and recruit participants, then semi-structured interviews to understand the reasoning behind those patterns.

Phase 1

Screener Survey

11 respondents across tech, education, and content industries. Age range 25–44. Experience range 1–5+ years.

Goal: quantify stress prevalence, EAP usage, attitudes toward digital platforms, and recruit for interviews.

Why a screener first: I needed to avoid selection bias — only talking to people already interested in wellness tools would skew findings. The screener let me recruit across attitudes and experiences.
Phase 2

Semi-structured interviews

3 participants (45 min each): a senior software engineer, a transitioning engineer, and a product manager. Different stress levels, company sizes, and EAP exposure.

Goal: understand the why behind survey data — reasoning, emotion, past experience.

Why semi-structured: I needed flexibility to follow unexpected threads. What people don't say is as important as what they do — a rigid script would miss it.
Supplementary

Competitive audit

3 platforms analysed: Headspace for Work, MindPeers, YourDost — evaluated on privacy approach, accessibility, and design tone.

Goal: understand existing design and privacy conventions in this space, and identify where the market was failing users.
View participant profiles

*All participant names are pseudonyms.

P1 — Vikram
Software Engineer, 5+ years
Had used EAP once. Described it as "transactional, like a call center." Has not sought employer support since. Privacy concerns dominant.
P2 — Anya
Software Engineer, transitioning
Sought therapy independently — never through employer. Fear that EAP data could affect performance reviews or become workplace gossip.
P3 — Rahul
Product Manager
"Workplace and stress are synonymous to me." Never had EAP available. Wants open-source transparency to verify privacy claims independently.

From raw transcripts to actionable themes

Phase 1: Survey Insights (n=11)

Before diving into interviews, the screener survey surfaced a critical contradiction:

  • 8 out of 11 respondents felt stressed "sometimes" or more frequently.
  • 100% of those who felt the need for support had never used employer-provided resources.

This proved that the core problem wasn't awareness. People knew the support existed; they were actively choosing not to use it.

Phase 2: Interview Themes (n=3)

I used the 3 interviews to understand the why behind the survey data. After each interview I wrote a detailed record: key quotes verbatim, themes by section, an overall summary, and design implications. Three themes emerged consistently across the interviews:

01

Across these interviews, privacy emerged as the strongest barrier — ahead of stigma or cost

All three participants raised privacy concerns without being prompted. The concern followed a consistent pattern: what if my employer gets this data and uses it against me? This was especially notable with Rahul, who had never used an EAP — suggesting the concern can exist before any product contact at all.

"It's a privacy concern more than anything. I don't find it okay that my company can have access to my mental health records."

— Rahul, Product Manager

"What if they use this data against me? Anonymity can't be guaranteed in a small startup."

— Vikram, Software Engineer
Design implication Privacy assurance cannot be buried in a policy page. It must be present at first contact, repeated at sensitive touchpoints, and demonstrated through controls — not just words.
02

For one participant, a single EAP experience shaped ongoing avoidance

Vikram described using his company's EAP once — an experience he characterised as transactional and call-centre-like. In his account, he had not sought employer support since, even when it was available in later roles. That single interaction appeared to create an association between "workplace mental health support" and "impersonal bureaucracy" that persisted across jobs.

"It felt very transactional — like I was in a conversation with a call center representative. The whole experience felt formal and detached."

— Vikram, on his single EAP experience
Design implication Tone is not a cosmetic decision — it is a trust mechanism. Language, pacing, and microcopy need to signal warmth and human understanding. "Securing your slot" not "Loading." "Schedule a session" not "Book."
03

Within this study, participants consistently wanted control over their data — not just assurance

Across the three interviews, participants described wanting the ability to verify and control their data rather than simply being told it was safe. Anya listed very specific structural requirements: personal email access, external therapists, anonymous chats, and a hard delete option. Rahul described wanting open-source transparency so he could verify privacy claims independently. The pattern suggests that assurance alone is unlikely to be sufficient.

"I wouldn't want to use my office email for it. I'd want to use my personal email so there's no tie to the company directory. I'd want external therapists—people who aren't on the company payroll. And I need a way to delete everything. My chats, my account, all of it. Maybe even anonymous chats where I don't have to give my name at all."

— Anya, Software Engineer
Design implication A dedicated "My Privacy" section is not a nice-to-have — it is load-bearing for the product's credibility. Users need to see controls, not read promises.

Where participants diverged

Good research doesn't only report consensus. While privacy was a universal concern, other needs and contexts varied significantly, which prevented me from designing a "one-size-fits-all" solution:

  • Current stress levels: While Anya and Rahul were actively navigating workplace stress, Vikram wasn't particularly stressed anymore because he had become more experienced and built stronger boundaries over his 5+ year career.
  • Attitudes toward AI: Rahul actively wanted an AI assistant to help match him with the right therapist and navigate resources. Anya never mentioned AI and was highly skeptical of automated recommendations, preferring direct human selection.

Systematically evaluating what existing platforms get wrong

Why these three platforms

I selected platforms representing the main strategic approaches in the Indian and global workplace wellness market: Headspace for Work as the leading global B2C-to-B2B crossover, MindPeers as the dominant India-first B2B platform, and YourDost as an anonymity-led hybrid model. Together, they cover the three most common approaches to this product problem — which meant any shared weaknesses were unlikely to be accidental.

Evaluation criteria

Each platform was evaluated against five criteria derived directly from the research findings — specifically, what participants described needing in order to trust a product like this.

1
Privacy transparency Is data handling explained clearly before the user commits? Are privacy controls accessible, or buried in settings?
2
Employer / employee data separation Does the platform make explicit what employers can and cannot see? Is this communicated proactively?
3
Onboarding clarity Is first use low-friction and trust-building — or does it feel administrative and form-heavy?
4
Language & tone Does copy feel human and supportive, or transactional and clinical?
5
User control Can users delete data, manage communication preferences, and access privacy settings independently?
View platform findings and scorecard

Platform findings

Headspace for Work
B2C & B2B · Global
GDPR-compliant, anonymised org dashboards, calm visual language
Privacy buried three levels deep in settings — not surfaced at onboarding
Generic B2B tone; limited personalisation for individual users
Employer/employee data separation unclear without reading full privacy policy
MindPeers
B2B-first · India-centric
India-specific context, personalised user journeys, gamified engagement
Onboarding is cluttered — high cognitive load at a moment requiring reassurance
Privacy choices not prominent; data handling explanation absent from key screens
Inconsistent contrast ratio; accessibility issues on several key screens
YourDost
B2C & B2B · Hybrid
Anonymity-first positioning; strong independent therapist network
Most explicit of the three about data not being shared with employers
Outdated UI reduces perceived credibility and trustworthiness
Vague encryption practices — "secure" stated but not substantiated

Summary scorecard

Rated against each research-derived criterion: ✓ Meets · ~ Partially meets · ✗ Does not meet

Criterion Headspace for Work MindPeers YourDost
Privacy transparency ~ ~
Employer / employee separation ~
Onboarding clarity ~
Language & tone ~ ~
User control ~ ~
Audit conclusion: No platform fully met the criteria most important to research participants. The consistent gap — across all three products — was privacy transparency and user control. YourDost came closest on employer/employee separation, but its credibility is undercut by poor visual execution. Headspace for Work had the strongest tone and onboarding, but hid privacy controls. This suggests a clear design opportunity: a platform that treats privacy controls as a primary feature — not a settings afterthought — would be differentiated in this market.

How findings shaped every major design decision

The research produced a set of constraints and priorities that restructured the entire product — not just a feature list.

Research finding

Privacy fear can be pre-formed — participants distrusted employer-provided tools before opening one

Design response

Privacy statement on welcome screen (not buried in settings). Onboarding step 2 is entirely about data handling. Encryption notice persists in footer of sensitive screens.

Research finding

Transactional tone after one bad EAP experience created lasting avoidance for one user

Design response

Full microcopy audit. "Book a session" becomes "Schedule your session." "Loading" becomes "Securing your slot." Warm language applied as a design token, not an afterthought.

Research finding

Interviewees wanted to verify and control their data, not just be assured it is safe

Design response

"My Privacy" becomes top-level navigation — not a settings sub-menu. Contains: data usage explanation, delete account/data option, communication preference controls.

Research finding

Participants in this study needed immediate relief tools — not a booking flow — when stressed

Design response

"Quick Access" elevated to home dashboard with direct paths to guided breathing, SOS/crisis line, self-assessments. Therapy booking deliberately secondary to immediate tools.

Structure as a trust signal

The IA was built from the user journeys — specifically from Anya's (privacy-focused) and Vikram's (immediate-relief-focused) paths. The driving question for every structural decision: does this placement reduce or increase anxiety for someone who already doesn't trust this product?

View information architecture map
Home DashboardCentral entry · privacy summary · mood overview · quick access shortcuts
Quick Access
Self-HelpSelf-AssessmentsResourcesEmergency / SOSPsych First Aid
Book Sessions
Therapist DirectoryTherapist ProfileAvailability & BookingConfirmation
Mood Tracker
Log MoodHistory & TrendsRecommendations
My Privacy ★
Data UsageMy Data & ControlCommunication Prefs
AI Assistant
Therapist MatchingBooking HelpGeneral Info
Journals
Text (Doc)Visual (Paint)Audio Notes

★ My Privacy is top-level navigation — not buried in settings. This was a direct research decision, not a visual one.

Testing the prototype against real users

After the Figma prototype was built, I ran moderated usability tests with 2 participants (Anya and Vikram — both from the original interview pool, which let me compare their stated preferences to their actual behaviour).

Sessions were 45 minutes, screen-recorded, with a think-aloud protocol. I tested 4 scenarios: onboarding, booking a therapist, using the mood tracker, and creating a journal entry.

View usability issues by severity
IssueSeverityResolution
Privacy flow unclear on homepage — no visible data handling informationHighAdded privacy statement to homepage and dedicated onboarding step
Saving a journal entry threw user out of current screen entirelyHighReplaced redirect with inline toast notification — user stays in flow
No way to edit journal or session entries after savingHighAdded edit option via ellipsis menu, top-right placement
Mood input not intuitive — wheel felt oversimplified, fields unclearHighAdded text tag option alongside wheel, privacy hint on input field
Therapist booking had no step indicator — users did not know progressMediumAdded "Step 2 of 3" progress indicator to booking flow
Filter controls too dense — difficult to scanMediumGrouped filters logically, used dropdown chips
"Book a session" CTA felt robotic and clinicalMediumLanguage audit: "Schedule your session," warm tone throughout
What I would do differently: With more time, I would run a third usability session with a participant who had no prior connection to the research. Both Anya and Vikram had context about the project goals that may have softened critical feedback. An external participant would give cleaner evaluative data.

What the research changed — and what I would investigate next

What changed because of research

  • Privacy surfaced as the first screen users see — not the sixth
  • "My Privacy" added as top-level navigation — not in the original concept
  • External therapist option added after Anya specified she would only trust non-employer-affiliated support
  • SOS widget added after Rahul described needing emergency access without navigating menus
  • Entire microcopy tone shifted from functional to warm — direct result of Vikram's EAP story
  • 7 high-severity usability issues resolved in the second design iteration

What I would investigate next

  • Longitudinal study: do users who engage with privacy features actually use the platform more? The hypothesis is yes — but it needs data.
  • HR stakeholder interviews: what do employers actually see in aggregate dashboards? Employees assume the worst — understanding real data access could reshape the privacy communication entirely.
  • Expand sample: 3 interviews surfaces dominant themes but not edge cases. A fourth participant from a non-tech industry would stress-test whether findings generalise.
  • Diary study: one-week study tracking real-time emotional responses to onboarding and privacy flow — not retrospective recall.

Researcher's note

The most interesting finding was in the gap between the screener and the interviews. The screener showed 8/11 people felt stressed. The interviews showed none had used employer support. I expected interviews to surface "I didn't know it existed." They didn't. Every participant knew support was available. The barrier was entirely psychological — and it existed before any product contact. That means no amount of UX polish fixes this without first addressing the trust problem. The research changed the entire framing of the product.

Research documentation

I included this appendix for anyone who wants to see how I structured the study, recruited participants, shaped the discussion guide, and moved from raw data to themes.

View research appendix

Research objective

To understand why employees in mid-to-large organisations avoid using employer-provided mental health support — even when they are aware of it and experience work-related stress — and to identify what a digital wellness platform would need to offer for them to use it with confidence.

Participant criteria

  • Currently employed full-time in an organisation with 50+ employees
  • Reported experiencing work-related stress at least "sometimes" (screener filter)
  • Had not used employer-provided mental health resources in the past 12 months
  • Age range 25–44; mix of seniority levels
  • At least one participant with prior EAP exposure; at least one without
  • Recruited across tech, education, and content industries to avoid single-sector skew

Discussion guide excerpt

The full guide covered 4 sections across approximately 18 questions. The following is a representative excerpt from the core section on barriers and trust. Questions were deliberately open-ended to avoid leading participants toward expected answers.

  1. Can you walk me through a time when you felt stressed at work? What did you do with that?
  2. Have you ever considered using your employer's mental health support? What happened — or what stopped you?
  3. What would need to be true about a platform like this for you to feel comfortable using it?
  4. If I showed you that your data was fully encrypted and your manager couldn't access it — would that change anything for you? Why or why not?
  5. Is there anything about how I've described this that doesn't sit right with you?

Coding approach

I used inductive coding — starting from the data rather than a predetermined framework. After each interview I wrote a structured record: key verbatim quotes, initial codes by guide section, a summary of the session, and early design implications.

A code was treated as a theme only when it appeared independently across at least two of the three interviews. I also specifically tracked contradictions — moments where a participant's stated attitude conflicted with their described behaviour — as these often surfaced the most analytically interesting material.

Analysis workflow

1Interview → same-day structured write-up (quotes, codes, implications)
2Cross-interview comparison → identify codes appearing in 2+ sessions
3Contradiction check → flag where stated preference ≠ described behaviour
4Theme consolidation → 3 primary themes confirmed across all participants
5Implication mapping → each theme → specific design constraint or requirement

GalleryGo: understanding hesitation before the first bid →