Krishna Doshi
Work
About
Play
Status
Practicing magic
UNICEF Laaha - 1 in 3 women face Gender Based Violence (GBV). 1.5M now have a safe space to turn to.
Client
UNICEF
Year
2023
Status
Live
Website
Skills
Product design
UX research
Website design
Interaction design
The platform
Laaha ("for her" in Arabic) is UNICEF's virtual safe space — a web platform giving adolescent girls and women in humanitarian settings access to information on gender-based violence, sexual and reproductive health, mental health, digital literacy, and local support services. Built by QED42 in partnership with UNICEF's GBV in Emergencies team, it operates across crisis regions where physical safe spaces are inaccessible or unsafe.
I joined during the platform's second major iteration (v2.0) to lead the UX and UI redesign across three product areas, after initial versions of the forum, chatbot, and content library had already been built and tested in the field.
Laaha before v2.0: what was already built
The v1 platform had established three core features: a moderated user forum for peer support, a chatbot (Levi) that guided users to relevant content, and a content library covering SRH, GBV, mental health, and rights topics. Field teams had taken the platform to Venezuela and Lebanon for live testing. The platform confirmed the need was real. What the testing revealed was that the experience couldn't yet reliably serve that need, especially for users with the most to lose if something went wrong.
v1 screens of onboarding, homepage, chatbot and need help page
The users
These weren't edge-case users. They were the entire design target.
Who they are: adolescent girls aged 12 to 17 and adult women 18 to 45, across conflict-affected regions in Latin America, the Middle East, and Eastern Europe. Many are displaced. Many are in communities where discussing body autonomy, relationships, or GBV is actively dangerous. Most access information primarily through shared mobile phones, often not their own.
Laaha deployment countries
Their situation in numbers
How they access technology
Most users arrive on shared phones: a parent's, a sibling's, a neighbor's. In several Laaha countries, VPNs are a standard access method because government surveillance or household monitoring makes direct access unsafe. Browsing history is a real risk. An account linked to a real identity is a real risk. Session data that could be accessed by a caregiver is a real risk. Every design decision existed inside these constraints, not as a theoretical concern but as a documented user behavior confirmed across both testing rounds.
"If my father hears me listening to any of these topics, it can be bad for me and I don't want to be in that position."
~Lebanon user testing participant
"If I was not married, my parents would not allow me to access this. My mother would just ask me these questions."
~Lebanon user testing participant
"They don't want us to know anything, and the day after we are married, they expect us to know everything."
~Lebanon user testing participant
Statements from user testing team
What we found when we tested
Venezuela & Lebanon · Phase 1 testing
What broke and where
The forum was the most powerful part of the platform. In both Venezuela and Lebanon, women wrote into it with raw honesty; some cried while writing, sharing experiences they had never been able to tell anyone. That confirmed the product's core purpose was right. What the testing revealed was that getting to that point was nearly impossible for most users.
Only 1 to 2 participants per group (out of 5 to 7 per session) managed to register for the forum unaided. Older teenagers with higher digital skills still found the structure complex and discouraging
What I was asked to solve
The brief covered four product areas across three phases, with the scope defined by a prioritized estimates document (QED42 x UNICEF, Laaha v2.0):
P1 - Homepage personalisation + carousel
P1 - Service Hub page (Need Help redesign) + Geolocation-based services
P2 - Search enhancements
P2 - Chatbot enhancements (voice + text)
P2 - Audio for text
The work was phased against a single underlying constraint that structured every decision: the platform could not track users, log sessions, or store personal data without creating privacy risks for the people it was trying to protect. Personalization, retention, and engagement, normally the toolkit for improving a platform, were all unavailable in their conventional forms.
What we changed
1. Chatbot (Levi) · flow and response system
The problem: Levi's original flow was a decision tree with no back-navigation. If a user picked the wrong option or changed her mind, she had to start over. The flow also asked for age and menstruation history upfront, which girls with early marriages found invasive and refused.
What changed:
Added in-conversation back-navigation so users could revise answers without resetting, matching the mental model of standard messaging apps
Restructured the opening questions to ask about what the user wants to know rather than demographic identifiers
Added a voiceover layer for low-literacy users so content could be listened to rather than read
Clarified the chatbot's scope upfront: it is a content guide, not a reporting mechanism for violence (users in both rounds expected to be able to file reports through it)
Chatbot responsive flow
Skipped
Free-text input for all chatbot queries
Age-gating sensitive content
Geolocation to auto-load services
Instead
Structured choice-based flow with back navigation
Disclaimer-based approach
Manual country/region fallback
Why
Given low digital literacy, open text increased abandonment
Users resented age questions as gatekeeping; a disclaimer respects their agency
Girls using VPNs would be shown wrong or no services if geo-detection was the sole method
2. Registration and anonymous onboarding
The problem: The existing registration required choosing a username from an auto-generated list with no editing option, creating a complex password meeting character/case/number requirements, entering it twice, answering a security question, entering the answer twice, and accepting community guidelines, which then reappeared every time a question was posted.
What shipped:
Theme-based username selection: users could choose from animal, nature, or fictional-character-based names (e.g., "WindRay," "BlueLagoon") rather than accept an assigned random string. The names conveyed personality without revealing identity.
Passphrase instead of complex password: "IlovecatsandDogs!" is easier to remember and more secure than "Amber222222!". Password requirements removed.
Single terms-of-use acceptance at registration, not repeated per post
Anonymous registration path: no email, no phone number required; the system supported full anonymity for users who wanted it
Login, registration and anonymous usage registration. Along with user forum onboarding and user profile
"The workflow is not intuitive overall on the forum." ~ Venezuela participant
3. Need Help · service discovery without exposure
The problem: The service hub page required users to select a country, then a region, then a city from a long dropdown: a high cognitive-load sequence on a slow connection, for a user who may have arrived mid-crisis. Service categories used acronyms (GBV Case Management, MHPSS) that testing confirmed most users didn't recognize. And services listed contact information but had no in-platform contact mechanism, which was correct but wasn't explained.
What changed:
Categories restructured around user situation language rather than service-provider taxonomy ("I need someone to talk to" rather than "Psychosocial Support")
Multi-path UX flow for users in countries not yet in Laaha's coverage: change country, change region, change language, or access global resources rather than reaching a dead end
Permanent "Exit site" button, always visible, always fixed to the same position regardless of scroll state
Contact methods listed with a plain explanation of why in-platform messaging isn't available
Unique screens for need help flow
4. Homepage · dynamic without a user profile
The problem: The homepage was static. Returning users saw identical content on every visit. The fold hierarchy was unclear: services were repeated twice, the Discover section was too compact, and the "How to use Laaha" section was wall-to-wall text. A heuristic evaluation found 11 WCAG-aligned issues including missing emergency exit controls, non-functional article clickability signals, and footer accessibility failures.
The redesign: A six-fold narrative structure, with each fold carrying one job:
The "dynamic content without tracking" problem was solved through a rotating carousel that surfaced different modules on each visit without storing which modules a user had already seen. Country-specific defaults loaded from IP geolocation but with a visible, frictionless override control for VPN users.
To see the final design click on 'See live'.
Ideation and first exploration draft
Exploration drafts presented.
Impact
Platform-level metrics from UNICEF's public reporting show Laaha reached 1.5 million users across 15 countries and 13 languages by 2024. Phase-specific attribution wasn't tracked: the platform operates in humanitarian contexts where A/B testing with vulnerable users raises real ethical questions, and measuring "conversion" against a user who may be in active danger is not the right frame.
Registration friction: from 6+ steps with repeated terms acceptance → 3 steps, single acceptance
Chatbot scope clarity: from 0 explicit scope statements → upfront disclaimer, content guide framing, violence reporting redirect
Service discoverability: from acronym-based categories → situation-language categories with multi-path fallback for unsupported regions
Accessibility: 11 WCAG-flagged issues identified and addressed in homepage redesign
"This is just what I wanted as I was looking at the content: a place where I could find exactly what I was looking for." ~ Lebanon participant
Reflection
Literacy-matched copy first, not later. The disclaimer redesign, bringing the language down to the platform's actual reading level, was scoped as a Phase 2 polish item. In retrospect it is a Day 1 access problem: a user who can't parse the safety warning can't make an informed decision about whether to use the platform at all.
Test the exit button earlier. The "Leave this site" button stayed fixed throughout the redesign. In Lebanon testing, it disappeared on scroll in the mobile prototype, a discovery that happened late enough that the fix was reactive, not designed in from the start. For a safety tool, the exit is not a secondary feature.
More time on the VPN use case specifically. The geolocation vs. VPN tension shaped a significant portion of the Need Help page architecture, but the actual incidence of VPN use across Laaha countries wasn't quantified in the materials I received. A clearer data point here would have let us design the fallback more confidently rather than treating it as a known-but-unquantified risk.
Other Projects.
Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the.
Lorem Ipsum Lorem Ipsum Lorem Ipsum Lorem Ipsum Lorem ;)
PREV
NEXT
Lorem Ipsum lorem ipsum
Page
Work
Play
About
Say Hi!
Resume
Krishna Doshi