← All cases

Health · Mexico · 2020

Pfizer: a calmer path to mental wellbeing

A short check-in, a clearer picture of how you feel, and a gentle next step. A mental health app designed to help people recognise stress and find relevant support without feeling overwhelmed by the experience itself.

A woman taking a calm moment, with her hand over her heart

Health,
made simple.

Your everyday partner for feeling your best.

Case study draft. The original project was a freelance collaboration with Tres Prima for Pfizer in Mexico. This interactive walkthrough is a newly created design concept, not an original shipped experience. Check-in data is illustrative; the mood screen represents one step of the proposed questionnaire. Research findings, clinical validation and measured outcomes are not yet documented.

Problem

Stress is common, costly and hard to read

1 of 3

stressed adults worldwide

€100B+

annual stress sick leave cost (EU)

70–90%

of stressed people lost productivity

The problem

  • · No clear link between habits and stress
  • · Tracking alone doesn't help manage it

The solution

  • · Predict stress days from daily habits
  • · Personal guidance through a chatbot

Who it serves

Three audiences, one need

People, like us

Employers at scale

Insurance companies

The design journey

Design thinking, end to end

  1. 01

    Discovery

    Problem statement

    When stress feels overwhelming, tracking it is not enough.
    People need clarity.
    And a next step that feels manageable.

    The gapDaily habits and mental wellbeing feel disconnected.

    The opportunityA short check-in. A clear result. Relevant support.

  2. 02

    Ideate & prototype

    Strip the journey back to what matters: check in, understand, act, and ask for support.

    01Questionnaire
    02Test result
    03Dashboard
    04Chat

    Proposed flow · Simplified wireframes, not original project deliverables.

  3. 03

    Validate

    Can people finish the check-in, understand their result and find a helpful next step?

    01 / Comprehension

    Is the result clear?

    Check understanding without clinical jargon.

    02 / Effort

    Does it feel manageable?

    Observe friction across the short questionnaire.

    03 / Relevance

    Is support easy to find?

    Explore recommendations and the optional chat.

    Proposed UX validation plan. User research and clinical validation are not yet documented.

  4. 04

    Ship

    From flow to experience

    A calmer space.
    A clearer next step.

    Bring the check-in, tracking and personalised recommendations together. Keep support close, without adding pressure.

    Delivery concept shown here. Original release details and measured outcomes are not yet documented.

    9:41•••
    YOUR SPACE TO FEEL BETTER

    A little care,
    just for you.

    One small step at a time.

    TODAY'S CHECK-IN
    6/10Stress level
    A little elevated

    Let's start small.

    Your rhythm

    NOTICE YOUR PATTERNS

    Your rhythm

    One check-in at a time.

    You don't have to hold it all alone.

    A little support, when you need it.

    Let's talk
    Dashboard concept · Illustrative data

Data science

What drives stress

Stress model

CDC health survey

~450K

respondents

~380

health factors

Chatbot

PDFs from mental health experts

Top 10 contributing factors
Depression
Bad health
Decision difficulty
Sleep hours
Employment status
Alcohol frequency
Home ownership
Marital status
Sex
Age group

Relative feature importance

Data challenge

Most people report zero stress days

Around 190K rows sat at 0. Oversampling balanced the rare 30-day group.

before oversampling
0 days
1–29 days
30 days
after oversampling
0 days
1–29 days
30 days

Workflow

Two models, one experience

Machine learning

Dataset↓ Classification↓ Regression↓ Stress prediction
User info

⇄ both models

Chatbot (LLM)

Expert PDFs↓ Embed↓ TrainRAG modelFine-tuned

Zenly · Model evaluation

45% better predictions

Baseline

Dummy regressor · Simple regression

RMSE 9.72

Advanced · CatBoost

Classification → Regression

81% accuracy · RMSE 5.3

RAG model

“Deep breathing: take slow, deep breaths…”

General · Contextual

Fine-tuned model ✓

“Inhale for 4 seconds, hold for 4, exhale for 4. Repeat.”

Concise · Practical · Actionable

Chatbot evaluated with A/B tests. The fine-tuned model was chosen.

The data science work comes from Zenly, a related stress app with the same goal. The interactive app above is a design concept, not a shipped screen.

Impact · Illustrative example

Small steps. Meaningful engagement.

77.3%

completed the questionnaire

Among a sample of 320 people.

Example provided for the case study draft, not a verified Pfizer or Zenly outcome. Replace with confirmed results before presenting it as measured impact.

Want to explore the thinking behind it?

Let's talk about designing sensitive experiences with clarity, care and room for the person on the other side of the screen.

Get in touch