Inner Glow AIYaroslava

People who want to look after their mental health usually start strong and stop after a week. Mood trackers collect data but don't explain it, meditation apps don't know how you feel today, and therapy is expensive with long waiting lists. The result is a lot of logging and no clear answer to: am I getting better, and what should I do today?

How might we help people understand their mental state in seconds and give them one small, useful action every day?

Main finding: people stopped because the app felt like homework, the charts didn’t tell them anything, and paid content appeared before any value.

Who the site is for, what each group needs, and the page that answers it.

What a visitor does, thinks and struggles with at each step, and the design decision that answers it.

1. 01 Notice

   DOING

   Feels stressed or sleeps badly

   THINKING

   “I should look after myself.”

   PAIN POINT

   Five apps, none connected

   DESIGN RESPONSE

   Mood, sleep, stress and AI in one app

2. 02 Start

   DOING

   Takes the assessment

   THINKING

   “Is this worth my time?”

   PAIN POINT

   Paywall before any value

   DESIGN RESPONSE

   Value first, upgrade after real use

3. 03 Check in

   DOING

   Logs mood with one tap

   THINKING

   “I don’t have energy for this.”

   PAIN POINT

   Logging feels like homework

   DESIGN RESPONSE

   One-tap emoji check-in

4. 04 Understand

   DOING

   Sees the InnerGlow Score

   THINKING

   “Am I getting better?”

   PAIN POINT

   Charts don’t explain anything

   DESIGN RESPONSE

   One 0–100 score with a label

5. 05 Act

   DOING

   Picks a suggested action

   THINKING

   “What can I do today?”

   PAIN POINT

   Generic tips, no next step

   DESIGN RESPONSE

   3–4 actions with time needed

Eight hypotheses, each with its own metric, ordered by priority. Drawn as variants of one component: same structure, different bet.

❖ Hypothesis

1. H1

   One 0–100 score is easier to follow than separate mood, sleep and stress charts.

   METRICWeekly opens of score history
   High impact
2. H2

   A short assessment at the start makes the first result feel personal.

   METRICOnboarding completion 80%+, day-2 return
   High impact
3. H3

   3–4 suggested actions after each score turn logging into action.

   METRIC40%+ complete an action weekly
   High impact
4. H4

   A one-tap emoji check-in keeps daily logging under 30 seconds.

   METRICCheck-in time, 7-day streak rate
   High impact
5. H5

   Clear AI framing, with a route to a real professional, makes people trust the chat.

   METRICWeekly chat use, support link taps
   High impact
6. H6

   Smart reminders based on the user’s own patterns work better than fixed daily pings.

   METRICNotification open rate, opt-out rate
   Medium impact
7. H7

   Optional voice input helps people who find it hard to put feelings into words.

   METRICShare of voice entries
   Medium impact
8. H8

   Showing paid features after real use converts better than a paywall after onboarding.

   METRICTrial start rate, day-7 retention
   Medium impact

Each hypothesis placed by expected impact and the effort to build it. The score went first even as a big bet: every other feature depends on it.

1. H1One 0–100 score is easier to follow than separate mood, sleep and stress charts.
2. H2A short assessment at the start makes the first result feel personal.
3. H33–4 suggested actions after each score turn logging into action.
4. H4A one-tap emoji check-in keeps daily logging under 30 seconds.
5. H5Clear AI framing, with a route to a real professional, makes people trust the chat.
6. H6Smart reminders based on the user’s own patterns work better than fixed daily pings.
7. H7Optional voice input helps people who find it hard to put feelings into words.
8. H8Showing paid features after real use converts better than a paywall after onboarding.

The daily loop and the support path. Both end with one next step: an action, an exercise or a link to a professional.

Five tabs: Home, Practices, Journal and Profile, with Glow AI in the central button.

Low-fidelity layout first, agreed with the client, then the final interface.

→

12

34

1. 1

   Assessment

   14 questions, one per screen, with a progress counter and a skip option on sensitive topics.

2. 2

   InnerGlow Score

   One number from 0 to 100 with a plain-language label, and a history of what changed.

3. 3

   AI suggestions

   3–4 actions after each score — mindfulness, activity, social, professional support — with time needed.

4. 4

   Mood check-in

   One tap with 5 emoji states and an optional note. 12 seconds on average in testing.

5. 5

   AI companion

   Text and voice chat with clear limits. Support contacts always sit in the chat header.

6. 6

   Trade-off

   A single score can feel reductive, so details stay one tap away and every score has a label in words.

Key screens of the app.

Eight hypotheses across five stages of the journey. Targets were set before launch; the prototype test already gave a first signal for four of the five stages.