Quantitative, anonymized data · Timezone: Israel (Asia/Jerusalem) · Measurement window: 12/07/2026 – today
Cohort: signed up on/after 12/07 (pilot wave)Window: 12/07–19/07/20269 participants · 9 active in periodUpdated: 19/07/2026 18:32
9/9
Active participants in period
signup-wave participants · actually used
52
Total conversations (sessions)
sessions with interaction
8
Used more than once
89% of active
6
Returned on another day
67% retention
10.5 min
Average conversation length
median 3.6 min
16.3
User messages per conversation
27 incl. replies
2.2
Sessions per day per user
on active days
48
Conversations in the last week
7 participants · 13/07–19/07
Activity over time
Conversations per day (violet bars) and unique active participants per day (teal line).
Conversations/day Active participants/day
Distribution by hour of day
When participants are active (by conversation start hour, Israel time).
Distribution by day of week
Sunday → Saturday.
Conversation length
Distribution of conversation durations (minutes).
Conversation depth
Number of user messages per conversation.
Conversations per participant
Number of conversations (sessions) per participant, anonymized.
Language & modality
Participants by language, and voice-vs-text message split.
Participants by language
Messages: voice vs text
Per-participant breakdown (anonymized)
Each row = one participant. Sorted by number of conversations. No names/emails — anonymous identifiers only.
Participant
Joined
Language
Conversations
Active days
Conv./day
Total min
Avg conv. (min)
Messages
First
Last
P5
2026-07-13
English
12
6
2.0
48
4.0
346
13/07
18/07
P6
2026-07-13
English
8
4
2.0
78
9.8
112
13/07
16/07
P8
2026-07-13
Hebrew
8
3
2.7
239
29.9
71
13/07
15/07
P3
2026-07-13
Hebrew
8
3
2.7
89
11.1
184
13/07
17/07
P7
2026-07-13
Hebrew
6
5
1.2
35
5.8
46
13/07
19/07
P9
2026-07-13
Hebrew
4
1
4.0
8
2.1
16
16/07
16/07
P1
2026-07-12
English
3
1
3.0
23
7.6
25
12/07
12/07
P4
2026-07-13
Hebrew
2
2
1.0
22
10.8
45
13/07
14/07
P2
2026-07-12
English
1
1
1.0
3
2.8
3
12/07
12/07
Feature usage
Adoption of the app's built-in features among the 9 active participants. Features are triggered by the user through conversation with the avatar ("set me a daily reminder", "add a goal", etc.).
8/9
Used a feature ≥1
89% of active
8
Check-ins / reminders
31 set · 30 actually fired
5
Rituals
13 total
3
Goals
17 total
2
Circles
6 people
How many participants used each feature
Number of unique participants who used each feature (out of 9 active).
Features unused in the period: bookmarks, supervisor sharing, concern alerts — 0.
Check-ins: breakdown
Check-ins/reminders that were set, by type. All in voice mode.
8 participants set 31 check-ins, of which 30 actually fired (97%) — i.e. the feature was not just configured but ran and produced a proactive outreach by the avatar. 9 still active (daily recurring).
Per-participant feature matrix (anonymized)
✓ = how many check-ins actually fired. "Memories" = items the AI saved automatically about the user (not a user action). "Coverage" = how many of the 4 built-in features were used.
Participant
Check-ins
Rituals
Goals
Circles
AI memories
Coverage
P5
7 (6✓)
·
11
5
8
3/4
P7
6 (6✓)
5
2
·
2
3/4
P6
8 (8✓)
2
·
1
14
3/4
P9
3 (3✓)
·
4
·
2
2/4
P3
2 (2✓)
3
·
·
12
2/4
P8
2 (2✓)
1
·
·
8
2/4
P4
1 (1✓)
2
·
·
3
2/4
P1
2 (2✓)
·
·
·
13
1/4
P2
·
·
·
·
2
0/4
Circles: relationship types
People participants added to their "circles", by relationship type.
2 participants defined 6 people in circles. Early-stage usage — room to grow.
AI memories: the AI saved 64 memories about 9 of the participants (7 on average per active user) — an indication of the depth of familiarity the system builds over time (saved automatically, not a user action).
Summary & insights
What we see: of 9 external participants who signed up, 9 actually used the app. 8 of them (89%) opened more than one conversation, and 6 (67%) returned on a separate additional day — an indication of retention. Usage is mostly voice (85% of messages by voice). Average conversation length 10.5 min (median 3.6 min), with 16.3 user messages on average per conversation.
Features: 8 of the 9 active participants (89%) used at least one built-in feature. The standout is check-ins/reminders — 8 participants set 31 reminders and 30 of them actually fired, i.e. the avatar initiated a follow-up. Next are rituals (5 participants), goals (3) and circles (2). Bookmarks and supervisor sharing were not yet used — an opportunity to surface/encourage them in future versions.
Methodology (full transparency):
• Measurement window: all data in the report — conversations, durations, features and memories — is filtered to the range 12/07/2026 through today (inclusive), Israel time.
• POC cohort: external users (gmail addresses) who signed up on/after 12/07 — the pilot signup wave, 9 participants. Internal Essence team accounts, test/load accounts, and the personal dev account were excluded. Since all accounts were created from 12/07 onward, all their activity falls inside the window anyway. "Active in period" = 9 participants with at least one real conversation (user message).
• "Conversation"/"session": computed via sessionization — a run of activity separated by 30 min of silence starts a new conversation. This replaces counting raw "conversation" rows, which are inflated ~2.6× by the app's automatic reconnects.
• Core numbers count conversations with real interaction (at least one user message): 52 conversations. Additionally there were 7 "greeting-only" opens (the avatar spoke, the user did not reply) — not included in the core numbers.
• Conversation length: derived from timestamps (first to last message in the conversation); true voice-call duration is not stored as a dedicated field.
• All times in Israel time. The data is fully anonymized.
Solumea POC Analytics · generated automatically from production data · 19/07/2026 18:32
Anonymous identifiers P1–P9 · no names, emails, or conversation content.