Registered devices
140
down 17%
Identity created. This starts the measured funnel, but it is not setup complete.
The outcome we want (lagging metrics)
The scoreboard. These tell us whether COPY is becoming a repeat behavior, not just a thing people try once.
Repeat conversations
Conversations people came back to, not just tried once.
48%
up 11 percentage points
Good
Repeat conversations
Conversations people came back to, not just tried once.
Current value
31 repeat conversations / 64 active conversations
Means: active conversations in the last 28 days that also had voice activity before this 28-day window divided by all active conversations in the last 28 days. Window: Jul 12, 2026 12:00 AM to Aug 8, 2026, 5:53 AM MST.
Read: Good: Repeat conversations is moving in the right direction (up 11 percentage points vs comparison).
Week-4 message-active retention
Shows whether newly registered devices keep finding value after the novelty window.
3%
up 1 percentage points
Watch
Week-4 message-active retention
Shows whether newly registered devices keep finding value after the novelty window.
Current value
registered devices in a signup cohort that were message-active four weeks after registration divided by registered devices in the same signup cohort. Window: mature signup cohorts only; newer cohorts show a dash until enough time has passed. Message-active means sent, played, or acknowledged a message.
Means: registered devices in a signup cohort that were message-active four weeks after registration divided by registered devices in the same signup cohort. Window: mature signup cohorts only; newer cohorts show a dash until enough time has passed. Message-active means sent, played, or acknowledged a message.
Read: Watch: Week-4 message-active retention is roughly steady or not decisive (up 1 percentage points vs comparison).
2+ people participated
Conversations are more valuable when another person actually joins in.
92%
up 5 percentage points
Good
2+ people participated
Conversations are more valuable when another person actually joins in.
Current value
59 conversations with 2+ people / 64 active conversations
Means: active conversations where at least two registered devices sent, played, or acknowledged a message divided by all active conversations in the last 28 days. Window: Jul 12, 2026 12:00 AM to Aug 8, 2026, 5:53 AM MST. This means two people participated; it does not necessarily mean both people sent messages back and forth.
Read: Good: 2+ people participated is moving in the right direction (up 5 percentage points vs comparison).
Invite reproduction
Measures whether accepted invitees later create their own invite within 24 hours.
28%
up 3 percentage points
Watch
Invite reproduction
Measures whether accepted invitees later create their own invite within 24 hours.
Current value
7 invitees who re-invited / 25 accepted invite devices
Means: registered devices that accepted an invite and then created another invite within 24 hours divided by registered devices that accepted an invite in the last 28 days. Window: Jul 12, 2026 12:00 AM to Aug 8, 2026, 5:53 AM MST. This is anonymized device-level behavior, not a person-level social graph.
Read: Watch: Invite reproduction is roughly steady or not decisive (up 3 percentage points vs comparison).
The signals we think predict it (leading metrics)
The levers. These are earlier behaviors we can improve now and test against the scoreboard later.
First conversation within 24h
Early channel creation is the first sign that setup turned into intent.
54%
flat
Watch
First conversation within 24h
Early channel creation is the first sign that setup turned into intent.
Current value
76 new devices with a conversation within 24h / 140 new registered devices
Means: newly registered devices that started a conversation within 24 hours of registration divided by newly registered devices in the last 28 days. Window: Jul 12, 2026 12:00 AM to Aug 8, 2026, 5:53 AM MST.
Read: Watch: First conversation within 24h is roughly steady or not decisive (flat vs comparison).
Invite events (activity ratio)
Raw invite event volume. The matched acceptance floor below is the conversion read.
14%
down 16 percentage points
Context only
Invite events (activity ratio)
Raw invite event volume. The matched acceptance floor below is the conversion read.
Current value
26 accepted invites / 189 created invites
Means: invite_accepted events divided by invite_created events. Window: Jul 12, 2026 12:00 AM to Aug 8, 2026, 5:53 AM MST. Activity, not conversion: creations fire before the share sheet opens and re-fire on re-issues, accepts are unmatched to their invite, and recent invites have not matured. Use the historical acceptance floor for conversion.
Read: Unknown: Invite events (activity ratio) needs more sample before reading the movement.
First playback milestone
Playback is the first evidence that voice reached the other side.
21%
down 11 percentage points
Needs focus
First playback milestone
Playback is the first evidence that voice reached the other side.
Current value
29 new devices with first playback / 140 new registered devices
Means: newly registered devices that reached the first-message-received milestone divided by newly registered devices in the last 28 days. Window: Jul 12, 2026 12:00 AM to Aug 8, 2026, 5:53 AM MST. This is a milestone proxy for first playback, not a complete message-by-message play rate.
Read: Needs focus: First playback milestone is moving the wrong way (down 11 percentage points vs comparison).
Second person participated
The strongest hunch for a durable conversation is another person participating.
92%
up 5 percentage points
Good
Second person participated
The strongest hunch for a durable conversation is another person participating.
Current value
59 conversations with 2+ people / 64 active conversations
Means: active conversations where at least two registered devices sent, played, or acknowledged a message divided by all active conversations in the last 28 days. Window: Jul 12, 2026 12:00 AM to Aug 8, 2026, 5:53 AM MST. This is participation, not a strict back-and-forth sender count.
Read: Good: Second person participated is moving in the right direction (up 5 percentage points vs comparison).
Historical acceptance floor (mature matched cohort)
Of fresh 1:1 invite channels created at least 7 days ago, the share that gained a second participant within 7 days. The channel and inviter floors only count creations made after second-participant instrumentation went live; earlier channels cannot be judged on an event that did not exist yet. This is a floor, not delivered-invite conversion: dismissed share sheets stay in the denominator. Full audit data ships beside this page as the invite-floor artifact.
The floors
| Channel floor (second participant, post-instrumentation cohort) | unavailable (sampled data; 9 of 98 raw, diagnostic only) |
|---|---|
| Corroborating floor (invite accepted, full window) | unavailable (sampled data; 100 of 479 raw, diagnostic only) |
| Inviter success floor (post-instrumentation cohort) | unavailable (sampled data; 8 of 54 raw, diagnostic only) |
| Fan-out failures (28d) | 0 |
| Redemption latency p50 / p95 (ok) | 384 ms / 2,975 ms |
Server redemption outcomes (28d)
| Outcome | Count | Share |
|---|---|---|
| ok | 54 | 100% |
Latency counts, shares, and percentiles are weighted by Analytics Engine sampling intervals (max interval 2).
Cohort audit waterfall
| Filter step | Rows | Weighted |
|---|---|---|
| All invite_created rows in window | 565 | 583 |
| Source is create_and_invite (excludes group add-member re-shares) | 497 | 515 |
| Channel type is 1:1 | 497 | 515 |
| Fresh token only (existing = 0 drops idempotent re-issues) | 497 | 515 |
| One member at creation | 497 | 515 |
| Deduplicated by anonymized channel prefix | 497 | 515 |
| Created at least 7 days before the reporting cutoff | 479 | 493 |
| Created after second-participant instrumentation went live (channel/inviter floors only) | 98 | 106 |
Duplicate creation rows dropped: 0. Channel prefixes seen under multiple inviters: 0. Sampling NOT verified (max interval 2).
This is a historical floor, not delivered-invite conversion: dismissed share sheets cannot be removed from the denominator with historical data.
Identities and channels are 8-character anonymized prefixes; collisions are detected and reported but cannot be repaired.
Analytics Engine sampling detected (max _sample_interval 2). Matched floors are unavailable: creation and outcome rows sample independently, so a missing outcome row is indistinguishable from a real non-conversion. Counts and weights are published for diagnosis only.
Onboarding / activation
The path from identity creation to real use. Registered device means an identity was created. It is not the same as fully completing onboarding.
Permission setup failures are a reliability guardrail. Lower is better; zero is the goal.
First conversation within 24h
54%
flat
76 devices started a conversation within 24 hours.
First message sent
28%
down 12 percentage points
39 new devices reached first-message-sent.
First playback
21%
down 11 percentage points
29 new devices reached first-message-received.
2+ people participated
92%
up 5 percentage points
Share of active conversations with at least two participating devices.
Permission setup failures
0
no failures
Coarse mic/push bridge failures. Lower is better; zero is the target.
What people do in the app
The last 28 days. Reach is the share of devices with a measured product session that did each action at least once.
101 devices across 1,064 measured sessions.
Who found the main customization surface. People who did this averaged 4.9 times.
Who moved from settings into appearance choices. People who did this averaged 3.7 times.
Who actually applied a different skin. People who did this averaged 2.9 times.
Who applied a different color variant. People who did this averaged 3.5 times.
Who moved the dial to another channel at least once. People who did this averaged 20.8 times.
Platform
Activity split by the platform prefix on updated app builds. Legacy rows stay visible instead of being guessed.
| Platform | Active devices | Sessions | Messages |
|---|---|---|---|
| iOS | 139 | 1,050 | 2,000 |
| Android | 3 | 14 | 15 |
Age of active devices
Days between registration and the latest measured session in this window. Limited to registrations in the 90-day report history.
| Device age | Active devices |
|---|---|
| 0-6 days | 31 |
| 7-13 days | 12 |
| 14-27 days | 16 |
| 28+ days | 24 |
Unlock progression
5 of 129 observed active identities completed all three (4%). Completion includes an earned achievement with no reward available, so progress is not mistaken for a missing claim.
| Unlock | Locked | Earned | Claimed | Completed |
|---|---|---|---|---|
| First Contact | 64 not earned | 34 ready to claim | 30 claimed a reward | 31 (24% complete) |
| Three's Company | 104 not earned | 18 ready to claim | 6 claimed a reward | 7 (5% complete) |
| Operator | 106 not earned | 14 ready to claim | 9 claimed a reward | 9 (7% complete) |
Usage over time
Weekly conversation activity. Default is 90 days; use shorter views to inspect recent movement.
0%
May 3
33%
May 10
58%
May 17
16%
May 24
29%
May 31
62%
Jun 7
68%
Jun 14
69%
Jun 21
57%
Jun 28
65%
Jul 5
53%
Jul 12
76%
Jul 19
82%
Jul 26
New this period2+ people participatedRepeat conversationsTop label is repeat conversation rate.
Good: Latest repeat conversation rate is moving in the right direction (up 6 percentage points vs comparison).
Cohort retention
Percent of each signup cohort that was message-active in later weeks. A dash means that week has not happened yet.
| Signup week | W0 | W1 | W2 | W4 |
|---|---|---|---|---|
| Jun 28-Jul 4 65 | 45% | 3% | 2% | 3% |
| Jul 5-Jul 11 59 | 27% | 8% | 0% | - |
| Jul 12-Jul 18 70 | 39% | 4% | 6% | - |
| Jul 19-Jul 25 40 | 23% | 3% | - | - |
| Jul 26-Aug 1 14 | 14% | - | - | - |
| Aug 2-Aug 8 16 | - | - | - | - |
Read each row left to right: Jun 28-Jul 4 had 3% of its signup devices message-active in week 4.
Hypothesis tracker
These are beliefs we are testing, not facts we should overstate.
| Hypothesis | Rationale | Evidence | Status |
|---|---|---|---|
| Groups retain better | Group channels create more chances for repeat voice use. | 14% of sends are in group conversations. | Mixed |
| Fast social activation matters | Channels become useful when another person joins quickly. | The mature matched invite floor is not yet quotable. The raw accept-to-create ratio is activity only, not conversion. | Unknown |
| Second participant predicts return | A conversation with two participating people is more likely to become repeat use. | 92% of active conversations had 2+ people participate. | Supported |
| Playback closes activation | A sent voice note matters more when someone hears it. | 21% first playback milestone rate. | Mixed |
Focus this week
Three practical reads from the current data. These are meant to guide product attention, not to be public copy.
1
Repeat conversations
Repeat conversations is currently marked good. Keep the work that supports this moving.
2
First playback milestone
First playback milestone is the clearest improvement area. Treat it as this week's product question.
3
First conversation within 24h
First conversation within 24h is watch. Watch it, but do not make it the whole story yet.
Social Pulls
Short lines Matt can send Alan or post as-is. They keep the numbers, but they should still sound like COPY and not like a dashboard.
48% of active COPY conversations in the last 28 days were repeat conversations. People are coming back to the same channel. That matters more than a one-off spike.
54% of new devices started a conversation within 24 hours. The first day still decides whether COPY feels alive or just feels like setup.
21% of new devices reached first playback. A sent note starts to matter when somebody actually hears it.
Metric glossary
Definitions used throughout this page. Counts are anonymized device or channel signals, not named people.
- Leading metric
- An earlier behavior we believe should predict a better product outcome later. These are hypotheses, not proof.
- Lagging metric
- An outcome that tells us whether COPY is already becoming useful and repeatable. It usually moves after the product experience changes.
- Percentage point
- The direct difference between two percentages. Moving from 20% to 30% is up 10 percentage points, not up 10%.
- Message-active identity
- An anonymized device identity that sent, played, or acknowledged a message during the window.
- Active conversation
- A conversation with voice activity during the window. This is a channel-level behavior, not a person count.
- Repeat conversation
- An active conversation that also had voice activity before the current window. This is the main repeat-behavior proxy.
- New conversation
- An active conversation with no earlier voice activity observed before the current week in this 90-day view.
- 2+ people participated
- A conversation where at least two anonymized identities sent, played, or acknowledged a message during the window. This is participation, not necessarily two-way sending.
- Cohort retention
- For each signup week, the share of those devices that were message-active in week 0, week 1, week 2, or week 4 after signup.
- WAU / MAU
- Weekly message-active identities divided by 28-day message-active identities. Higher means more of the recent audience is active this week.
- Registered device
- A device identity created on the server. This starts the measurable onboarding funnel, but does not prove the person finished setup.
- Onboarding permission failure
- A coarse mic or push permission bridge failure mirrored from client diagnostics. It excludes raw native error text and should be read as a reliability guardrail.
Archive policy
Pulse is the rolling live read. Weekly and monthly scorecards are frozen to completed Phoenix periods, with dated URLs so the comparison does not move after the period closes. The public archive is regenerated from recent Analytics Engine history each night; GitHub artifacts and Cloudflare deployments remain the deeper audit trail.