Biggest drop: Account created to Paywall accepted (£0) (100% lost)
Counts since the beta opened, real users only. Each app step is a subset of the one above. The web-to-app step is directional: it depends on PostHog stitching the web visitor to the app account. Active in week 2 counts only testers whose second week has finished.
Internal-data synthesis across Vercel, PostHog, GSC, MailerLite, Sentry, Adapty, Linear, Convex, and BigQuery. Anomalies and Wins reference today vs the trailing 14-day window. Written 07:00 UK by the analytics-cockpit Routine.
Snapshot fetched at 07:11 UTC (first attempt timed out at 60s, succeeded on retry in 26.5s). Prior briefings: Nov 20 (anomaly_count: 2, action_count: 7), Nov 19 (2, 7), Nov 18 (0, 5), Nov 17 (0, 5).
Day 20 of the public beta (opened 2 November 2026).
"CSRF token or cookie absent on unsafe request" (136893534) reads 20 events_7d and `users_affected: 20`, last seen `2026-11-21T04:46:03Z`. This is the same saga that moved 0 → 8 → 14 → 20 across the last four readings — today's +42.9% move on `users_affected` keeps the run going rather than levelling off as it has repeatedly since October. Four consecutive daily readings of growing real-user impact, with nothing in today's engineering batch (see Win 1) touching it, is the strongest case yet for a code-level look rather than waiting for it to go quiet.
`cod_backlog` rises to 94 from 65 (+44.6%) — bigger than 20 November's record 18.75% drop, just in the opposite direction. Six COD issues completed in a tight batch at 06:42:43–47 UTC today (see Win 1), well short of what would explain a 29-item net rise, so the move reads as a filing spike rather than a stalled queue: `created_last_7d` rises to 165 from 140 (+17.9%) while `completed_last_7d` eases slightly to 99 from 103 (−3.9%). Nothing in the snapshot names what generated the new tickets — a bulk triage or sibling-sweep pass is the obvious hypothesis given the volume, but that's a hypothesis, not a confirmed cause.
Six COD issues complete within 4 seconds of each other (06:42:43–47 UTC). Two are priority 1/2 and materially real: COD-5650 — the GDPR export cleanup cron deleted any blob with content type `application/json`, so a user who attached a `.json` document lost the file within 24 hours; COD-5672 — only 2 of the 6 community screens carried the session-replay no-capture label, so the DM inbox and topic lists would have been recorded despite the policy stating messages are hidden in full. The other four (COD-5673, COD-5668, COD-5648, COD-5647) are backend search-indexing and routing fixes.
Nine items today, up from seven on both 19 and 20 November — now well past the five-item threshold this brief treats as a sign the backlog of standing decisions is building rather than clearing.
1. Decide whether to formally triage and close the backend "Not authenticated" cluster and UNREAD_COUNT_DENIED now that both have been quiet for a sixth day, rather than leaving 10 issues unresolved on the chance they resume
→ App Health · Sentry 128651664
2. Take a code-level look at the web CSRF warning now — `users_affected` has grown for a fourth straight day (0 → 8 → 14 → 20) and nothing in today's engineering batch touches it
→ App Health · Sentry 136893534
3. Confirm why `cod_backlog` jumped 44.6% in a day (65 → 94) — the largest single-day move this briefing has recorded in either direction — and whether it reflects a bulk filing pass or a real quality regression
4. Investigate the Adapty profile-fetch-timeout error on mobile (109003255) — events more than doubled week-on-week (13 → 28) and it's now the largest mobile Sentry issue by volume, with 5 users affected; Adapty governs paywall/subscription state so this could be touching onboarding, not just logging noise
| Surface | Metric | Value | vs Prior (20 Nov) |
|---|---|---|---|
| Beta | Testers | 23 | 23 (flat) |
| Beta | Weekly active | 1 of 23 (4.3%) | 1 of 23 (4.3%, flat) |
| Mobile | Crash-free sessions (7d) | 100% | 100% |
| Source | Status | Notes |
|---|---|---|
| Cockpit snapshot endpoint (`/api/snapshot/all`) | ⚠️ ok on retry | First attempt timed out (504, FUNCTION_INVOCATION_TIMEOUT) at 60s; second attempt succeeded at 07:11 UTC in 26.5s |
| Sentry (mobile, web, backend) | ✅ ok | All three projects returned; mobile's Adapty-timeout issue more than doubles, see Anomaly 3; web's CSRF warning escalates for a fourth day, see Anomaly 1; web `sessions_7d` (performance monitoring) drops 24.8% while PostHog unique visitors hold flat — reads as a sampling artefact in Sentry's own session tracking rather than a real traffic fall, but noted here rather than silently dropped; backend keeps decaying out of the rolling window as the frozen cluster ages |
Snapshot fetched at 07:11 UTC, single request succeeded (6.5s). Prior briefings: Nov 19 (anomaly_count: 2, action_count: 7), Nov 18 (0, 5), Nov 17 (0, 5), Nov 16 (3, 6).
Day 19 of the public beta (opened 2 November 2026).
Markers say what happened on a day, not what caused a change.
"Error: Adapty profile fetch timed out after 8000ms" (109003255) reads 28 events_7d against 13 in the prior 7-day window (+115.4%), `users_affected: 5`, last seen `2026-11-21T01:17:29Z`. It's now the single largest issue in mobile's top-issues list by event volume, ahead of the already-flagged "onError" issue (see Watch). Adapty handles subscription and paywall state, so a timing-out profile fetch is a plausible contributor to onboarding friction rather than cosmetic noise — worth a look given the beta's activation numbers are already soft.
5. Triage the "onError" mobile issue (161195809) — now 26 events_7d (vs 16 yesterday), still 5 users affected, still no clue what's actually failing after two days of growth
6. Decide whether Adapty's rate limit is a permanent constraint — today's reading climbs to 4 of 11 fields `ok: true` from 3 of 11, the same partial-recovery pattern that has reverted every time since 12 November
7. Confirm whether the monthly PropertyData bulk-refresh's 18.2% API-call failure rate (2,360 of 12,990 calls, 13 November) left any of the 1,776 processed postcode districts with stale data
Unconfirmed since first flagged on 14 November (now 7 days); the cron still reports `fullRunComplete: true` and outcome `success` despite the failure rate, and no run has superseded it.
→ CCI
8. Confirm status on COD-2792, now 175 days old and still the oldest open COD issue, alongside DEV-196 at 311 days, the oldest open DEV issue
9. Wire Meta/Google Ad Spend credentials, still unset
`META_ACCESS_TOKEN`/`META_AD_ACCOUNT_ID` and the Google Ads credential set remain unset, with no attribution data for paid channels.
| Mobile | Sessions (7d) | 136 | 136 (flat) |
| Mobile | Sentry events (7d) | 57 | 73 (−21.9%) |
| Mobile | Unresolved issues | 6 | 6 (flat) |
| Web | Sentry events (7d) | 53 | 47 (+12.8%) |
| Web | Sessions (7d) | 413 | 549 (−24.8%, see Source Health) |
| Web | Unresolved issues | 4 | 4 (flat) |
| Backend | Sentry events (7d) | 65 | 73 (−11.0%) |
| Backend | Unresolved issues | 10 | 10 (flat) |
| App Health | Axiom `tasks:listByUserPaginated` (21 Nov, partial) | 224 | 222 (20 Nov, partial) |
| Marketing | Web pageviews (7d) | 106 | 110 (−3.6%) |
| Marketing | Unique visitors (7d) | 47 | 48 (−2.1%) |
| Marketing | GSC clicks (7d) | 8 | 9 (−11.1%) |
| Marketing | GSC impressions (7d) | 2,933 | 2,326 (+26.1%, see Watch) |
| Marketing | Avg position | 17.73 | 16.64 (+6.6% worse, see Watch) |
| Marketing | Email subscribers | 10 | 10 (unchanged) |
| Engineering | COD backlog | 94 | 65 (+44.6%, see Anomaly 2) |
| Engineering | Completed last 7d | 99 | 103 (−3.9%) |
| Engineering | Created last 7d | 165 | 140 (+17.9%) |
| Engineering | DEV in-flight | 1 | 1 (flat) |
| Engineering | BET open | 0 | 0 (flat) |
| Engineering | High priority open | 92 | 91 (+1.1%) |
| Population | Total users (BigQuery mirror) | 27 | 27 (flat) |
| Population | Reported spend (app) | £18,369 | £18,369 (unchanged) |
| Monetisation | MRR | £0 | £0 (data: 4 of 11 fields ok, up from 3 of 11, see Watch) |
| CCI | Drift queue | 0 pending | 0 pending |
| Spend | Total monthly (projected) | £536.13 | £538.35 (−0.4%) |
| Axiom | ✅ ok | Call volumes and error rate returned; no cron failures today; `tasks:listByUserPaginated` continues in a normal partial-day range |
| PostHog (Marketing, Beta) | ✅ ok | Beta holds flat; marketing traffic eases back slightly, within normal range |
| GSC | ✅ ok (lag) | The usual ~2-day ingestion lag; impressions jump, clicks and position both ease the wrong way, see Watch |
| MailerLite | ✅ ok | 10 subscribers; automations still 0 activated |
| Ad Spend (Meta/Google) | ⚠️ stale/not configured | Credentials still unset, see Action 9 |
| Adapty | ✅ ok, partial | 4 of 11 fields `ok: true` (revenue, mrr, arr, refund_events), up from yesterday's 3 of 11, see Watch and Action 6 |
| Linear (Engineering) | ✅ ok | Returned successfully; a sixth sibling-fix batch lands including a GDPR data-loss fix and a privacy gap, see Win 1; cod_backlog jumps unusually, see Anomaly 2; COD-2792 reaches 175 days, DEV-196 reaches 311 days, see Action 8 |
| BigQuery (Population) | ✅ ok | Returned successfully; reads 27, tracking the live tester count with its usual small lag |
| Convex (CCI) | ✅ ok | Drift queue empty, submissions queue empty; no new cron runs since 15 November's clean ONS run; the monthly PropertyData bulk-refresh's failure rate remains unconfirmed, see Action 7 |
| Spend | ✅ ok | All rows returned; projected £536.13/month, down slightly on 20 November |
"CSRF token or cookie absent on unsafe request" (136893534) reads 14 events_7d, `users_affected: 14`, last seen `2026-11-20T06:18:45Z` — under an hour before this snapshot. This is the same saga flagged yesterday when `users_affected` moved from a multi-week run of zero to 8; today it moves again, from 8 to 14 (+75%). Three consecutive daily readings of growing real-user impact (0 → 8 → 14) is a materially different shape from the "quiet, then a blip" pattern this issue showed through October and early November, and nothing in today's engineering batch (see Win 1) touches it. Escalating from yesterday's 🟡.
A fresh Sentry issue titled simply "onError" (161195809) appears for the first time this morning: 16 events_7d, `users_affected: 5`, last seen `2026-11-20T01:18:51Z`. It's the single largest contributor to mobile's total event count moving to 73 from 58 (+25.9%), a day-over-day jump that clears the 20% band on its own. The bare "onError" title gives no clue what's failing — this needs a code-level look before it's dismissed as noise, since it already outranks several named issues on this dashboard by volume.
Ten COD issues complete across two tight clusters — six between 01:16–01:31 UTC (a grouped sweep closing out raw-millisecond date arithmetic and Convex-Request-ID correlation gaps) and four more at 06:47 UTC. The standout is COD-5643 (priority 0): the mobile job-duration estimate counted wall-clock time as days rather than working days, so a genuine two-day job was quoted at 8 hours instead of 16 — a real pricing-confidence bug, not cosmetic. `cod_backlog` falls to 65 from 80 (−18.75%), the largest single-day drop recorded here. The rolling `completed_last_7d` figure actually eases to 103 from 132 (−22%) over the same move — the trailing 7-day window dropped an older high-volume day as today's batch landed, so the live backlog count is the number to trust, not the rolling rate.
Seven items today, unchanged in count from 19 November.
1. Decide whether to formally triage and close the backend "Not authenticated" cluster and UNREAD_COUNT_DENIED now that both have been quiet for a fifth day, rather than leaving 10 issues unresolved on the chance they resume
2. Take a code-level look at the web CSRF warning now — `users_affected` has grown for a third straight day (0 → 8 → 14) and nothing in today's engineering batch touches it
3. Triage the new "onError" mobile issue (161195809) — 16 events_7d, 5 users affected, appeared overnight, and the title gives no clue what's actually failing
4. Decide whether Adapty's rate limit is a permanent constraint — today's reading climbs to 3 of 11 fields `ok: true` from 1 of 11, the same partial-recovery pattern that has reverted every time since 12 November
5. Confirm whether the monthly PropertyData bulk-refresh's 18.2% API-call failure rate (2,360 of 12,990 calls, 13 November) left any of the 1,776 processed postcode districts with stale data
Unconfirmed since first flagged on 14 November (now 6 days); the cron still reports `fullRunComplete: true` and outcome `success` despite the failure rate, and no run has superseded it.
→ CCI
6. Confirm status on COD-2792, now 174 days old and still the oldest open COD issue, alongside DEV-196 at 310 days, the oldest open DEV issue
7. Wire Meta/Google Ad Spend credentials, still unset
`META_ACCESS_TOKEN`/`META_AD_ACCOUNT_ID` and the Google Ads credential set remain unset, with no attribution data for paid channels.
| Surface | Metric | Value | vs Prior (19 Nov) |
|---|---|---|---|
| Beta | Testers | 23 | 23 (flat) |
| Beta | Weekly active | 1 of 23 (4.3%) | 1 of 23 (4.3%, flat) |
| Mobile | Crash-free sessions (7d) | 100% | 100% |
| Mobile | Sessions (7d) | 136 | 137 (−0.7%) |
| Mobile | Sentry events (7d) | 73 | 58 (+25.9%, see Anomaly 2) |
| Mobile | Unresolved issues | 6 | 7 (−14.3%) |
| Web | Sentry events (7d) | 47 | 41 (+14.6%, see Anomaly 1) |
| Web | Sessions (7d) | 549 | 507 (+8.3%) |
| Web | Unresolved issues | 4 | 4 (flat) |
| Backend | Sentry events (7d) | 73 | 72 (+1.4%) |
| Backend | Unresolved issues | 10 | 10 (flat) |
| App Health | Axiom `tasks:listByUserPaginated` (20 Nov, partial) | 222 | 456 (19 Nov, partial) |
| Marketing | Web pageviews (7d) | 110 | 110 (flat) |
| Marketing | Unique visitors (7d) | 48 | 48 (flat) |
| Marketing | GSC clicks (7d) | 9 | 11 (−18.2%) |
| Marketing | GSC impressions (7d) | 2,326 | 2,706 (−14.0%) |
| Marketing | Avg position | 16.64 | 15.98 (+4.1% worse, see Watch) |
| Marketing | Email subscribers | 10 | 10 (unchanged) |
| Engineering | COD backlog | 65 | 80 (−18.75%, see Win 1) |
| Engineering | Completed last 7d | 103 | 132 (−22.0%) |
| Engineering | Created last 7d | 140 | 152 (−7.9%) |
| Engineering | DEV in-flight | 1 | 1 (flat) |
| Engineering | BET open | 0 | 0 (flat) |
| Engineering | High priority open | 91 | 94 (−3.2%) |
| Population | Total users (BigQuery mirror) | 27 | 27 (flat) |
| Population | Reported spend (app) | £18,369 | £18,369 (unchanged) |
| Monetisation | MRR | £0 | £0 (data: 3 of 11 fields ok, up from 1 of 11, see Watch) |
| CCI | Drift queue | 0 pending | 0 pending |
| Spend | Total monthly (projected) | £538.35 | £539.60 (−0.2%) |
| Source | Status | Notes |
|---|---|---|
| Cockpit snapshot endpoint (`/api/snapshot/all`) | ✅ ok | Single request succeeded, 07:11 UTC, 6.5s |
| Sentry (mobile, web, backend) | ✅ ok | All three projects returned; mobile gains a new high-volume "onError" issue, see Anomaly 2; web's CSRF warning escalates for a third day, see Anomaly 1; backend keeps decaying out of the rolling window as the frozen cluster ages |
| Axiom | ✅ ok | Call volumes and error rate returned; no cron failures today; `tasks:listByUserPaginated` continues in a normal partial-day range |
| PostHog (Marketing, Beta) | ✅ ok | Beta holds flat; marketing traffic and visitors both flat day-over-day |
| GSC | ✅ ok (lag) | The usual ~2-day ingestion lag; clicks and impressions both ease back, position keeps drifting worse, see Watch |
| MailerLite | ✅ ok | 10 subscribers; automations still 0 activated |
| Ad Spend (Meta/Google) | ⚠️ stale/not configured | Credentials still unset, see Action 7 |
| Adapty | ⚠️ error (429), partial | 3 of 11 fields `ok: true` (revenue, mrr, arr), up from yesterday's 1 of 11, see Watch and Action 4 |
| Linear (Engineering) | ✅ ok | Returned successfully; the largest sibling-fix batch yet lands, including a real mobile estimate bug, see Win 1; COD-2792 reaches 174 days, DEV-196 reaches 310 days, see Action 6 |
| BigQuery (Population) | ✅ ok | Returned successfully; reads 27, tracking the live tester count with its usual small lag |
| Convex (CCI) | ✅ ok | Drift queue empty, submissions queue empty; no new cron runs since 15 November's clean ONS run; the monthly PropertyData bulk-refresh's failure rate remains unconfirmed, see Action 5 |
| Spend | ✅ ok | All rows returned; projected £538.35/month, flat on 19 November |
Snapshot fetched at 07:12 UTC, single request succeeded (7.5s). Prior briefings: Nov 18 (anomaly_count: 0, action_count: 5), Nov 17 (0, 5), Nov 16 (3, 6), Nov 15 (3, 6).
Day 18 of the public beta (opened 2 November 2026).
"CSRF token or cookie absent on unsafe request" (136893534) reads 8 events_7d against 1 in the prior window (+700%), last seen `2026-11-19T02:50:14Z`, under five hours before this snapshot. What's new isn't the volume — this issue has touched 44 events_7d at its peak back in early November — it's `users_affected: 8`. Every prior reading of this saga this briefing has recorded carried `users_affected: 0`, consistent with an instrumentation or bot-traffic effect rather than a real visitor hitting it. Eight distinct users is a different shape of signal, and worth a code-level look rather than waiting for it to go quiet again as it has repeatedly since October.
`high_priority_open` rises to 94 from 83 (+13.3%). The four prior daily readings (86, 84, 83, and now 94) moved by only 1–3 items a day, so an 11-item jump breaks that pattern even though it sits just under the 20% band. `created_last_7d` also ticks up to 152 from 148 (+2.7%) while `completed_last_7d` eases slightly to 132 from 134 (−1.5%), which is consistent with more being filed than cleared today rather than a backlog re-count. Two P0s completed this morning (Win 1) could plausibly have generated sibling tickets, but nothing in the snapshot names what moved into the high-priority set.
Two COD issues complete within a second of each other (06:42:32–06:42:33 UTC): COD-5636 closes a gap in COD-5600's fix — the web community search-drawer query was still reaching Sentry unscrubbed through the navigation breadcrumb and request URL, a real PII leak rather than a cosmetic one; COD-5635 fixes a mobile DIY composite-job estimate that kept the 12.5% contractor materials margin the DIY task route is supposed to strip, meaning DIY users were quoted a contractor-inflated price. Both are priority 0.
GSC impressions_7d rises to 2,706 from 2,139 (+26.5%) and clicks_7d to 11 from 9 (+22.2%) — both crossing the favourable 20% band. Average position worsens slightly over the same window, to 15.98 from 14.00 (+14.1% worse), so the extra visibility is coming at least partly from lower-ranking queries rather than uniformly better rankings. "bathroom cost calculator" (position 4.6, 2 clicks, 121 impressions) and "honely" (position 2.6, 2 clicks, 34 impressions) remain the strongest-ranking queries.
Seven items today, up from five on 18 November — more than five, which is a signal that the backlog of standing decisions is building up rather than being cleared.
1. Decide whether to formally triage and close the backend "Not authenticated" cluster and UNREAD_COUNT_DENIED now that both have been quiet for a fourth day, rather than leaving 10 issues unresolved on the chance they resume
→ App Health · Sentry 128651664
2. Take a code-level look at the web CSRF warning now that it has affected 8 distinct users for the first time in its multi-week saga, rather than waiting for it to go quiet again
Every prior reading showed `users_affected: 0`; today's does not.
→ App Health · Sentry 136893534
3. Check Linear for what moved `high_priority_open` from 83 to 94 in a single day — the snapshot doesn't name which issues entered the high-priority set
4. Decide whether Adapty's rate limit is a permanent constraint — today's reading fell back from 2 of 11 fields `ok: true` to 1 of 11, the same revert pattern seen since 12 November
5. Confirm whether the monthly PropertyData bulk-refresh's 18.2% API-call failure rate (2,360 of 12,990 calls, 13 November) left any of the 1,776 processed postcode districts with stale data
Unconfirmed since first flagged on 14 November (now 5 days); the cron still reports `fullRunComplete: true` and outcome `success` despite the failure rate.
→ CCI
6. Confirm status on COD-2792, now 173 days old and still the oldest open COD issue, alongside DEV-196 at 309 days, the oldest open DEV issue
7. Wire Meta/Google Ad Spend credentials, still unset
`META_ACCESS_TOKEN`/`META_AD_ACCOUNT_ID` and the Google Ads credential set remain unset, with no attribution data for paid channels.
| Surface | Metric | Value | vs Prior (18 Nov) |
|---|---|---|---|
| Beta | Testers | 23 | 23 (flat) |
| Beta | Weekly active | 1 of 23 (4.3%) | 1 of 23 (4.3%, flat) |
| Mobile | Crash-free sessions (7d) | 100% | 100% |
| Mobile | Sessions (7d) | 137 | 144 (−4.9%) |
| Source | Status | Notes |
|---|---|---|
| Cockpit snapshot endpoint (`/api/snapshot/all`) | ✅ ok | Single request succeeded, 07:12 UTC, 7.5s |
| Sentry (mobile, web, backend) | ✅ ok | All three projects returned; web's CSRF warning spikes with real user impact for the first time, see Anomaly 1; mobile gains a new "App Hanging" issue, see Watch; backend keeps decaying out of the rolling window as the frozen cluster ages |
| Axiom | ✅ ok | Call volumes and error rate returned; no cron failures today; `tasks:listByUserPaginated` continues in a normal partial-day range |
Snapshot fetched at 07:11 UTC, single request succeeded (39.6s). Prior briefings: Nov 17 (anomaly_count: 0, action_count: 5), Nov 16 (3, 6), Nov 15 (3, 6), Nov 14 (2, 5).
Day 17 of the public beta (opened 2 November 2026).
No anomalies — all tracked metrics inside trailing 14-day bands.
Pageviews rise to 113 from 70 (+61.4%), unique visitors to 50 from 31 (+61.3%). The bathroom calculator remains the single biggest page (47 of 113 views, 41.6%, with 8 of the week's 9 calculator completions against it), but the new growth sits elsewhere: `utm_source=instagram` logs 20 visits — the first Instagram-attributed traffic this briefing has recorded — and `m.facebook.com`/`www.facebook.com` referrers add a further 18 between them, against zero social referral in every prior reading. None of it ties to a known campaign (`campaign_daily_14d` shows only 9 visitors against the dormant `2026-08-organic-launch` and `2026-08-beta-recruitment` tags), so the source is real but unattributed — see Watch.
Two COD issues complete within 2 minutes (06:35:02–06:36:49 UTC): COD-5634 (priority 0) fixes a mobile job call-out fee being multiplied by the legacy room-count scale when the pricing engine charges it once per visit — a real overcharge rather than a cosmetic bug. COD-5633 (priority 0) closes out COD-5620, which completed on 15 November but left the native reauth retry never firing on a Clerk network failure; today's fix is the corrected version of that same path. COD backlog ticks up slightly to 81 from 80 (+1.25%), with `high_priority_open` easing to 83 from 84 (−1.2%) and `completed_last_7d` at 134 against `created_last_7d` of 148.
`unresolved_count` drops to 6 from 8 (−25%) with no new Sentry issue entering the top list and crash-free sessions/users both still at 100% (144 sessions_7d). The long-running Adapty-fetch-timeout issue (109003255) stays quiet, `last_seen` still frozen at 14 November 20:14 UTC — its `events_7d` reading of 16 is the rolling 7-day window decaying rather than new activity, consistent with that quiet. What exactly cleared from the unresolved list isn't named in the snapshot, so this is reported as a count move rather than an attributed fix.
Five items today, unchanged in count from 17 November.
1. Decide whether to formally triage and close the backend "Not authenticated" cluster and UNREAD_COUNT_DENIED now that both have been quiet for roughly three days, rather than leaving 10 issues unresolved on the chance they resume
`unresolved_count` has held at 10 throughout, including today.
→ App Health · Sentry 128651664
2. Decide whether Adapty's rate limit is a permanent constraint, given today's partial 2-of-11 recovery mirrors 12 November's reading that reverted within a day
Revenue and subscriptions_active now read `ok: true`; the other nine fields, including MRR, still don't.
3. Confirm whether the monthly PropertyData bulk-refresh's 18.2% API-call failure rate (2,360 of 12,990 calls, 13 November) left any of the 1,776 processed postcode districts with stale data
Unconfirmed since first flagged on 14 November; the cron still reports `fullRunComplete: true` and outcome `success` despite the failure rate, and no run has superseded it.
→ CCI
4. Confirm status on COD-2792, now 172 days old and still the oldest open COD issue, alongside DEV-196 at 308 days, the oldest open DEV issue
5. Wire Meta/Google Ad Spend credentials, still unset
`META_ACCESS_TOKEN`/`META_AD_ACCOUNT_ID` and the Google Ads credential set remain unset — with today's unattributed Instagram/Facebook burst (Watch), knowing which channel to credit matters more than usual.
| Surface | Metric | Value | vs Prior (17 Nov) |
|---|---|---|---|
| Beta | Testers | 23 | 23 (flat) |
| Beta | Weekly active | 1 of 23 (4.3%) | 1 of 23 (4.3%, flat) |
| Mobile | Crash-free sessions (7d) | 100% | 100% |
| Mobile | Sessions (7d) | 144 | 151 (−4.6%) |
| Source | Status | Notes |
|---|---|---|
| Cockpit snapshot endpoint (`/api/snapshot/all`) | ✅ ok | Single request succeeded, 07:11 UTC, 39.6s |
| Sentry (mobile, web, backend) | ✅ ok | All three projects returned; mobile's unresolved count eases, see Win 3; web keeps drifting up, see Watch; backend holds flat as the quiet cluster continues to decay |
| Axiom | ✅ ok | Call volumes and error rate returned; no cron failures today; `tasks:listByUserPaginated` continues in a normal partial-day range |
| Mobile | Sentry events (7d) | 58 | 66 (−12.1%) |
| Mobile | Unresolved issues | 7 | 6 (+16.7%, see Watch) |
| Web | Sentry events (7d) | 41 | 37 (+10.8%, see Anomaly 1) |
| Web | Sessions (7d) | 507 | 503 (+0.8%) |
| Web | Unresolved issues | 4 | 4 (flat) |
| Backend | Sentry events (7d) | 72 | 88 (−18.2%, see Watch) |
| Backend | Unresolved issues | 10 | 10 (flat) |
| App Health | Axiom `tasks:listByUserPaginated` (19 Nov, partial) | 456 | 517 (18 Nov, partial) |
| Marketing | Web pageviews (7d) | 110 | 113 (−2.7%) |
| Marketing | Unique visitors (7d) | 48 | 50 (−4.0%) |
| Marketing | GSC clicks (7d) | 11 | 9 (+22.2%, see Win 2) |
| Marketing | GSC impressions (7d) | 2,706 | 2,139 (+26.5%, see Win 2) |
| Marketing | Avg position | 15.98 | 14.00 (+14.1% worse, see Win 2) |
| Marketing | Email subscribers | 10 | 10 (unchanged) |
| Engineering | COD backlog | 80 | 81 (−1.2%) |
| Engineering | Completed last 7d | 132 | 134 (−1.5%) |
| Engineering | Created last 7d | 152 | 148 (+2.7%) |
| Engineering | DEV in-flight | 1 | 0 (+1) |
| Engineering | BET open | 0 | 0 (flat) |
| Engineering | High priority open | 94 | 83 (+13.3%, see Anomaly 2) |
| Population | Total users (BigQuery mirror) | 27 | 27 (flat) |
| Population | Reported spend (app) | £18,369 | £18,369 (unchanged) |
| Monetisation | MRR | £0 | £0 (data: 1 of 11 fields ok, down from 2 of 11, see Watch) |
| CCI | Drift queue | 0 pending | 0 pending |
| Spend | Total monthly (projected) | £539.60 | £542.10 (−0.5%) |
| PostHog (Marketing, Beta) | ✅ ok | Beta holds flat; marketing traffic eases back slightly from 18 November's social-driven jump, within normal range |
| GSC | ✅ ok (lag) | The usual ~2-day ingestion lag; impressions and clicks both jump, position eases, see Win 2 |
| MailerLite | ✅ ok | 10 subscribers; automations still 0 activated |
| Ad Spend (Meta/Google) | ⚠️ stale/not configured | Credentials still unset, see Action 7 |
| Adapty | ⚠️ error (429), partial | 1 of 11 fields `ok: true` (revenue), down from yesterday's 2 of 11, see Watch and Action 4 |
| Linear (Engineering) | ✅ ok | Returned successfully; a seventh sibling-fix batch lands including a security-leak fix and an overcharge fix, see Win 1; high_priority_open jumps unusually, see Anomaly 2; COD-2792 reaches 173 days, DEV-196 reaches 309 days, see Action 6 |
| BigQuery (Population) | ✅ ok | Returned successfully; reads 27, tracking the live tester count with its usual small lag |
| Convex (CCI) | ✅ ok | Drift queue empty, submissions queue empty; no new cron runs since 15 November's clean ONS run; the monthly PropertyData bulk-refresh's failure rate remains unconfirmed, see Action 5 |
| Spend | ✅ ok | All rows returned; projected £539.60/month, down slightly on 18 November |
| Mobile | Sentry events (7d) | 66 | 67 (−1.5%) |
| Mobile | Unresolved issues | 6 | 8 (−25.0%, see Win 3) |
| Web | Sentry events (7d) | 37 | 33 (+12.1%, see Watch) |
| Web | Sessions (7d) | 503 | 502 (+0.2%) |
| Web | Unresolved issues | 4 | 4 (flat) |
| Backend | Sentry events (7d) | 88 | 89 (−1.1%) |
| Backend | Unresolved issues | 10 | 10 (flat) |
| App Health | Axiom `tasks:listByUserPaginated` (18 Nov, partial) | 517 | 743 (17 Nov, partial) |
| Marketing | Web pageviews (7d) | 113 | 70 (+61.4%, see Win 1) |
| Marketing | Unique visitors (7d) | 50 | 31 (+61.3%, see Win 1) |
| Marketing | GSC clicks (7d) | 9 | 10 (−10.0%) |
| Marketing | GSC impressions (7d) | 2,139 | 2,504 (−14.6%) |
| Marketing | Avg position | 14.00 | 13.76 (eases slightly) |
| Marketing | Email subscribers | 10 | 10 (unchanged) |
| Engineering | COD backlog | 81 | 80 (+1.25%) |
| Engineering | Completed last 7d | 134 | 140 |
| Engineering | Created last 7d | 148 | 148 (flat) |
| Engineering | DEV in-flight | 0 | 0 (flat) |
| Engineering | BET open | 0 | 0 (flat) |
| Engineering | High priority open | 83 | 84 (−1.2%) |
| Population | Total users (BigQuery mirror) | 27 | 27 (flat) |
| Population | Reported spend (app) | £18,369 | £18,369 (unchanged) |
| Monetisation | MRR | £0 | £0 (data: 2 of 11 fields recover, see Watch) |
| CCI | Drift queue | 0 pending | 0 pending |
| Spend | Total monthly (projected) | £542.10 | £545.55 (−0.6%) |
| PostHog (Marketing, Beta) | ✅ ok | Beta holds flat; marketing traffic jumps on a new social channel, see Win 1 |
| GSC | ✅ ok (lag) | The usual ~2-day ingestion lag; clicks and impressions ease slightly, position eases slightly |
| MailerLite | ✅ ok | 10 subscribers; automations still 0 activated |
| Ad Spend (Meta/Google) | ⚠️ stale/not configured | Credentials still unset, see Action 5 |
| Adapty | ⚠️ error (429), partial | 2 of 11 fields `ok: true` (revenue, subscriptions_active), echoing 12 November's reverted partial recovery, see Watch and Action 2 |
| Linear (Engineering) | ✅ ok | Returned successfully; a sixth sibling-fix batch lands including a priority-0 overcharge fix, see Win 2; COD-2792 reaches 172 days, DEV-196 reaches 308 days, see Action 4 |
| BigQuery (Population) | ✅ ok | Returned successfully; reads 27, tracking the live tester count with its usual small lag |
| Convex (CCI) | ✅ ok | Drift queue empty, submissions queue empty; no new cron runs since 15 November's clean ONS run; the monthly PropertyData bulk-refresh's failure rate remains unconfirmed, see Action 3 |
| Spend | ✅ ok | All rows returned; projected £542.10/month, down slightly on 17 November |
Crash-free 100.00% over 136 sessions · 6 open Sentry issues · backend errors 0.2%.
47 web visitors · 0.0% requested early access · top channel Direct