Weekly Digest 007
Built from ops/weekly/2026-10-05.json vs ops/weekly/2026-10-12.json (the two newest
dated snapshots — latest.json duplicates 10-12), GitHub issue history for
2026-10-05 → 2026-10-12, and decisions/ entries dated this week (one:
2026-10-07-gir-cloud-run-minors-video-implied-consent.md).
1. Drivers that moved
Diff of the drivers block, 2026-10-05.json → 2026-10-12.json:
| Driver (JSON key) | 10-05 | 10-12 | Change |
|---|---|---|---|
| reviews_per_active_parent_per_month | 1.333 (n=6, month 2026-08) | 1.333 (n=3, month 2026-10) | value coincidentally unchanged — n and month both moved, not like-for-like |
| parents_invited_per_coach_per_month | 0.2 (n=5, month 2026-09) | 0 (n=6, month 2026-10) | dropped to zero; n moved 5→6 |
| invite_to_first_purchase_conversion | 6 (n=1, month 2026-08) | 2 (n=1, month 2026-10) | dropped 6→2; n held at 1 — single-sample, see caution below |
| parent_monthly_retention | 0 (n=6, month null) | 0 (n=6, month null) | none |
| coach_monthly_churn | 0 (n=5, month 2026-10) | 0 (n=5, month 2026-10) | none |
| price_per_review.usd | 15.00 (n=8) | 13.64 (n=11) | dropped; n +3 |
| price_per_review.cad | null (n=0) | 8.00 (n=1) | first CAD-denominated review ever recorded |
A real week of volume, after several flat ones. volume.active_parents went 0 → 3,
volume.reviews_delivered_in_period 0 → 4, volume.gmv.usd $0 → $30, volume.gmv.cad
$0 → $8. All four deliveries trace to two roster parents: P-1840925a (3 reviews,
reviews_in_period: 3) and P-0aa9b668 (1 review, joined this period via
origin: platform_matched). That reconciles the price_per_review split above: 3 USD
reviews pulled the USD average from $15.00 to $13.64 (n 8→11), and the 1 CAD review is
the first entry ever in that leg (n 0→1).
volume.blended_take_rate.cad reads 43.87%, roughly double the 20% base_take_rate
decision (model/assumptions.yaml) — on a single CAD transaction (n=1) this is most
likely a small-n/fixed-fee artifact on an $8 ticket rather than a real rate change, but
it is new this week and worth the export owner's eye before a second CAD data point
confirms or dissolves it.
parent_monthly_retention holds at 0% — the same flat zero first measured last week
(cohorts[0], 2026-08 cohort, still m1_retained: 0). A new 2026-10 cohort
(cohorts[1], size 2 — P-1840925a and P-0aa9b668) has appeared but is too new for a
month-1 read (m1_retained: null). Nothing this week contradicts last week's first real
retention data point.
parents_invited_per_coach_per_month fell to zero — n moved 5→6 as the newest
coach (C-41eb4e06, joined 2026-10-07) rolled into the window, and all six active
coaches show invites_sent_in_period: 0. Read next to the volume jump above: this
week's growth came from an existing relationship repurchasing and a platform_matched
parent, not from a coach-driven invite — worth noting since driver:invites is the
official name for this lever and it hit its floor in the same week volume moved.
invite_to_first_purchase_conversion moved 6→2 on n=1 both weeks — same single-sample
caution digest 006 raised elsewhere: this is one invite's outcome changing, not a trend.
Coach-disappearance anomaly (digests 003–006) — still unresolved, 5th week flagged.
C-0f4b1805 remains absent from coaches/parents/roster in 2026-10-12.json. No new
contact_evidence or decisions/ entry explains it.
Guess-basis baseline — now 7 weeks unowned. Checked directly against
model/assumptions.yaml (grep -n "owner:" returns nothing in the file):
coach_monthly_churn, parents_invited_per_coach_per_month,
invite_to_first_purchase_conversion, parent_monthly_retention, and
reviews_per_active_parent_per_month are all still basis: guess with no owner anywhere
in the file. Digest 002 (2026-08-24) started this clock; it is now 7 calendar weeks with
no movement toward measured/actual, or an assigned owner, on any of the five.
A driver-taxonomy change is already live in the tracker, ahead of its own decision
record. This week's issues introduce two labels absent from CLAUDE.md's canonical
driver list (driver:retention, driver:invites, driver:frequency,
driver:conversion, driver:infra-cost, driver:coach-churn, driver:matched-premium,
driver:price): driver:g2r (8 issues this week) and driver:dc (9 issues this week) —
see §3. The decision record meant to formalize this (#292, "G2R + Delivering Coaches
north stars, pods, and the driver label set") is still open. Carried into §4.
2. Evidence logged
One new contact_evidence row since digest 006:
| Handle | Date | Channel | Driver tags | Direction | |---|---|---|---|---| | P-0aa9b668 | 2026-10-07 | email | (none) | unclear — no driver tag attached |
This is the parent's first-ever contact, logged the same day they joined and purchased
(roster[].joined: 2026-10-07, last_purchase: 2026-10-07). No driver_tags were
attached, so there is no basis to say which driver it pushed — stating that plainly
rather than guessing. No other contact activity this week: none of the six parents named
on digest 006's touch list received a logged touch (see §5).
3. Backlog shipped and started
Issue history for 2026-10-05 → 2026-10-12 (GitHub closed:/created: search against the
period boundary).
Shipped (closed this week): 26 issues — a sharp reversal from digest 006's zero. By
track: 18 track:driver (all carrying a driver:x tag — no hygiene gap), 7
track:table-stakes, 1 track:obligation.
| # | Title | Track / driver | |---|---|---| | 329 | iOS first-tap stall on silent audio primer | table-stakes, P1, surface:review | | 327 | Parent playback: one audio hiccup flips review to captions (sticky audioError) | table-stakes, P1, surface:review, feedback | | 318 | Family signup on /families ahead of #280 invites | driver:conversion, P1, surface:parent | | 311 | Coach can submit (and be transferred for) an unpaid review request | track:obligation, P0, surface:payments | | 297 | Integrity checks that fail the Friday report loudly | driver:dc, P0, surface:platform | | 296 | Canonical G2R + Delivering Coaches queries and the Friday report script | driver:g2r, P0, surface:platform | | 295 | Make connection origin truly immutable + append-only origin_corrections | driver:g2r, P0, surface:records | | 294 | Coach state machine with dated vetted_at, coach_type, capacity_per_week | driver:dc, P0, surface:coach | | 293 | Paid-submission + delivery timestamps on review_requests | driver:g2r, P0, surface:payments | | 291 | Viewer CI: intermittent vitest worker crash | table-stakes, P2, surface:platform | | 288 | GIR undiscoverable in /my-clips until a clip is sent | table-stakes, P1, surface:records, feedback | | 286 | GIR 'Quiet' pill on a recommended 'good' shift (classifier split-brain) | table-stakes, P1, surface:records, feedback | | 284 | GIR report crashes on a quiet shift (null clock) | table-stakes, P1, surface:records, feedback | | 277 | Parent review-playback fixes (note-resume freeze, play/pause lock, re-watch) | driver:retention, P1, surface:review/parent, feedback | | 276 | GIR: no parent entry point to START a Game In Review | driver:frequency, P2, surface:capture | | 263 | GIR: home-surface game grouping | table-stakes, P2, surface:platform | | 262 | GIR: coach side — grouped bench request | driver:coach-churn, P2, surface:coach | | 261 | GIR: confirm + send 1–N clips to a coach | driver:conversion, P2, surface:parent | | 260 | GIR: free report — ingest → parse → candidacy → view | driver:frequency, P2, surface:review | | 259 | GIR: verify coded-export shapes (Hudl/Instat variants) | driver:frequency, P2, surface:capture, spike | | 242 | Cap review clip length for goalie coaches | driver:coach-churn, P2, surface:capture, feedback | | 228 | Coach home V5.2: Link each kid to a team | driver:g2r, P0, surface:coach/records | | 227 | League schedules as games (G2R denominator) | driver:g2r, P0, surface:records | | 78 | Coach roster-need expression | driver:matched-premium, P3, surface:matching | | 74 | Coach-absorbed billing fallback | driver:price, P2, surface:payments | | 68 | Free-review allowance: first N per coach per month | driver:price, P1, surface:payments, proposed |
Note: #276's closure (2026-10-07) lines up with the same-day decision
decisions/2026-10-07-gir-cloud-run-minors-video-implied-consent.md, which unblocked the
prod enablement that #276 depended on.
Started (opened this week): 50 issues — the largest weekly intake in the digest's history, in four clusters:
- "Two north stars (G2R + Delivering Coaches) + pods" epic — 18 issues (#292–#309;
#309 is the rollout index, #292 its decision record). All P0 except #308 (P1). Five
opened and closed the same day (#293–#297); 13 remain open, including #292 and #309
themselves. Source of the new
driver:dc/driver:g2rlabels flagged in §1. - Payments/obligation hardening — 4 issues, all
track:obligation: #311 (closed same day), #312 (refunds/disputes outside the app don't block coach transfer), #313 (Stripe webhook advances a request without checking payment_status), #314 (coach can still fetch a signed playback URL for unpaid/cancelled requests) — #312–#314 open. - Live-beta playback/platform hardening — 15 issues: #291, #310, #315, #316, #319,
#320, #324, #325, #326, #327, #329, #331, #332, #333, #334. Mostly
track:table-stakes; five carryfeedback(#320, #326, #327, #331, #332), one is a plainbug(#319). #327 and #329 closed same day; the rest (13) are open, including #331 ("iPhone fullscreen review playback is broken — ... decide the playback architecture"), which reads as the open-ended parent of several of the others (#324, #325, #326, #332, #333, #334 all touch the same playback surface). - Game In Review polish + parent funnel — 13 issues: #271–#277, #280, #284, #286,
#288, #317, #318. Six closed same day (#276, #277, #284, #286, #288, #318); seven
remain open, including #317 ("Free first review signup bonus", explicitly
PARKED).
4. Decisions needed
One decisions/ entry dated this week —
decisions/2026-10-07-gir-cloud-run-minors-video-implied-consent.md: Matt enabled GIR
self-serve ingest on prod Cloud Run, keeping the cheap hygiene controls (no retention,
no PII in logs, least-privilege R2) and dropping the dedicated-GCP-project + counsel-review
gate, reasoning that the implied-consent source feeds (Hudl/InStat/LiveBarn) already cover
consent. This is a decision made, not one needed — noted here because it unblocked #276
(§3). Reversal condition: parents objecting to third-party processing of their child's
video.
Carrying forward digest 006's five open items, elapsed weeks updated:
- Price-per-review derivation unconfirmed (
decisions/2026-08-19-price-per-review.md) — carried, 7 weeks. Matt still needs to confirm or complete the $120/hr → $15 derivation. - Season-one plan adopted without Alex, Danny, or Derek (
decisions/2026-08-20-season-one-plan.md) — carried, 7 weeks. The free-allowance removal and $20 September default still need those three conversations. - Market analysis + seasonality layer (
decisions/2026-08-20-market-analysis-and-seasonality.md) — carried, 7 weeks. Still flagged for Alex's review at onboarding, still unreviewed. - Startup capital exclusion (
decisions/2026-08-20-startup-capital-excluded-from-operating-model.md) — carried, 7 weeks. Reversal condition is explicitly Alex's to invoke as CFO; still open. - Residual Tier-1 confirmation on incorporation (
decisions/2026-09-09-incorporation-in-place-no-coach-reonboard.md) — carried, 4 weeks. Still need Stripe/counsel confirmation that the sole-prop → incorporated swap doesn't trigger a platform re-KYC that pauses payouts, and still need to resolve the jurisdiction conflict the decision file itself flags.
New this week:
- G2R + Delivering Coaches north-star adoption — decision record still open while its
labels already ship.
driver:dc/driver:g2rare live on 23 issues opened or closed this week (§1, §3) — in the Friday report script, the coach state machine, and schema changes already merged — but the decision record meant to formalize the two north stars and the driver label set (#292) is stillopen, and neither label appears in CLAUDE.md's canonical driver list. Flagging now rather than waiting for it to compound the way the other five already have.
Six open decisions, five of them now 7 weeks unmoved.
5. Touch points this week
From the roster block (2026-10-12.json). Ordering by primary_flag (SILENT,
ONE_AND_DONE, NEW, NEVER_CONTACTED, CADENCE_DUE): 5 SILENT parents plus 1 NEW fill the
cap of six — 3 cut (P-4719cb7d, CADENCE_DUE; P-c4005031 and P-64d03653, both
NEVER_CONTACTED).
| Handle | Why (primary_flag) | Owner | Suggested opener |
|---|---|---|---|
| P-2952f6f8 | SILENT | Alex | Last purchase 8/3, one delivered review 8/4, last contacted 8/20 — now 69 days since purchase, no follow-up logged since. |
| P-c2dc0179 | SILENT | Alex | Most paid-up parent on record — 3 paid, 3 delivered, all by 8/7 — contacted 9/11, now 65 days since purchase, no follow-up since that call. |
| P-494a7944 | SILENT | Alex | 2 paid, 1 cancelled, last delivery 8/16, last touched 8/20 — now 58 days since purchase; the cancellation plus continued silence is still worth a direct check-in. |
| P-7d3431ec | SILENT | Alex | One delivered review 8/14, last touched 8/19 — now 60 days since purchase, no repeat purchase since. |
| P-f1c95a7d | SILENT | Alex | One delivered review 8/14, last touched 8/20 — now 59 days since purchase, no repeat purchase since. |
| P-1840925a | NEW | C-763bfd8b | Joined 10/1, 3 paid + 3 delivered by 10/5 — the most active parent in the beta by volume — and never contacted (last_contact_at: null). Worth a check-in while the experience is fresh, routed through their own coach. |
Follow-through on digest 006's touch list (6 handles): the one new contact_evidence
row this week (§2) was for P-0aa9b668, who was not on that list — so zero of digest
006's six were followed up on.
- P-2952f6f8 — no new contact since 8/20. Fifth week running.
- P-c2dc0179 — no new contact since 9/11. Fourth week running.
- P-494a7944 — no new contact since 8/20. Fifth week running.
- P-7d3431ec — no new contact since 8/19. Fifth week running.
- P-f1c95a7d — no new contact since 8/20. Fifth week running.
- P-4719cb7d — no new contact since 8/19, still zero — but this parent repurchased on
10/7 (
last_purchasemoved 8/18 → 10/7) entirely without any outreach from the team. The cadence gap closed itself through the parent's own action, not through follow-up; the contact miss continues even though the commercial outcome improved. Dropped off this week's capped list as a result (nowCADENCE_DUE, notSILENT— see cut list above).
Net: five of digest 006's six parents have now gone five weeks (four for P-c2dc0179)
with zero logged contact; the sixth repurchased on their own initiative with the outreach
gap still unaddressed. This is the fifth consecutive digest naming this same gap.
6. Mid-week pulse
No pulse entries this week.