Mindbloom Dossier · 06 · The quarter, sequenced

The first 90 days.

Everything else in this dossier is analysis or prototype. This is the order I would actually do it in, with what each phase proves before the next one spends anything.

Days 1 to 15

Instrument and listen. Ship nothing.

  • Pull the real baselines this dossier had to assume: stall rate after the eligibility check, consult show rate, opt-out rate, alumni re-engagement rate. Replace the illustrative defaults in the Ledger with truth.
  • Stand up the weekly funnel-voice memo by hand first: read every reply and review myself, publish the memo internally each Monday, per piece 01.
  • File the site-versus-checkout price mismatch from the corpus as a bug with engineering, and confirm where it stands.
  • Meet the guide and clinical-comms teams. Every cadence in this plan gets their sign-off before a single send.
Days 16 to 45

The two cheapest tests on the property.

  • Run the payment-plan experiment exactly as written: one sentence, pre-registered kill criteria, clinical comms co-owning the copy.
  • In parallel, test the quiz-door reassurance line from the teardown: time estimate, no-commitment language, measured on start rate.
  • Pilot the session-day logistics cadence with one guide pod and a small cohort: sitter confirmation, checklist nudge, nerves flagged to the guide, nothing clinical from the agent, ever.
Days 46 to 75

Earn the right to automate.

  • Run the crisis-classifier bake-off against the red-team suite: whatever detector the team chooses must beat both baselines on recall without losing the precision, and ambiguous still routes to humans.
  • Only then: pilot the guarded re-engagement cadence from Tend on stalled leads, with the human review queue live and every send logged with its reasoning.
  • Automate the Monday funnel-voice memo I have been writing by hand, now that a month of hand-written editions defines what good looks like.
Days 76 to 90

Point the machine at the relationship.

  • Pilot the alumni win-back thread with guide review on every reply, and the referral-timing test behind genuinely great outcomes only.
  • Write the quarter memo: every experiment's numbers, every kill hit or avoided, and the next quarter's ranked list, in the async written form the company already runs on.

What day 90 has to show

Real baselines where assumptions were. At least one shipped copy change with a measured lift and flat guardrails. A crisis-detection bar that is written down and passed, not vibes. A weekly written memo the team actually reads. And zero incidents of automation crossing a line a human should have held. If the quarter produces those five things, the second quarter funds itself with the Ledger's arithmetic; if it does not, the memos will say exactly why, which is worth almost as much.