All 75 campaigns audited. 12 safe fixes are already live. Everything left that needs a human is below, sorted by how you action it: clicks in Customer.io, 2 messages to send, one-word approvals for Claude, and 4 bigger calls. Ticks save in this browser.
Campaign 50, reverse trial. The day-14 A/B still splits 50/50 but the variant (action 529, "tomorrow frame") is a draft, so half that arm gets nothing. Enable 529, or set the split to 100/0.
Campaign 73's E2 pair (engaged and disengaged, actions 479 and 480) and campaign 1's action 410 were built, QA'd and never enabled. Everyone passes through unsent. Enable them or delete them; either beats the current silent skip.
Customer.io blocks agent edits to Bee-editor bodies. Open each template's text block and swap the string. Copy buttons give you the exact before and after.
ready to go — SOAP notes → ready to go: SOAP notesyour practice — all in one place → your practice, all in one placeautomatically — so you stop chasing no-shows → automatically, so you stop chasing no-showsclinical templates — SOAP notes → clinical templates: SOAP notesHi {{customer.first_name}}, → Hi {{customer.first_name | default: "there"}},Hey {{customer.first_name}}, → Hey {{customer.first_name | default: "there"}},Campaign 85's first cohort: 19 sent, 9 bounced. Either flip Email 1 (action 680) to draft until addresses pass a MillionVerifier check, or accept the burn while volume is small (115 matched).
Campaign 7: one send ever, no conversion goal, 49 people matching but nobody entering. Stop it.
This is the live revenue leak. Evidence.
Both waits are named "Wait 7 days" but set to 1 day and 2 days, so the "day 7" SMS fires around day 3. I set both to 604,800 seconds. People parked mid-wait keep the new timer.
Two n8n failures: campaign 80's webhook is 0 for 425 with 2,199 failed attempts, and campaign 3 never receives voicedrop_delivered (its voicemail follow-up SMS has zero sends ever). It is your n8n rig, so this is a joint session: we trace both flows, then fix. While we are in there we confirm what the flow keys on, and I repair the invalid holdout (both arms currently send identical payloads saying action_id 481; template 393 should say 482).
15 untitled empty shells (oldest untouched since November 2025) plus campaign 34, Essentials onboarding, for a plan that no longer exists. None has ever sent anything. I delete them and list what went.
The drip sells $1 a month for 3 months while the July broadcasts sell 50% off 12 months, to overlapping audiences in the same week. Pick one as July's offer; I re-time or pause the other on your word.
Campaigns 1 and 3 have identical triggers, so every new admin gets the onboarding series plus a voicemail, SMS and sales emails in week one. Proposal: campaign 3 only for high-intent segments (team size, insurance interest, tier-1 phone), starting day 2. Say yes and I draft the exact trigger change for review.
Ready to go: Plus onboarding (35), Advanced onboarding (36), monthly to annual (37, 7,798 people match), insurance upsell (39), and the accepting_insurance backfill trio (75 to 77, start then stop when drained). Rank them and I prepare each for your enable click.
Nothing emails people whose payment fails (billing_status Unpaid). A recovery sequence protects the net-subscriber number the June plan cares about. Say go and I build it as drafts on the standing rule.
Campaign 73 computes the current month's code in Liquid (JAN1 to DEC1). All 5 live emails and the SMS now emit JUL1. PostHog shows 13 people arriving at checkout with promoCode=JUL1 since 3 July and zero JUL1 redemptions, while JULY4 redeems normally through the same checkout. JUNE1 also never got its expiry: redemptions kept landing until 4 July. The rotation has failed twice in 3 months (MAY1 in the Outlook links in June, JUL1 now).
The same expired-trial population can receive "$1 a month for 3 months" from campaign 73 and "50% off 12 months" from the July broadcasts in the same week. Nothing suppresses one when the other fires.
Both wait steps are named "Wait 7 days". The first is set to 86,400 seconds (one day), the second to 172,800 (2 days). The SMS named "day 7, non-bookers" fires around day 3, before most people have had a chance to book. The stopped twin (campaign 84) has the same values, so the bug came from the original build.
Campaign 80's live voice-drop webhook has 425 sends, 0 delivered and 2,199 failed attempts with no HTTP status recorded, which reads as connection failures to the n8n endpoint. Separately, campaign 3 never hears back: the SMS that says "following the voicemail I just left" has zero sends ever, because voicedrop_delivered never arrives inside its 10-minute wait. The receiver (campaign 72) shows 5,822 of 7,133 journeys never activating.
The live arm and the 20% holdout arm send byte-identical webhook payloads. Both declare action_id 481. I verified the 2 templates directly. Unless n8n suppresses on some other signal, the holdout receives real voicemails and the test measures nothing.
The day-14 test in the reverse-trial journey still splits 50/50, but the variant action (529) is in draft. Half the people randomised into that arm receive nothing, and have since the test went live. The same built-but-never-enabled pattern sits in campaign 1 (action 410) and campaign 73 (both E2 variants).
First cohort: 19 sent, 9 bounced. Abandoned signups are typo-prone by nature, but at this rate the campaign damages sender reputation for a tiny audience (115 matched so far).
Grammar on both trial-end SMS templates. Preheaders on the 10 emails that had none or had a wrong one (272 to 277, 396, 397, 109, and campaign 70's cancellation email, which promised a reschedule link the body does not have). Capterra review link verified working; the June 404 concern is closed. No sending states, triggers, delays, splits or goals were changed.
Human opens and bounce as a share of delivered emails. Hover for each week's detail. The final week is partial.
| Week starting | Delivered | Human opens | Bounce |
|---|
Opens and clicks as a share of delivered, across each campaign's email sends. Campaigns under a 20% open rate need work.
| Campaign | Human open | Human click |
|---|
Campaigns 1 and 3 share identical entry triggers, and campaign 50 starts the same day. Up to 15 touches land in the first 8 days, with no frequency cap anywhere in the workspace.
This week's work is the queue above. These phases follow.