Your automations were perfect the day you built them. That was four months ago. Since then: you changed your phone system (the missed-call trigger broke silently), your field-service software updated (the reminder template reverted), a sequence has been texting customers about your "spring special" since July, and your review requests stopped firing when someone toggled a setting during training. Nothing alerted you. Nothing ever alerts you. That's what the weekly health check is for.
This is a Hub post in the 8-system framework — the maintenance rhythm for Pillar A: Capture Every Lead, Pillar B: Turn Estimates Into Signed Jobs, Pillar C: 5-Star Review Engine, and Pillar D: Full & Calm Calendar. Thirty minutes, same time every week, forever.
The setup: when and who
- Same 30 minutes every week. Friday afternoon or Monday morning — pick the slot with the fewest interruptions and protect it. "Whenever I get to it" means never.
- One owner. The office manager, the owner, whoever owns the systems. Not "whoever's free" — accountability needs a name.
- The checklist lives in one place. A printed sheet, a recurring task in your project tool, a note in the setup guide binder. The checklist is the system; the person just runs it.
- Log the results. Date, what you checked, what you found, what you fixed. Three lines. The log is how you spot patterns ("the Facebook integration breaks every other month") and how the backup person runs the check when the owner is out.
Minutes 0–8: delivery and volume (the vital signs)
Open each sending tool's dashboard and check the week's numbers:
- Text delivery rate. What % of texts sent actually delivered? Healthy is high-90s. A sudden dip means a carrier issue, a registration problem, or a content filter — investigate before it becomes a week of silent failures.
- Email delivery and bounce rate. Same check for email sends. A bounce spike means list rot or an authentication problem (SPF/DKIM).
- Volume sanity. Roughly the expected number of texts/emails went out? Zero sends from a system that should send daily = a broken trigger. 10x normal volume = a loop or a misconfigured blast. Both are emergencies.
- Opt-out count. A normal trickle is fine; a spike after a specific send tells you which message annoyed people. Note it, adjust the message.
You're not doing deep analysis — you're glancing at the gauges. Anything red gets a note for the fix-it block at the end.
Minutes 8–16: trigger and sequence spot-checks
Pick two automations per week on rotation (so every system gets checked monthly). For each:
- Is the trigger still live? Open the automation and confirm it's enabled and the trigger condition is intact. Triggers die silently when connected tools update, when someone "just toggles this to test something," or when API connections expire.
- Did it fire this week? Check the run/send history. An enabled automation with zero runs is either correctly idle (no triggering events) or broken — verify which by checking whether triggering events actually occurred.
- Is the content still current? Read the actual messages. Seasonal references ("spring special" in October), old pricing, the previous technician's name, a promotion that ended — stale content in live automations is embarrassingly common.
- Are the links working? Click every link in the sequence. Booking links, review links, payment links — they break when pages move, and a broken review link means weeks of lost reviews nobody noticed.
Rotation schedule: Week 1 — Systems 1 & 2 (missed-call, lead response). Week 2 — Systems 3 & 4 (estimate follow-up, reviews). Week 3 — Systems 5 & 8 (confirmations, dispatch texts). Week 4 — Systems 6 & 7 (after-hours, reactivation). Monthly, every system gets eyes on it.
Minutes 16–22: the customer-facing surfaces
- Google Business Profile: any new reviews needing responses? (The 24-hour response rule.) Any suggested edits or questions pending?
- Website form: submit a test lead monthly (not weekly — but check the last submission date weekly; no submissions in 2 weeks on a normally-busy form = something broke).
- Phone system: is the after-hours greeting current? Did the holiday schedule get left on after the holiday? (The classic: "we're closed for Thanksgiving" playing in mid-December.)
- Missed-call log: any missed calls that didn't get the text-back? Spot-check two — the automation's coverage should be near-total.
Minutes 22–30: fix-it block and the log
The last 8 minutes are for fixing what you found — or scheduling the fix:
- Fix the 5-minute items now: re-enable the toggled automation, update the stale price, fix the broken link. Don't defer what takes minutes.
- Schedule the bigger items: "rebuild the Facebook integration" goes on the calendar with a date, not on a mental to-do list.
- Write the log: date, systems checked, issues found, actions taken/scheduled. Three to five lines.
- Flag patterns: same issue two weeks running = systemic, not incidental. The integration that breaks monthly needs replacing, not re-fixing.
The monthly deeper checks (add to week 4)
Once a month, extend the check by 15 minutes:
- Run a test lead end-to-end. Submit a website form, verify the notification, the CRM entry, and the automated response — the full chain, timed.
- Review the suppression lists. Opt-outs honored? Any manual re-adds that shouldn't be there?
- Check tool bills. Any subscription price changes, unused seats, or tools you forgot you're paying for? (The "check what you own" audit, lite version.)
- Review the month's metrics trend. Delivery rates, response times, review velocity — direction matters more than absolute numbers. Declining trends get investigated before they become problems.
What the health check catches (the greatest hits)
- The expired integration token. Facebook leads stopped flowing 3 weeks ago; nobody noticed because the "connected" indicator was green. Caught by the volume sanity check.
- The seasonal message in the wrong season. "Beat the summer heat!" texting in November. Caught by the content read-through.
- The broken review link. Google changed the review URL format; the text still sends the old one to a dead page. Caught by the link check — three weeks of lost reviews otherwise.
- The duplicate sender. Someone re-enabled the old tool's reminders during a settings dive. Caught by the volume check (2x normal reminder volume).
- The quiet-hours violation. A new sequence scheduled without time restrictions, texting at 6am. Caught by the send-time review — before a customer complaint catches it.
Every one of these is a real pattern, and every one is caught by a checklist that takes 30 minutes. The alternative is discovering them via customer complaints — the most expensive monitoring system ever designed.
Compliance note
The health check is a compliance tool as much as a performance one: it verifies opt-out handling still works, quiet hours are still configured, and 10DLC business-texting registration (carrier registration for texts sent by software; 1–7 day approval, ~$15–$20 in fees) is still valid after tool changes. Add a quarterly line item: confirm consent records are intact and suppression lists migrated correctly after any tool change. General information, not legal advice.
Related guides in this series
The maintenance rhythm for every pillar — Pillar A: Capture Every Lead, Pillar B: Turn Estimates Into Signed Jobs, Pillar C: 5-Star Review Engine, Pillar D: Full & Calm Calendar. Pair with the overlap audit (the deeper quarterly version) and the annual automation checkup.