Missed-call recovery playbook. First 30 seconds, 3-touch cadence, booking handoff.

The point of the first 30 seconds is the same point you'd want any dispatcher to clear if a call rolled over from voice mail into a live voice — acknowledge by name, confirm the issue, lock the address, and set the booking expectation. The voice agent is front-desk-of-record for the missed call; it doesn't deflect, doesn't 'a tech will call you back', and doesn't open with a marketing line.

Below the surface the voice agent is also qualifying the call — ticket type (water heater leak / AC out / panel tripped), severity (active leak vs. routine install), timing (same-day vs. tomorrow vs. flexible), and tenant vs. owner. The same form the audit-request scheduler drop-down walks; every recovered call lands with the same fields the dispatcher needs to put a tech on it.

• Confirm the address from your CRMThe voice agent reads the address tied to the caller's number and confirms it on the call — 'showing 412 Maple — is that the house?' — not 'what's the address'. Same shape as the audit-request drop-down is checking: the address isn't re-typed, it's verified.
• Set the booking expectation, not a promiseLock the window with a question the dispatcher can answer: "any tech free before 5 pm today?" or "we have 9 — 11 am tomorrow, would that work?" — never "we will be there in 30" if a tech isn't on the board. Honesty on the booking window is the same calibration the missed-call-recovery playbook keys off.
• Don't oversell. Don't deflect.No upsell pitch on the first 30 seconds. No marketing-side route to voicemail. The first 30 seconds is the call solved, the address locked, and a window the tech can meet — the rest comes after the slot is on the dispatch board.

A missed call that goes unreturned is the most expensive line item on the weekly ROI report. The 3-touch cadence is the format that closes it: a first SMS within 60 seconds of the missed call, a follow-up at +2 hours if no reply, and a final same-day touch at +6 hours. Each message is a single SMS — opt-out line in every one.

The cadence is tuned to the recipient's timezone and quiet-hours windows (same rules as the TCPA playbook). The first touch is the lead-generator — 'we just tried you, here for the appointment' — and the second and third are each a separate push, not a duplicate of the first. If the third touch doesn't get a reply, the lead goes into a longer-window follow-up, not a fourth intra-day SMS.

• No upsell. No marketing carousel. One ask per touch.Each of the 3 messages is one SMS that asks for the booking — 'we can have a tech out tomorrow morning — does 9 — 11 work?' — not a sequence of marketing angles. Anything longer than a sentence is harder to write back to and harder to rebook from.
• Opt-out and quiet-hours rules apply — even on +6hEvery SMS in the 3-touch cadence carries the same 'Reply STOP to opt out · Reply HELP for help' line, and the +6h push lands during the recipient's waking hours. The cadence respects the rules on /resources/tcpa-sms-compliance even when the call started off-hours.

Once the caller replies with a yes (any 'yes', '?', or specific time), the voice agent lands the slot directly on the dispatch board the shop runs — ServiceTitan, Housecall Pro, or Jobber. The tech sees one job, alongside everything else on the schedule. Same shape as every other job the shop books.

The handoff closes with an SMS confirmation (a separate 'see you tomorrow at 9' push), the per-booked-job bonus gets tagged to the lead, and the weekly ROI report reads the booking as one recovered-call conversion. The dispatch board is the source of truth for 'did the booking land', and the per-booked-job bonus bills against that — same calc as the per-shop worked example.

• Native writeback to whichever scheduler the shop runsServiceTitan, Housecall Pro, OR Jobber — same three the audit-request scheduler dropdown confirms a shop runs. The voice agent doesn't open a third-platform sideboard for the booking.
• Tag the lead for per-booked-job attributionThe lead is tagged on the first reply as 'recovered — missed-call', and the tag flows through the booking so the weekly ROI report attributes the recovered-call revenue to the same booking the tech sees on the dispatch board. The tagging is the audit trail for the bonus, not a separate analytics layer.
• A booking that doesn't land is a call Holdfast doesn't get paid onSame rule as the per-shop worked example: the bonus bills on the booking, not on the recovery. A call that doesn't book doesn't extend the cadence into a 4th touch — the per-booked-job risk sits with Holdfast, not with the shop.