Automated reminders cut missed appointments by 29-39% — then they plateau, and here's why
A meta-analysis of 29 studies (cited June 4, 2026) quantifies the effect of automated reminders: -28.9% no-show rate on average for an automated SMS, -39.1% for a human phone call. The reference study, Koshy, Car & Majeed (2008), shows a drop from 18.1% to 11.2% no-shows with a simple SMS reminder. But no study arm ever gets close to zero: reminders solve forgetting, not the friction of rescheduling.
The most-cited figure in this body of research (Koshy, Car & Majeed, 2008) still holds up nearly 20 years later: an SMS reminder brings the no-show rate from 18.1% down to 11.2%, a 38% relative reduction. The broader meta-analysis (29 studies) confirms the order of magnitude with a weighted mean of -28.9% for an automated message and -39.1% for a human call — live back-and-forth outperforms a one-way send.
What these numbers also reveal is the ceiling: in the best-performing arm of a three-group study, 13.6% of appointments were still missed after the reminder. A reminder that only offers "reply C to confirm" does nothing for the customer who genuinely can't make it anymore — they go silent, and the slot dies with them. The real friction isn't forgetting, it's the lack of a simple path to cancel or reschedule inside the same exchange. That's exactly the terrain for an AI agent at a branch desk: mail pickup, onboarding appointments, an agrément renewal deadline — every reminder needs to allow rescheduling on the spot, not just acknowledgment.
For AppH
- AppH's agent treats every reminder as a two-way exchange: confirm, cancel, or reschedule inside the same thread, never a one-way message — the factor the literature flags as most decisive, more so than reminder frequency.
- When a customer doesn't respond to the automated message, the agent escalates to a human at the branch instead of stacking silent follow-ups — consistent with the finding that a human call outperforms SMS alone.
Against / the honest limit
- These figures come from clinical studies (medical appointments), not French business domiciliation; the actual rate of slots "saved" by an AppH agent still needs to be measured branch by branch, and hasn't been published yet.
- Even with a well-designed two-way reminder, no study cited here gets close to a 0% no-show rate — AppH doesn't promise to eliminate missed appointments, only to cut the share caused by forgetting and by the lack of a rescheduling path.
What strikes us about this documented ceiling (never below roughly 11-14% no-shows, whatever the method) is that it confirms what we keep telling branch owners: a reminder only helps if it opens a simple way out. An agent that sends a text and stops there solves half the problem; one that closes the loop — reschedule, cancel, escalate — solves the other half.
Reviewed by a human at AppH