Why the data comes from medicine
There is no reliable data on test-drive no-shows. The “40% of test drives are no-shows” figure repeated by blogs comes with no source. The only measurement with a described sample is from Foureyes, a US software vendor for dealers: about 700 dealers in the second half of 2023. 56% of customers from internet leads and 62% of those who phoned showed up for new-car appointments (Foureyes, February 2024). This is a vendor’s data, and visits are counted from what dealers entered into their CRM.
Medicine has studied the same problem for decades: a booking, a slot, a person who did not come. So what follows is its data. Carry it over to a showroom with care — more on that at the end.
What one reminder does
A 2013 Cochrane systematic review pooled 8 randomised trials with 6,615 participants (Cochrane, CD007458). Without reminders 67.8% attended, with an SMS reminder 78.6%, with a phone call 80.3%. SMS and calls worked almost the same, but in two trials an attendance cost 55% and 65% less by SMS. The authors rated the evidence as low to moderate quality and said plainly that it is not yet enough to inform policy.
The 2016 meta-analysis by Robotham and colleagues used more data: 21 studies, 8,345 patients with reminders and 7,731 without (BMJ Open). Attendance rose from 54% to 67%, and the no-show rate fell from 21% to 15% — roughly a quarter. Several reminders raised attendance more than one. Voice reminders appeared more effective than text.
Human or machine
Here the data disagree. In Parikh’s 2010 trial of 9,835 patients, reminders came three days ahead. Without a reminder 23.1% did not come, after an automated call 17.3%, after a call from clinic staff 13.6% (American Journal of Medicine). The live call was clearly stronger. In the Cochrane review, SMS and calls worked almost the same.
The practical takeaway: automation covers the bulk, and a person is needed where a visit is expensive or the risk of a no-show is high.
When and how often to remind
The largest trial on reminder timing ran at Kaiser Permanente Colorado: 54,066 patients, 25 clinics (Steiner and colleagues, 2018). One reminder three days ahead — 5.8% missed visits; one a day ahead — 5.3%; both — 4.4%. Two reminders worked better than one and did not lower patient satisfaction. The difference between “three days ahead” and “one day ahead” was small. A 2012 meta-analysis by Guy and colleagues also found no difference between reminders 24, 48 and 72 hours ahead (Health Services Research). The rule “remind exactly one day before” has no evidence behind it.
The wording matters too. In two UK trials with almost 20,000 patients, the reminder arrived five days before the appointment (Hallsworth and colleagues, 2015). When the message added what a missed appointment costs the health system, no-shows fell from 11.1% to 8.4%. The second trial replicated the effect — 8.2%. The same idea stated in general terms, without the figure, worked noticeably worse — 9.9%. Specifics work better than a polite formula.
A cancellation is not a failure
A person who cancels in advance is not a loss: the slot can go to someone else. In a 2001 trial, a call the day before cut no-shows from 26% to 19%, while cancellations rose from 9.9% to 17%. Other patients took the freed slots, and the extra revenue paid for the calls (Hashim and colleagues).
A recent example is Penn Medicine in 2026 (NEJM Catalyst). First they surveyed 186 patients who had missed a visit. The most common reason — 22% of answers — was “I did not know I had an appointment”. Then they ran a trial with 59,994 patients at high risk of not showing up. The usual SMS was supplemented with an automated call that let people confirm, cancel or reschedule. No-shows fell from 11.3% to 9.6%, and completed visits rose from 75.9% to 77.8%. After the trial the scheme was switched on for everyone.
What reminders do not fix
A reminder cures forgetfulness, not unwillingness. In the same Kaiser Permanente trial the authors write that for the highest-risk patients two reminders are not enough and further measures are needed. A customer who booked a test drive “just in case” and has already bought a car next door will not come after a reminder. At least they can cancel the booking — if it takes one tap.
Popular figures without a source
- “Reminders cut no-shows by 50–90%.” The pooled estimates are about a quarter (Robotham) and about a third (Cochrane).
- “40% of test drives are no-shows” and “AI reminders cut no-shows by 30%.” Both appear in a Convin blog post of February 2025, with no source for either.
- “SMS raises attendance by 50%.” This is how the odds ratio of 1.48 from the Guy et al. meta-analysis is sometimes retold. An odds ratio is not an increase in attendance. The relative increase in attendance in the reviews is 14–23%.
- “The best time is exactly 24 or 48 hours before.” In the Kaiser Permanente trial reminders one day and three days ahead gave similar results, and in Guy et al. timing made no difference.
What carries over to a showroom, a service shop or any booking
- Two reminders. For a booking tomorrow — the evening before and a couple of hours ahead. For a booking several days out — two or three days ahead and on the day.
- “Reschedule” and “Cancel” buttons in every reminder. A freed slot is offered straight away to the next person in line.
- Specifics in the text. “The car will be ready at 14:00, a manager will meet you at the entrance” beats “this is a reminder of your booking”.
- A live call for expensive visits. The system reminds everyone; a manager calls those who have not confirmed.
- Three numbers before and after: the show rate, the share of bookings cancelled in advance, and the share of freed slots filled again.
What we don’t know
Almost all the data comes from medicine: different customers, different reasons for not coming, a different price of a visit. The Hallsworth trials ran in a single location in England, and accurate phone numbers were available for only 20% of patients. We found no trials at car dealerships, service shops or beauty salons. The Foureyes data comes from a software vendor. None of these trials tested messenger bots with buttons; at Penn Medicine the confirm, cancel and reschedule options came through an automated call. There is no data for Uzbekistan. We have no numbers of our own yet.
Where reminders sit in the whole chain — from request to visit — is on Sales automation with AI.
What such reminders look like in a conversation is shown in the demo on our page for car dealerships and on the page about chatbots. To have bookings in your business broken down, use the questionnaire on our home page or message us on Telegram.
