The evidence behind what we claim
Reminder software is sold with big round numbers. This page holds ours, with the sample size and the date next to each one, and links to the research we lean on so you can read it yourself rather than take our word for it.
It includes a trial that found no effect, and a list of the claims we are not making.
What we measure ourselves
Taken from the production database on the date shown, and refreshed monthly. Real numbers, printed as they are.
- Sample
- 719 messages that had a receipt, out of 753 sent by 93 businesses
- Measured
- 3 October 2026
- How it is counted
- Every message we send carries a delivery receipt from the mobile network. This counts the receipts that came back marked delivered, against those marked failed. Messages still awaiting a receipt are excluded rather than counted as successes.
- What it does not show
- 30 days of data. We started recording receipts on 3 September 2026 and update this figure monthly, including when it moves against us. It has: the first reading, six days in, was 90.3% from 113 receipts. Of the 79 texts that failed, 32 went to a landline, which cannot receive a text: the number saved as the customer's mobile was a landline. 12 went to a mobile number the network did not recognise or could not reach. 29 failed on our side, to countries we are not yet set up to send to: this group has doubled since the 22 September reading (14), and it is ours to fix, not bad customer data. We have no cause for the other 6.
- Sample
- 1,340 appointments that received at least one reminder
- Measured
- 3 October 2026
- How it is counted
- Counted across every appointment that received at least one text. 1,036 got a single message, 268 got two, 36 got three or more.
- What it does not show
- This says how often a reminded customer hears from a business, not what share of all appointments get a reminder. It is here because the usual objection to reminders is that they annoy people, and just over one message is not a campaign of harassment.
- Sample
- 902 of 7,273 timed appointments held by 120 businesses, starting between June 2025 and December 2026, of which 2,454 are bookings still in the future
- Measured
- 22 September 2026
- How it is counted
- Every appointment synced from a business's own calendar, converted to that business's time zone rather than to ours. All-day entries are excluded, along with anything lasting 20 hours or more, because a calendar stores an all-day event at midnight and it would otherwise count as an early start.
- What it does not show
- This says when appointments are booked, not how often they are missed. We do not record no-shows, so it doesn't support the claim that an early or late slot is missed more often than a midday one. The trade mix is also uneven: it is dominated by whichever businesses sync a calendar.
- Sample
- 107 businesses that have sent at least one text message, holding 293 campaigns created by our own account setup
- Measured
- 3 October 2026
- How it is counted
- New accounts are seeded with three campaigns whose sending hour depends on the trade, 6pm for most and 9am or 10am for the rest. This counts how many of those campaigns still carry the seeded hour, among businesses that have gone on to send a real message to a real customer. Accounts that never sent anything are excluded, because a default kept by a dormant account says nothing.
- What it does not show
- A kept default is weak evidence and can mean the setting was never opened. Note also the direction of the exceptions: 12 of the 14 changes moved the send time earlier into the working day, from 6pm to somewhere between 8am and 3:35pm. The other two moved a 9am default later, to 10am and to 4:22pm.
- Sample
- 1,270 people, across 200+ businesses
- Measured
- 3 October 2026
- How it is counted
- Text messages accepted by the network and billed to a business, from the first one on 23 November 2025. Emails are counted separately and are not in this figure.
- What it does not show
- Starting sample taken from service business that registered to Remindlo, not from a random sample of the UK population. The figure is the total number of reminders sent, not the number of people who received one.
- Sample
- Every contact named this way across 39 businesses, from 11 September to 3 October 2026. The model turned down another 190 entries rather than guess at a name
- Measured
- 3 October 2026
- How it is counted
- When a calendar entry has no guest to take a name from, its title goes to an AI model, which reads out the customer's name or declines. A name is written to the contact only if every word of it already appears in the entry. We checked all 856 against the entry each came from, and none contains a word that was not there. 127 of them are a surname with no first name, which is the right answer for a business that greets customers by surname.
- What it does not show
- Nobody has graded these names, so this counts names read, not names that are right. The check proves nothing was invented, not that the right person was picked from an entry that mentions two. Businesses have edited 1 of the 856 since, which we do not read as accuracy: most never open a contact they did not type. 154 of these customers have been sent a reminder since, 201 texts in all.
What the research says
All of it is healthcare, because that is where reminders have been tested in randomised trials. A dental practice is a closer match to those trials than a tyre shop is, and we say so rather than quietly transplanting the numbers.
A text reminder before a booked appointment increases the share of people who turn up.
- 78.6% attended after a text reminder, against 67.8% with no reminder
- 8 randomised controlled trials, 6,615 participants
- A phone call from a person did about as well (80.3%), but the text messages cost 55% and 65% less per attendance in the two trials that measured it
- Who was studied
- Patients with an already booked healthcare appointment, across trials in several countries.
- What it does not show
- The review authors rate the evidence low to moderate quality. It measures attendance at a booked appointment, which is not the same as a past customer coming back.
Gurol-Urganci I, de Jongh T, Vodopivec-Jamsek V, Atun R, Car J. Mobile phone messaging reminders for attendance at healthcare appointments. Cochrane Database of Systematic Reviews 2013, Issue 12. Art. No.: CD007458. Read it
A text reminder sent at the wrong moment does nothing. Timing is the product, not the text message.
- 40.5% responded with a text reminder, against 39.9% without: odds ratio 1.03 (95% CI 0.94 to 1.12), p=0.56
- First-time invitees were the exception: 40.5% against 34.9%, odds ratio 1.29 (95% CI 1.04 to 1.58)
- 8,269 adults aged 60 to 74, across 141 general practices in England
- Who was studied
- People sent a bowel screening kit by post, reminded seven weeks into the screening round if they had not returned it.
- What it does not show
- This one found no overall effect and it is on the page for that reason. The reminder went to people who had already ignored a posted kit, seven weeks after the fact, with no appointment to attend.
Hirst Y, Skrobanski H, Kerrison RS, et al. Text-message Reminders in Colorectal Cancer Screening (TRICCS): a randomised controlled trial. British Journal of Cancer 2017;116(11):1408-1414. Read it
Reminding somebody that a service has come round again increases the number who take it up.
- Reminders raised the proportion immunised: risk ratio 1.28 (95% CI 1.23 to 1.35), a risk difference of 8 percentage points, from 55 studies and 138,625 participants
- Text messages on their own: risk ratio 1.29 (95% CI 1.15 to 1.44), six studies, 7,772 participants, rated high certainty evidence
- Postcards 1.18, autodialled calls 1.17, letters 1.29, a person telephoning 1.75
- 75 studies across 10 countries
- Who was studied
- Children, adolescents and adults due for a vaccination, mostly in primary care.
- What it does not show
- The closest published work to a recall campaign, because the recipient has nothing booked. It is still vaccination rather than a service business, and a vaccination is free at the point of use where a service is not.
Jacobson Vann JC, Jacobson RM, Coyne-Beasley T, Asafu-Adjei JK, Szilagyi PG. Patient reminder and recall interventions to improve immunization rates. Cochrane Database of Systematic Reviews 2018, Issue 1. Art. No.: CD003941. Read it
A more recent synthesis finds the same size of effect for text reminders, but cannot rule out no effect at all.
- Reminders against none: risk ratio 1.11 (95% CI 1.05 to 1.19), 10 studies, 8,236 participants
- Text messages on their own: risk ratio 1.14 (95% CI 0.99 to 1.31), which does not reach statistical significance
- Telephone reminders: risk ratio 1.11 (95% CI 1.04 to 1.19)
- Who was studied
- Outpatients at hospital appointments.
- What it does not show
- Heterogeneity is high (I² 83%). The point estimate for text messages matches the 2013 Cochrane review, but on this data the honest reading is 'probably helps, not proven'.
Al-Turbag M, Mooney M, Corry M. A systematic review and meta-analysis of appointment reminders for enhancing hospital attendance. Journal of Hospital Management and Health Policy 2026;10:2. Read it
The reminder can only work on the people you hold a mobile number for, and that is usually the binding constraint.
- 59.1% attended after the standard invitation, against 64.4% when a text was sent 48 hours before: odds ratio 1.26 (95% CI 1.05 to 1.48), p=0.01
- Only 456 of the 1,122 women in the reminder arm (41%) had a mobile number on their GP record, so the rest were never sent one
- Among the women who were actually texted, uptake was 71.7% against 59.8%: odds ratio 1.71 (95% CI 1.29 to 2.26), p<0.01
- 2,240 women, single-blind randomised trial, London Borough of Hillingdon, November 2012 to October 2013
- Who was studied
- Women invited for their first routine breast screen in an ethnically diverse London borough where uptake sat below the national target.
- What it does not show
- The larger figure is a per-protocol comparison and is not like for like: women whose surgery holds a mobile number differ from those whose surgery does not. The trial's own headline result is the smaller one, 59.1 against 64.4.
Kerrison RS, Shukla H, Cunningham D, Oyebode O, Friedman E. Text-message reminders increase uptake of routine breast screening appointments: a randomised controlled trial in a hard-to-reach population. British Journal of Cancer 2015;112:1005-1010. Read it
Missed appointments are large enough to be counted nationally, every month, in a system that already sends reminders.
- 89.5% of appointments in general practice in England were attended in July 2026
- 34.9 million appointments were recorded across general practice that month
- Published 27 August 2026 by NHS England as official statistics in development
- Who was studied
- General practice in England.
- What it does not show
- The remaining 10.5% is not all no-shows. The status field holds Attended, Did Not Attend or Unknown, and practices with poor data quality are suppressed entirely, so treat it as a ceiling rather than a no-show rate.
NHS England. Appointments in General Practice, July 2026. Published 27 August 2026. Read it
A returning customer is worth more than the visit in front of you, because the same customer becomes more profitable each year they stay.
- Retaining 5% more customers raised profits by almost 100% across the businesses studied
- In auto servicing, a fourth-year customer produced more than three times the profit of a first-year customer
- Drawn from more than 100 companies across two dozen industries
- Who was studied
- Service businesses in the late 1980s: credit cards, industrial laundry, industrial distribution and auto servicing.
- What it does not show
- It is 1990, the figures are illustrative rather than a controlled experiment, and note what the article does not contain: the endlessly repeated line that keeping a customer is five times cheaper than winning one appears nowhere in it.
Reichheld FF, Sasser WE. Zero Defections: Quality Comes to Services. Harvard Business Review, September-October 1990. Read it
What we do not want to claim
We avoid numbers that are not measured, and we withdraw claims that are not supported by evidence. Because of that you won't find made-up claims on this site.
Recover up to 30% of lost customers
There's a decent chance that the 30% is something you could actually achieve. We'll do our own measurement and publish it, but until then we cannot say it.
A specific percentage fall in no-shows for your business
The published trials are healthcare, and the effect varies enough between them that a single number would be a guess dressed as a measurement.
Keeping a customer is five times cheaper than winning a new one
Traced to Reichheld and Sasser's 1990 Harvard Business Review article, which does not say it. What that article does show is that retaining 5% more customers lifted profits by almost 100%, and that in auto servicing a fourth-year customer was worth more than three first-year ones. Those are the figures we use.
The rule we hold ourselves to
A number goes on this site only if it is one of three things:
- 01something we measured, printed with its sample size and the date it was taken
- 02something somebody else measured, with a citation and a link
- 03an assumption you can change yourself, which is what the calculator is for
The honest version of the pitch
Reminders help people turn up, the effect is real but modest, and timing is most of it. The free plan sends 10 messages a month with no card, which is enough to find out what happens with your own customers.
Start for free