
Notification engines: training reminders that cut no-shows in ILT
Training reminders work best when a notification engine generates them from session data instead of a person working through a spreadsheet. Give each reminder a rule relative to the session start, compute its send time in the right time zone, allow one message per registration, rule and channel, and check consent before every send.
Reminders very likely raise attendance, but the studies we found come from clinics and adult education, not instructor-led training. This guide covers the engine design, two tested code snippets, a consent table checked against primary sources, and a way to measure your own results.
Key takeaways
- Evidence supports reminders in adjacent settings, and we found no ILT study, so measure your own results.
- Plan reminder jobs from events and rules, and replan whenever the session changes.
- Store session start as local time plus an IANA zone, and choose per rule between a fixed offset and a local-clock time.
- Put a unique key on every planned send so duplicate events and retries cannot produce duplicate messages.
- Consent rules differ by country and are changing in the US as of November 2026, so have counsel review.
Do training reminders actually reduce no-shows?
Training reminders probably do, but the evidence is indirect. The best studies measure clinic attendance and adult-education courses, not instructor-led training. Read the figures below as a direction and a design hint, not as a forecast for your classes.
A 2016 systematic review in BMJ Open pooled randomized studies of text reminders for clinic appointments. It included 26 articles and used 21 in the main analysis. Attendance was 67% with reminders and 54% without (risk ratio 1.23, 95% CI 1.10 to 1.38), and no-shows were 15% versus 21%. Results varied widely between studies (I² of 83% for attendance). The review found multiple notifications more effective than a single one and did not compare message content.
Closer to training, a University of Bristol working paper followed 1,179 learners in UK government-subsidized adult numeracy and literacy courses. Encouraging text messages cut the share of learners who stopped attending by 36% and raised average attendance by 7% against a control group. The authors call these initial results, and the abstract does not give the message wording.
| Source | Setting | Reported result | Limit for ILT |
|---|---|---|---|
| BMJ Open review (2016) | Clinic appointments, randomized studies | Attendance 67% vs 54%; no-shows 15% vs 21% | Patients, not trainees; high variation between studies; content not tested |
| Bristol working paper | UK subsidized adult basic-skills courses | 36% fewer learners stopped attending; 7% higher average attendance | Ongoing weekly courses, not one-off paid sessions; preliminary results |
| Vendor articles | LMS and training software blogs | Completion or no-show gains quoted without a named study | Not usable as evidence; we leave the figures out |
Which events should trigger training reminders?
Generate training reminders from domain events, not from a nightly job that scans a table. A registration event creates a plan of reminder jobs from rules, and a session-change event supersedes that plan. Every message then traces to one registration, one rule and one version of the session.
The engine produces two kinds of message. Reminders are time-based, from a rule such as “48 hours before start”. Notices are event-based and go out immediately, such as a room change. Keep separate templates and rules for each. We suggest notices skip quiet hours when the session starts within a day.
| Event | Engine action |
|---|---|
| RegistrationConfirmed | Send confirmation; plan all reminder jobs for the session |
| SessionRescheduled | Increment session version; supersede planned jobs; send change notice; replan |
| SessionCancelled | Cancel planned jobs; send cancellation notice |
| RegistrationCancelled | Cancel planned jobs for that registration only |
| SeatOffered | Send the offer with its expiry; plan reminders only after acceptance |
| JoinLinkOrRoomChanged | Send a notice now; the next reminder reads the new value |
The event list assumes your data model already separates sessions from registrations, as covered in our guide to a TMS data model for sessions and enrolments. Seat offers follow your waitlist and cancellation rules.
How should training reminder schedules handle time zones and DST?
Store each session start as local wall time plus an IANA zone name, then compute send times from that whenever you plan or replan. A fixed offset such as 48 hours before and a local-clock rule such as two days before at 09:00 give different answers when daylight saving time changes in between.
The function below supports both rule types. We ran it on Python 3.13 with the standard zoneinfo module.
from datetime import datetime, time, timedelta
from zoneinfo import ZoneInfo
UTC = ZoneInfo("UTC")
def session_start(local_iso, tz):
return datetime.fromisoformat(local_iso).replace(tzinfo=ZoneInfo(tz))
def send_at(start, rule):
if rule[0] == "offset": # ("offset", hours_before)
return start.astimezone(UTC) - timedelta(hours=rule[1])
_, days, hh, mm = rule # ("local", days_before, hour, minute)
day = (start - timedelta(days=days)).date()
return datetime.combine(day, time(hh, mm), tzinfo=start.tzinfo).astimezone(UTC)
For a session starting Monday 2 November 2026 at 09:00 in America/New_York, the 48-hour offset fires on Saturday 31 October at 10:00 EDT. The local-clock rule “two days before at 09:00” fires at 09:00 EDT the same day. The hour of difference exists because US clocks changed on 1 November.
Neither rule is wrong. Offsets suit “two hours before” reminders, where elapsed time matters. Local-clock rules suit day-before reminders, which should arrive at a sensible hour. Avoid anchoring them near the changeover hour, where a local time can be skipped or repeated.
Apply a daytime sending window in the recipient’s time zone when you know it, and in the session’s zone when you do not. We use 08:00 to 20:00 as an illustrative default, and local law may set its own limits.
Which channel should carry each reminder?
Choose the channel from the consent the learner gave and the failures you can detect, not from habit. Email carries detail and is cheap, SMS is short and immediate, push reaches only learners with your app, and WhatsApp adds its own opt-in and template rules. Each has a gate to pass before sending.
| Channel | Gate before sending | Watch for |
|---|---|---|
| An authenticated sending domain; Google’s sender guidelines require SPF or DKIM from all senders and a spam rate under 0.3% | Bounces and spam complaints; senders above 5,000 messages a day to Gmail also need DMARC | |
| SMS | Recorded consent for the number; in India, registered headers and templates | Opt-out replies and undelivered receipts |
| Push | Learner has the app and allows notifications | Learners without the app never see it, so treat it as an add-on |
| Opt-in under the platform policy; approved templates outside the 24-hour window | Opt-outs made on or off WhatsApp must be honored |
We found no study comparing channels for ILT. Fall back to the next consented channel on a hard failure such as a bounce or an invalid number, not on a timer, which would send the same reminder twice.
Pro tip: Test the quiet failure. Register a test learner with a bouncing email and an invalid phone number. Confirm every job ends in a visible failed state with a reason that operations staff can see. Silent success is the failure most reminder systems never test.
What should training reminders contain?
A training reminder answers where, when, how to join, what to bring and what to do if the learner cannot come, in that order, with no promotion. Mixing marketing into a logistics message can change how some laws classify it.
- Date, start time, the time zone name, and local time for virtual sessions.
- Venue, room and directions, or the join link.
- Preparation steps and what to bring.
- A cancel link that releases the seat, so a likely no-show becomes a seat the waitlist can fill.
- A named contact and, for SMS, the opt-out instruction.
Store the template ID and version on every job so you can audit what a learner received. An illustrative SMS template:
Reminder: {course_title}, {start_local} {tz_abbr}, {room}.
Can't make it? Release your seat: {cancel_link}
Reply STOP to opt out.
How do you retry without sending duplicate reminders?
Design for at-least-once delivery and make duplicates impossible in the data layer. Give every planned send a natural key of registration, rule, channel and session version, enforce it with a unique constraint, and claim jobs with one atomic update before calling a provider. Retry only transient failures, and stop at a cutoff.
The snippet below shows the planning and claiming steps. We ran it on Python 3.13 with SQLite.
import sqlite3
db = sqlite3.connect(":memory:")
db.execute("""CREATE TABLE reminder_job (
registration_id TEXT, rule_id TEXT, channel TEXT, session_version INTEGER,
send_at_utc TEXT, status TEXT NOT NULL DEFAULT 'planned',
PRIMARY KEY (registration_id, rule_id, channel, session_version))""")
def plan(reg, version, sends): # safe to replay
for rule_id, channel, when in sends:
db.execute("INSERT OR IGNORE INTO reminder_job "
"(registration_id, rule_id, channel, session_version, send_at_utc) "
"VALUES (?,?,?,?,?)", (reg, rule_id, channel, version, when))
def supersede(reg, new_version):
db.execute("UPDATE reminder_job SET status='superseded' "
"WHERE registration_id=? AND session_version<? AND status='planned'",
(reg, new_version))
def claim_due(now):
cur = db.execute("UPDATE reminder_job SET status='claimed' "
"WHERE status='planned' AND send_at_utc<=?", (now,))
return cur.rowcount
In our run, delivering the same registration event twice still produced two rows, not four. After a reschedule, version 1 rows became superseded and version 2 rows were planned. The first poll claimed one due job and the second poll claimed none. The snippet omits the provider call, delivery webhooks and a sweep that returns stale claims after a worker crash.
Apply these retry rules around it:
- Retry timeouts, server errors and rate limits with exponential backoff and jitter.
- Treat bounces, invalid numbers and opt-outs as permanent: mark the job failed or suppressed.
- Set a cutoff per rule, since a “day before” reminder sent after the session starts is noise.
- Pass the job’s natural key to the provider as an idempotency key if it accepts one.
What do consent and opt-out rules require for training reminders?
The rules depend on country, channel and whether a message contains promotion, and several are moving. This section reflects primary sources read on 16 November 2026 and is not legal advice. Record consent per learner and channel with the wording shown, and have counsel review it before launch.
| Rule | What the source says | Design consequence |
|---|---|---|
| US TCPA consent | 47 CFR 64.1200: autodialed or prerecorded calls to lines including cell numbers need “prior express consent” (a)(1); calls with an advertisement or telemarketing need prior express written consent (a)(2) | Capture phone consent at registration, separate from marketing consent. Whether your sender counts as an autodialer is a legal question we did not research |
| US TCPA opt-out | Paragraph (a)(10): revocation by any reasonable method; “stop,” “quit,” “end,” “revoke,” “opt out,” “cancel” and “unsubscribe” qualify; honor within ten business days at most. The FCC adopted changes on 30 September 2026 (FCC 26-67): an opt-out to an informational message may be treated as covering only that category, and a caller may designate an exclusive method if it discloses it in the text. Effective 30 days after Federal Register publication, which was scheduled for 26 October 2026 | Honor STOP at once and treat it as “no more texts”. We have not seen the published version, so confirm the effective date (about 25 November 2026 by our arithmetic) |
| US CAN-SPAM | FTC guide: category is decided by a message’s primary purpose; transactional or relationship messages are largely exempt but “the law views these categories narrowly”; commercial email needs a postal address and opt-out honored within 10 business days; no business-to-business exception | A reminder for a booked session plausibly fits the transactional category. Keep promotion out |
| UK PECR | ICO guidance: “You must not send marketing emails or texts to individuals without specific consent”; a soft opt-in exists for existing customers with an opt-out offered at collection and in every message. The page covers marketing and does not address service messages | Keep reminders free of marketing, and offer an opt-out anyway |
| India TRAI | TRAI direction of 20 August 2024 under the 2018 regulations: headers and content templates must comply; one template may not link to more than one header; from 1 September 2024 messages with non-whitelisted URLs, APKs, OTT links or callback numbers are not permitted | Register templates with fixed slots and whitelist your link domain. We did not check later extensions |
| WhatsApp policy | Business Messaging Policy: you need the person’s number and “opt-in permission”; conversations start with an approved template; outside the 24-hour window only templates are allowed; opt-outs on or off WhatsApp must be respected | Store the opt-in event; map each reminder to an approved template. The page shows no date |
EU rules are not in the table because we did not read the primary EU texts for this post. Personal-data duties for phone numbers and consent records are in our guide to data privacy for training provider platforms.
How do you measure whether training reminders work?
Define a no-show as a confirmed registrant who neither attended nor cancelled before a stated cutoff, then compare cohorts that differ only in the reminder plan. Class sizes are small, so a weekly chart misleads. Plan for thousands of registrations per group before trusting a difference.
Here is an illustrative calculation, not a benchmark. Assume a 10% baseline no-show rate. With the standard two-proportion normal approximation, 5% two-sided significance and 80% power, we calculated about 2,005 registrations per group to detect a fall to 7.5%, and about 3,213 to detect a fall to 8%. At 200 registrations a week, filling both groups for the first case takes roughly 20 weeks.
Withhold only an optional tier from the comparison group, such as the final SMS, never confirmations or change notices. Track these alongside the no-show rate:
- Late cancellations and seats refilled from the waitlist.
- Opt-out rate per channel and template.
- Bounce, complaint and delivery-failure rates.
Pro tip: Log what you did not send. Record every suppressed reminder with its reason, such as no consent, opted out or session cancelled. Without suppression records, a no-show analysis counts learners who never received a reminder as learners who ignored one.
What should you do next?
List your session events and write each reminder as a rule with an offset or local time, a channel order, a cutoff and a template version. Add per-channel consent fields to registration. Then adapt the two snippets above to your own stack and run them against your real time zones.
Record your no-show rate for eight to twelve weeks before changing cadence. If you are evaluating platforms instead of building, ask each vendor to show its reminder rules, send ledger and opt-out handling. SimpliTrain’s training scheduling page, for example, states per-session time zone display adjusted for each learner and automatic delivery of join links to enrolled learners; confirm reminder behavior in a demo.
Frequently asked questions
How many reminders should you send before a training session?
The review we cited found several notifications more effective than one, but it did not identify an ideal count. A reasonable starting plan is a confirmation, a reminder a day before and a final one a couple of hours before. That is our suggestion, so test it against your own data.
Is SMS better than email for training reminders?
We found no study comparing channels for instructor-led training, so we cannot say. Email holds more detail, while SMS is short and immediate. Use each where you hold consent, let delivery failures decide the fallback, and compare your own results once you have enough registrations.
Do you need consent to send training reminders by text?
Usually you should hold recorded consent for the phone number. In the US, autodialed texts to cell numbers need prior express consent under 47 CFR 64.1200, and promotional content raises the bar. Rules differ elsewhere, for example under UK PECR and India’s TRAI regime, so have counsel review your registration form.
What should happen when a learner replies STOP?
Suppress further texts at once and log the request. The FCC rule text lists “stop,” “quit,” “end,” “revoke,” “opt out,” “cancel” and “unsubscribe” as valid revocation words and allows at most ten business days to honor a request. The FCC’s September 2026 changes let callers limit an informational opt-out to a category, so ask counsel before relying on that.
How do you stop duplicate reminders after a session is rescheduled?
Include the session version in each reminder’s unique key. When the session changes, increment the version, mark planned jobs for the old version as superseded and plan new ones. Replayed events hit the unique constraint and do nothing, so learners get one reminder per rule.
