Live sessions are the highest-cost, highest-value moment in any technical training program. A senior engineer is blocking two hours to walk fifteen learners through Kubernetes RBAC failures, or a data platform architect is auditing a learner's dbt project on shared screen. When a session gets cancelled, whether by the learner, the tutor, or Refonte Learning itself, the cost lands somewhere: on the learner's momentum, on the tutor's paid hours, or on the program's completion timeline. A cancellation policy is not a legal footer. It is an operational contract that says who absorbs which cost, under which conditions, and how fast the makeup happens.
This article documents how the Refonte Learning live session cancellation policy actually works in 2026. It is written for enrolled learners who need to reschedule, for tutors who need to know their notice obligations, and for prospective students evaluating whether the program is operationally serious. The policy sits inside the broader live session quality standards framework, and if you have not read that pillar, the cancellation rules will make more sense once you understand what a live session is supposed to deliver in the first place.
Why cancellation policies matter more than learners realize
Most prospective learners skim the cancellation section of an enrollment agreement the way they skim a car rental contract: they assume it will not apply to them. It always applies to somebody. In a 24-week cohort with weekly live sessions, roughly one in four learners will trigger at least one cancellation event, and roughly one in ten cohorts will see a tutor-side cancellation because of illness, travel disruption, or an emergency. Multiply that by hundreds of active cohorts and the policy is running dozens of times a week.
The stakes are not symmetric. A learner who cancels a 1-on-1 session on ninety minutes notice has cost the tutor a two-hour billable block that cannot be resold. A tutor who no-shows a cohort session has cost fifteen learners two hours of coordinated calendar time, plus the psychological cost of a broken commitment. A platform that cancels an entire cohort week has cost everyone. Each of these events needs a different rule, because pretending they are the same event produces either learner-hostile policies or tutor-exploitative policies.
Refonte Learning treats the cancellation policy as a load-bearing document. It is versioned, it is dated, and every enrolled learner and contracted tutor receives the current version at onboarding. When the policy changes, the change is announced with at least fourteen days notice and the previous version stays accessible for anyone who enrolled under it. That grandfather clause matters: if you enrolled in September 2025 under the previous rescheduling window, you are not silently downgraded when the 2026 revision ships.
The policy is also mechanically enforced through the scheduling system. When a learner clicks reschedule at eight hours before session start, the system knows it is inside the twelve-hour paid window, applies the credit rule automatically, and logs the event. There is no ambiguity where a support agent has to interpret intent, which is where most edtech cancellation disputes actually happen. Removing human judgment from the enforcement layer removes most of the acrimony.
The three cancellation categories
Every cancellation at Refonte Learning falls into one of three categories, and the entire policy is structured around this taxonomy. Getting the category right is the first step in any dispute, because the rules diverge sharply from that point forward.
Learner-initiated cancellation covers any case where the enrolled learner cannot or does not want to attend a scheduled session. This includes rescheduling requests, outright cancellations, and no-shows where the learner never joined the call. The subcategories matter because a rescheduling request made four days in advance is treated very differently from a no-show.
Tutor-initiated cancellation covers cases where the contracted tutor cannot deliver the session. Illness, family emergency, technical failure on the tutor's side, and scheduling conflicts all fall here. Refonte Learning holds tutors to a stricter notice standard than learners, because tutors are the paid party in the relationship and because a tutor cancellation affects a full cohort, not a single seat.
Platform-initiated cancellation covers cases where Refonte Learning itself cancels the session. This is rare and usually involves either a force majeure event (widespread infrastructure outage, regional emergency) or a quality intervention where the platform has determined that the scheduled tutor should not deliver the session (typically after a serious complaint or a technical readiness failure). Platform cancellations always trigger full makeup obligations at no cost to the learner, and the tutor is paid according to the terms of their contract regardless of whether they delivered.
A fourth pseudo-category exists in practice: mutual reschedule. This is when the tutor and all cohort members agree to move a session, typically because of a public holiday or a scheduling conflict that surfaces after the calendar is published. Mutual reschedules do not trigger the notice or credit rules at all, provided they are logged in the system and all parties confirm. Most of the smoothest schedule changes at Refonte Learning are mutual reschedules that never touch the formal cancellation machinery.
Learner-initiated cancellation: the notice windows
Learner cancellations are governed by a tiered notice window that maps to how much operational cost the cancellation imposes. The 2026 policy uses four windows, measured from the scheduled session start time in the learner's local timezone.
More than seventy-two hours notice: free reschedule, no session credit consumed, no fee. This window covers the vast majority of legitimate cancellations. If you know on Sunday that you cannot make Wednesday's session, you reschedule for free.
Between twenty-four and seventy-two hours notice: free reschedule for cohort sessions, one session credit consumed and returned upon attendance of the makeup for 1-on-1 sessions. The asymmetry is deliberate. A cohort session runs regardless of whether one learner reschedules, so the operational cost is near zero. A 1-on-1 session inside seventy-two hours means the tutor has blocked the calendar and may have started preparation, so a small credit gate discourages casual last-minute moves without penalizing genuine emergencies.
Between twelve and twenty-four hours notice: session credit consumed, tutor is paid in full for the scheduled block, makeup session available at learner cost or from remaining credit balance. This is the window where the operational cost becomes real. The tutor has almost certainly completed preparation and has committed the calendar block in a way that cannot be resold.
Less than twelve hours notice, or no-show: session credit fully consumed, no automatic makeup, tutor paid in full. The learner can request a discretionary makeup, and Refonte Learning grants these liberally for documented emergencies (medical, bereavement, force majeure), but the default is that the session is spent.
The distinction between cohort sessions and 1-on-1 sessions runs through every window, which is why we recommend reading the 1-on-1 mentoring vs cohort sessions breakdown before you look at your own schedule. If your program mixes both formats, the same calendar week can contain sessions governed by different sub-rules.
All notice is measured from the moment the reschedule or cancellation is submitted in the scheduling system, not from the moment an email is sent to support. Learners who email support at eleven hours before session start and then complain that they gave twenty-four hours notice will find that the system timestamp governs. This is not bureaucratic rigidity, it is the only enforceable standard.
No-shows: the hardest category
No-shows are the failure mode that a cancellation policy exists to price. A no-show is defined as a learner who has not joined the session within fifteen minutes of scheduled start and has not submitted a cancellation or reschedule request. At that point the tutor is authorized to end the session for a 1-on-1, or to proceed with the remaining attendees for a cohort session.
For 1-on-1 sessions, the fifteen-minute rule means the tutor waits, attempts to reach the learner through the scheduled channel, and then closes the session. The learner is charged a full session credit. There is no partial credit for late arrival past fifteen minutes because the operational reality is that a two-hour session cannot compress into ninety minutes without losing structure.
For cohort sessions, no-shows are recorded but the session continues. The absent learner loses the live interaction and is expected to catch up through the session recording, which is published within seventy-two hours per the live session format documentation. Cohort no-shows do not consume additional credit because cohort attendance is not credit-metered in the same way as 1-on-1 time.
The hard case is the repeat no-show. A learner who no-shows three or more times in a rolling ninety-day window triggers an academic review. This is not punitive, it is diagnostic. Repeat no-shows usually signal one of three underlying issues: the learner's life circumstances have shifted and the program schedule no longer fits, the learner has disengaged and needs either a program pause or an exit, or there is a specific interpersonal or content issue with the current tutor that is not being surfaced through normal channels. The academic review exists to catch the third case, because a learner who quietly stops showing up because of a tutor conflict is a learner that Refonte Learning has failed.
Documented emergencies always override no-show consequences. Medical documentation, evidence of a family bereavement, or verifiable force majeure (regional power outage, natural disaster) restore the consumed credit and remove the no-show from the academic record. The bar for documentation is reasonable, not adversarial. A note from a clinic or a screenshot of an outage map is sufficient.
Tutor-initiated cancellation: stricter rules, faster makeup
Tutors operate under stricter notice standards than learners because they are the paid party and because their cancellation has a multiplied effect. A tutor cancelling a cohort session of fifteen learners has cancelled thirty person-hours of committed calendar time, not two.
The standard tutor notice window is seventy-two hours for anything short of a genuine emergency. Tutors who need to reschedule for personal reasons, scheduling conflicts with other work, or non-emergency travel are contractually required to give seventy-two hours notice and to propose at least two alternative session times that fall within the same cohort week or, at maximum, the following week.
Inside seventy-two hours, tutor cancellations are treated as emergency events. The tutor is required to notify Refonte Learning operations directly, not just the learners, so that a substitute tutor can be considered. For high-demand technical topics where substitution is difficult, the session is rescheduled with the same tutor for the earliest mutually available slot, and learners are offered a partial session credit as compensation for the disruption.
Tutor no-shows are treated with a completely different severity. A no-show tutor triggers immediate operations contact, and the session is either recovered through a substitute delivery within twenty-four hours or fully rescheduled with makeup credit to every affected learner. Repeat tutor no-shows are a termination trigger under the tutor contract. Refonte Learning does not run a policy where tutors can quietly drift into unreliable delivery, because unreliable delivery is the single fastest way to destroy a cohort's trust in the program.
Technical failures on the tutor side are a specific subcategory. If a tutor's internet, audio, or screen-sharing setup fails at session start and cannot be recovered within ten minutes, the session is cancelled and rescheduled at no learner cost. Tutors are expected to have tested their setup against the technical requirements for live sessions before every session, and repeat technical failures trigger a mandatory equipment and connection review.
Platform-initiated cancellation and force majeure
Refonte Learning cancels sessions itself in three scenarios, and each has a defined recovery path.
The first is force majeure: infrastructure outages affecting the video platform, regional emergencies affecting a significant portion of a cohort, or events that make delivery impossible or inappropriate. The 2020-2022 period taught the entire edtech industry that force majeure clauses need to be specific and operational, not vague. Refonte Learning defines force majeure narrowly: events outside the reasonable control of both the platform and the tutor that make session delivery impossible. Force majeure cancellations trigger full makeup with no credit consumption for learners and full contracted payment for tutors.
The second is quality intervention. If Refonte Learning determines, based on recorded evidence or serious complaints, that a scheduled tutor should not deliver a specific session, the platform cancels and either substitutes or reschedules. This is a rare but important power, because it is the enforcement mechanism behind the quality standards. Quality-intervention cancellations do not affect the tutor's contracted pay for the cancelled session but may trigger contract review depending on the underlying issue.
The third is scheduling error. Every operational system produces errors: double-booking, timezone mistakes, calendar sync failures. When Refonte Learning cancels because of its own error, the makeup is prioritized above other scheduling and the affected learners receive a documented apology plus a discretionary credit. The platform does not attempt to hide behind the tutor or the learner when the fault is operational.
All platform-initiated cancellations are logged in a monthly operations review. The pattern matters: if a specific cohort, tutor, or timezone band is showing an elevated platform cancellation rate, the underlying cause is investigated. This is the same continuous-improvement loop that runs on tutor quality scores and learner NPS.
Refunds versus makeups: the money question
The cancellation policy is separate from the refund policy, and confusing the two is the single most common learner misunderstanding. A cancellation is a scheduling event that affects a single session or a small number of sessions. A refund is a financial event that affects the enrollment contract as a whole.
Makeup sessions are the default remedy for almost every cancellation category. Refonte Learning would rather deliver the session than refund the enrollment, because the learner enrolled to gain the skill, not to get their money back. The makeup pathway is fast: for tutor-initiated and platform-initiated cancellations, the makeup is typically scheduled within one week. For learner-initiated cancellations with credit consumed, the makeup is available whenever the learner has remaining credit balance and can find a mutually available slot.
Refunds enter the picture in three cases. First, when a learner exits the program under the enrollment agreement's withdrawal terms and has unused session credits. Second, when platform-initiated cancellations accumulate beyond what makeup can reasonably cover (a very rare threshold, but it exists to protect learners from a systematically failing cohort). Third, when a documented Refonte Learning error has caused loss that a makeup cannot repair, such as a certification deadline missed because of a platform scheduling failure.
The refund calculation is pro-rated against delivered sessions, not calendar time. A learner who has completed sixty percent of scheduled sessions has consumed sixty percent of the enrollment value, regardless of whether they finish in twenty weeks or thirty. This calculation is documented in the enrollment agreement and shown in the learner dashboard, so there is no surprise at exit.
Disputes over refund calculations follow the formal dispute resolution mechanism, which was built for tutor contract disputes but applies to any structured disagreement over Refonte Learning policy application. The mechanism is designed to be fast and evidence-based rather than adversarial.
Rescheduling: the mechanics that make or break the policy
A cancellation policy is only as good as the rescheduling system that implements it. If learners cannot easily find replacement slots, the policy becomes a source of friction regardless of how fair the written rules are.
Refonte Learning uses a two-tier rescheduling system. Tier one is the standard self-service reschedule, available through the learner dashboard. The learner sees available slots for their assigned tutor (or, for cohort makeups, available cohort sessions on the same topic) and selects a new time. The system enforces the notice window rules automatically. Most reschedules resolve at tier one within minutes.
Tier two is assisted rescheduling, used when the standard system does not surface an acceptable slot. This happens most often for 1-on-1 sessions with specialist tutors whose availability is limited, or for learners in unusual timezones. Assisted rescheduling is handled by operations, typically within one business day, and expands the search: alternative tutors with equivalent expertise, out-of-band scheduling with the primary tutor, or cross-cohort makeup for cohort sessions.
There is no cap on how many times a single session can be rescheduled, but there is a cap on how long a session credit remains valid: twelve months from issue. This is generous compared to industry norms and it exists so that learners dealing with genuine life disruptions (job change, family event, health issue) have a real recovery window. A learner who needs to pause the program can bank their credits and resume, provided the resumption fits within the twelve-month window.
The rescheduling system respects timezone metadata rigorously. Every session is stored in UTC and displayed in each participant's declared local timezone. Daylight saving transitions, which historically caused the most cancellation confusion, are handled by anchoring to UTC rather than to local wall-clock time. Learners who cross timezones during a program (for travel or relocation) can update their timezone in the dashboard and see all future sessions immediately reflect the change.
Edge cases the policy is built to handle
A cancellation policy that only handles the common cases is not a policy, it is a marketing document. The 2026 revision includes explicit handling for the edge cases that produce most support tickets.
Illness mid-session. A learner who becomes ill during a session can leave and is entitled to a partial credit restoration if they left before the halfway mark. Tutors experiencing acute illness mid-session are expected to end the session and reschedule; no learner credit is consumed.
Connectivity failure on the learner side. A learner who cannot connect due to their own internet failure is expected to attempt reconnection for up to fifteen minutes. If reconnection fails, the session is treated as a partial cancellation with a discretionary credit review. Repeated connectivity failures suggest a technical setup issue that operations will help troubleshoot.
Cohort session with tutor present but learner-side quorum lost. If a cohort session is scheduled and fewer than half the enrolled learners join, the tutor may choose to proceed as an office-hours-style session or to reschedule. This decision is at the tutor's discretion and neither option consumes learner credit.
Cross-timezone cohort scheduling conflicts. Cohorts with learners spanning wide timezone ranges occasionally hit scheduling conflicts that cannot be resolved with a single session time. The policy allows for split cohort delivery, where the same session content is delivered twice at different times, or for a permanent time shift agreed by cohort vote.
Public holiday overlaps. Sessions scheduled on public holidays in the tutor's country are automatically flagged for reschedule at cohort creation. Sessions scheduled on public holidays in learner countries that were not caught at creation are treated as mutual reschedules if raised at least seventy-two hours in advance.
Program pause requests. Learners can request a formal program pause of up to ninety days for documented reasons (medical, bereavement, major life event). Paused programs stop the session schedule entirely and do not consume credits during the pause. Resumption picks up where the program left off, adjusted for any curriculum updates in the interim.
Each of these edge cases has been reduced to a specific rule because ambiguity in edge cases is where policies fail. A learner who is genuinely ill should not have to argue their case, they should read the rule and know what happens next.
How the policy interacts with certification and completion deadlines
Refonte Learning programs that lead to certification or portfolio completion have deadline structures that interact with the cancellation policy in specific ways. A learner who has consumed cancellations up to the point where remaining scheduled sessions cannot fit before a certification deadline is flagged in the dashboard, and operations reaches out with recovery options.
Recovery options include intensified scheduling (two sessions per week instead of one for a defined recovery period), substitution of asynchronous project work for a subset of live sessions (subject to program requirements), or a formal deadline extension. Extensions are granted based on the pattern of cancellations: a learner whose cancellations were mostly platform-initiated or emergency-documented will receive extensions readily, while a learner with a pattern of casual short-notice cancellations will be offered scheduling intensification rather than an easy extension.
This is a place where the policy actively defends the learner's outcome. Refonte Learning does not want a learner to enroll, cancel their way through half the program, and then discover they cannot certify. The dashboard shows a projected completion date that updates in real time based on scheduling and cancellation history, so learners can see the trajectory well before it becomes a crisis.
For the AI Engineering program specifically, where the internship placement component has its own deadline structure, the cancellation policy is tuned tighter. Learners in the placement track have a lower cancellation allowance before operations intervenes, because a placement match depends on the learner being certified on a specific timeline that placement partners have committed to. This tighter policy is disclosed at enrollment and is part of what makes the placement track viable.
Comparing the 2026 policy to prior versions
The 2026 revision made three substantive changes to the previous policy, and it is worth understanding what changed and why.
First, the notice window for free rescheduling was extended from forty-eight hours to seventy-two hours. This was learner-friendly and reflected data showing that most legitimate rescheduling needs surface more than two days in advance, so the tighter window was penalizing edge cases without meaningfully protecting tutor calendar cost.
Second, the tutor no-show threshold for contract review was tightened. Under the previous policy, a tutor's third no-show in a twelve-month window triggered review. Under 2026, the second no-show triggers review. This reflects a strategic decision that tutor reliability is a first-order product feature and that the platform will not tolerate drift.
Third, credit validity was extended from nine months to twelve months. This was in response to learner feedback that nine months did not accommodate the reality of career transitions and life disruptions that often precipitate a program pause.
The policy did not change in several places where changes were considered and rejected. The twelve-hour hard credit-consumption window was retained despite pressure to move it to twenty-four hours, because the operational cost of last-minute cancellation is real and shifting it further out would either raise program prices or reduce tutor pay. The no-show fifteen-minute rule was retained because shorter windows produce more tutor-side disputes and longer windows waste too much scheduled time.
What this policy tells you about Refonte Learning
Cancellation policies reveal what a training program actually optimizes for. A policy that makes cancellation impossible is optimizing for revenue capture at the expense of learner circumstance. A policy that makes cancellation free at any time is optimizing for enrollment marketing at the expense of tutor reliability and program completion. Neither extreme produces a program that actually delivers outcomes.
Refonte Learning's policy sits in the middle by design. It absorbs learner-side friction generously outside seventy-two hours because that is where legitimate need lives. It holds a line inside twelve hours because that is where operational cost becomes real. It holds tutors to a stricter standard than learners because tutors are the paid party and their reliability defines the program. It ships makeups aggressively because delivery of the skill is the actual product, not the calendar event.
The policy also reveals what Refonte Learning treats as non-negotiable. Learners can cancel, tutors can cancel, the platform itself can cancel, but the underlying commitment to deliver the curriculum content to every enrolled learner does not cancel. Every cancellation in the system triggers a recovery path, and the recovery paths are engineered to complete before program deadlines. This is why the policy is bundled into the broader quality standards rather than being a separate legalistic document.
If you are evaluating whether to enroll, read the cancellation policy alongside the AI Engineering program curriculum and the tutor standards. The three documents together describe what you are actually buying: a scheduled delivery of instruction, backed by operational commitments about what happens when scheduling breaks down. That combination is what makes a training program a program rather than a sequence of videos.
About Refonte Learning
Refonte Learning is an EdTech platform delivering professional training in AI, data engineering, cloud infrastructure, DevOps, and software engineering. Programs combine live cohort instruction, 1-on-1 mentoring, and structured internship placements. The company is operated by Refonte Infini Infiniment Grand, a French SAS registered under SIREN 949 841 605, with an operational office at 1 Poulton Close, Dover, Kent, United Kingdom, CT17 0HL. All policies referenced in this article, including the cancellation policy, are versioned documents available to enrolled learners and contracted tutors through the platform dashboard.