The industry’s silent risk: why we rely on predictions we can’t track
September brought its familiar strain of quiet panic across international education. Results come out, conditional offers fall through, and everyone scrambles into damage control – agents fielding calls from furious parents, universities recalculating their yield numbers, pathway providers explaining themselves.
And every year, the sector shrugs it off as bad luck, instead of admitting it’s the predictable result of a system built on forecasts nobody actually checks.
International admissions rest on a structural paradox: universities award life-changing conditional offers based on a predicted grade – an educated guess.
The anatomy of an educated guess
The scale of this issue within the UK alone is systemic. UCAS data spanning three years found that only 16% of applicants received fully accurate predicted grades across their best three A-Levels, while 75% were overpredicted. The UK isn’t an outlier in relying on forecasts rather than results, it’s simply where the data happens to be public.
Framing this as a fault of overly generous teachers misses the point. The deeper issue is that a prediction, once made in October, is treated as a fixed judgement rather than a forecast that should update as new evidence arrives.
By the time a trajectory visibly deteriorates on a periodic assessment, the window for meaningful intervention has usually closed
The problem is not bad intention; it is an infrastructure gap. In almost every high-stakes global sector, operational risk is managed through continuous telemetry. In international education, academic progress is still evaluated through sparse, backward-looking checkpoints – a mock exam in January, a term report in March.
Between these isolated markers lies a black box. By the time a trajectory visibly deteriorates on a periodic assessment, the window for meaningful intervention has usually closed.
What we learned at ATMO: moving from checkpoints to tracking
When ATMO became a UCAS-registered center and began issuing our own predicted grades, accuracy ceased to be an abstract debate about admissions ethics; it became our daily responsibility.
Transitioning from episodic assessments to continuous progress tracking changed everything. Across ATMO’s tracking of more than 50,000 learning records and 120 million data points-assessment attempts, concept-mastery checks, and mentor session records – we shifted from periodic guesses to real-time academic visibility.
When algorithms flag a micro-drop in concept mastery in February, our mentors step in immediately – turning data telemetry into actionable teaching while there is still time to intervene. Discovering an academic deficit in August leaves room only for apologies; identifying it in real time preserves the placement.
None of this is a silver bullet, and it isn’t free – not every provider can build this overnight. But the starting point doesn’t have to be a full platform; it can be a shared calendar of continuous checkpoints instead of two annual ones. What isn’t sustainable is treating a once-a-year mock exam as the only signal.
Building a new trust infrastructure
For agents staking their brand equity on student conversion, and for universities managing enrolment pipelines, this kind of visibility is no longer an administrative luxury. It is the new trust infratructure. Families investing significant capital expect greater visibility into the likelihood of success, not just a number handed down once and left unchecked until results day.
To build a more resilient sector, the goal shouldn’t be to craft a better guess; it must be to replace speculation with a forecast worth trusting.
Universities can start by asking pathway providers a simple question: how do you know a predicted grade is accurate before results day, not after? Agents can ask the same before recommending a pathway provider to a family who’s paying for a trustworthy answer. The offer letter arrives before the learning is done. The question is whether anyone’s watching what happens in between.

About the author: Tetiana Serebrianska is the founder of ATMO School, a data-driven online school delivering live education to students across 39+ countries, powered by 120 million data points. With over seven years of experience in online education, Tetiana has built and managed fully digital schools serving more than 6,000 students.