Every month inside a learning platform builds an asset the price page never mentions: your learner model, the map of what you got wrong, when you last saw each topic, which explanations you needed, where your weak domains sit and how they trend. That asset is why month six of any decent platform feels better than month one, and it is also the lock: change resources and the model stays behind, resetting you to a stranger the new system must relearn, which is a switching cost measured in weeks of miscalibrated practice. This article names the cost, the questions that price it before purchase, and the portability standard the category should be held to, ourselves included.
What the learner model actually contains, and what leaving costs
Decompose the asset and the switching cost becomes concrete. Performance history: item-level rights and wrongs, the raw material of every analytic. Weakness map: the derived picture of domains below your floor, which drives adaptive targeting. Scheduling state: where every topic sits in its spacing cycle, the invisible layer whose loss silently resets months of retention engineering. Created material: notes, flags, flashcards, saved explanations, your own annotations on the platform's content. And tutoring history: the record of your misconceptions and their repairs. Leave a platform and the created material sometimes exports; the model, weaknesses, scheduling, misconception history, usually does not, and the new platform's first month is spent rediscovering what the old one knew, during which its adaptive features run on noise. The cost is largest exactly when switching is most tempting, mid-preparation, which is the lock's actual geometry.
The export questions to ask before subscribing
Five, answerable from documentation or one support message, and worth asking while you are still a prospect. Can I export my question history, item-level, in a usable format, or only summary statistics? Can created content, notes and flashcards leave in a standard format that another tool can import? Does anything preserve scheduling state, or does spacing restart from zero elsewhere? What happens to all of it on cancellation and on account deletion, retained, frozen, destroyed, and on what timetable? And is any of this documented in writing, because export capabilities described only by sales conversations have a way of versioning. Closed Qbank ecosystems, upload-based systems and interoperable study tools answer these very differently, and the answers should weight long subscriptions especially: the more months you will invest, the more the exit terms matter, which inverts how most students shop.
Reducing exposure, and the standard worth demanding
Practical mitigations first. Keep the durable layer portable by construction: your own error log, the misconception list from tutoring sessions, the weak-domain map, maintained in your own notes in your own format, ten minutes weekly that makes any future platform's cold start warm. Prefer platforms whose scope reduces switching frequency, one reason multi-exam continuity matters, a learner model that carries from UKMLA through postgraduate banks amortises across a career stage rather than a single season, which is iatroX's architecture and, stated with matching honesty, the same lock-in geometry as everyone else's within its walls: continuity inside the platform is real, and users should check the current account tools for what can and cannot presently be exported rather than take any vendor's direction of travel, ours included, as a shipped feature. And the category standard worth demanding out loud: item-level history export in open formats, created-content portability, and documented deletion, because learning history is the student's professional development record in embryo, and the vendors who treat it that way first will deserve the loyalty the lock currently simulates.
Frequently asked questions
Is lock-in always bad?
The model's value is real, better targeting, honest analytics, so depth in one platform is often rational; the problem is unpriced exit, and the questions above convert it from surprise into informed trade.
Do free tools avoid the problem?
They invert it: general chatbots hold little durable model, so there is nothing to lose and nothing accumulating, the context-rebuilding tax paid daily instead of at exit; both costs are real, and naming them is the comparison.
What single habit most protects a multi-year learner?
The weekly portable log: errors, misconceptions, weak domains in your own file; it is the one copy of your learner model no subscription decision can touch.
Do interoperability standards exist for learning data?
Fragments do, flashcard formats travel, question histories mostly do not, and the absence of a common learner-model standard is why the portable personal log remains the only reliable interchange format.
Should schools negotiate export rights in procurement?
Explicitly, on students' behalf: institutional contracts are where portability clauses actually get agreed, and the rubric's item nine exists precisely so the question is asked before signature.