What a good study plan looks like
What usually gets handed over as a plan
Ask for a study plan and you usually get a syllabus: a list of grammar topics in a conventional order, a vocabulary set per unit, and an implied promise that working through it produces fluency. There is nothing wrong with the contents. What is missing is any information at all about the person following them.
A real plan answers four questions that a syllabus does not. What are you trying to be able to do. By when. How many hours a week do you actually have. And what, specifically, is currently in the way. Change any one of those answers and the correct sequence changes with it — which means a plan that would come out identical for two different people has told you it is a table of contents.
It starts from questions asked before any speaking happens
The inputs are not exotic, and most of them can be collected in a couple of minutes of form-filling before a first conversation ever happens.
The goal, in the sense of a domain rather than an aspiration. Work, an exam, a relocation, travel. These produce genuinely different courses: relocating means bureaucracy, housing, healthcare and schools; work means meetings, updates, negotiation and written register carried into speech; an exam means a specific format with specific tasks. “Improve my English” is not a goal, it is a mood.
The timeline. A deadline does more than apply pressure; it is the only thing that makes pacing decisions answerable at all. Whether to spend three weeks consolidating the past tenses or move on and come back is an unanswerable question in the abstract and an easy one if there is a date.
The weekly time budget, honestly stated. This is the input people most often inflate and the one that most determines whether the plan survives contact with reality. A plan pitched at five hours a week for someone who has two will simply be abandoned, and abandonment costs more than slow progress ever does.
Interests. Not decoration. Most objectives can carry almost any subject matter — justifying an opinion or narrating a past event works as well through cycling as through supply chains — so wherever that choice is open, spending it on a topic the learner is indifferent to is a wasted parameter. Some objectives are bound to their domain and cannot move: ordering food, asking directions, telling the time. Those come as they are.
A self-assessed level and a daily practice goal, both of which are rough and both of which are useful — the first as a prior for the initial estimate, the second because a daily target you set yourself is the one you might keep.
Then the first conversation refines all of it. A form records what you believe about your English; a conversation records what you produce under time pressure. Where those two disagree, the plan needs both — the belief explains what you will find discouraging, and the production decides what to teach.
Personalized inside a fixed structure, not invented from scratch
Here is the design decision that matters most, and it is one that cuts against the instinct to make everything bespoke.
The pedagogical sequence should not be generated fresh for each learner. Underneath every personal plan sits a curriculum skeleton for the band, and it is a fixed document: a numbered set of can-do objectives with stable identifiers, an ordered grammar sequence, the vocabulary domains that belong to the band, and explicit exit criteria naming which objectives must be mastered before the next band opens. It is written once per band, it is the same skeleton for everyone at that band, and it does not change because a learner is interested in cycling.
What personalizes is everything above it: which objectives come first given the goal, how fast the sequence moves given the timeline and the hours available, which scenarios carry the practice, which review items are due, and what the material is about. That is a lot of personalization. It is just not personalization of the pedagogy.
The reason for the split is quality consistency. A system that invents the progression per learner will produce a good sequence for some people and a subtly incoherent one for others, and nobody will be able to tell which they got. A fixed skeleton makes the failure mode visible: either the plan follows the sequence or it does not.
It is a living document, and the changes should be legible
A plan written once and followed for twelve weeks is a forecast, and forecasts about learning are bad. The plan should change after every lesson, because every lesson produces evidence — objectives that moved from being introduced to being practiced, errors that turned out to be entrenched rather than incidental, a level estimate that firmed up or moved.
But “it adapts” is exactly the kind of claim that hides a lack of mechanism. The useful test is whether you can see what changed and why. A plan revision should be a new version, not an overwrite, with a record of what generated it. If the sequence moved because the analysis of your last lesson found a recurring pattern, that should be inspectable rather than a mysterious reshuffle.
There is a second reason for versioning. If a human teacher edits a plan — reorders objectives, adds a focus, slows the pacing — an automated process must not silently undo that on its next pass. The safe rule is that automated analysis may keep updating the mechanical state, the review schedule and the mastery flags, but structural changes to a human-edited plan are proposed rather than applied. Automation that quietly overrules a teacher is worse than automation that asks, because the teacher stops trusting it and stops editing.
Progress framed against your own deadline
How progress is reported shapes what learners optimize for, which is a good reason to report it carefully.
The tempting format is a number that goes up. It is satisfying, it is easy to render, and it is almost always indefensible — “your level grew three points this week” invites the question “measured how, against what,” and there is rarely a rigorous answer. A number nobody can defend is a number that will eventually be used to make a decision it cannot support.
The alternative is to frame movement against the learner’s own stated goal and date. On track for a given band by a given month, or not on track and here is the gap. That framing is falsifiable, which is uncomfortable and correct. It also keeps the plan honest about the hours: if the weekly budget has not been met for a month, the timeline has moved, and saying so is more useful than quietly re-baselining.
Alongside it, the specifics: which objectives are done, which are in progress, what the next session targets. Those are the things a learner can act on. A short written note from the tutor tying the week together — what improved, what is next, how it relates to the deadline — does more for continuity than any progress bar, because it is addressed to a person rather than displayed at one.
What a good plan does not contain
No promise about how long fluency takes. The honest variable is hours of production, and nobody can predict yours.
No list of topics without an order and a reason. If the plan cannot say why this before that, the sequence is decorative.
And no fixed content for lesson twelve. A plan that already specifies a lesson three months out either is not reading the evidence from lessons one through eleven, or is going to ignore what it reads.
What it should contain is unglamorous: a goal, a date, an honest hour count, an ordered set of can-do objectives with their current states, the scenarios that will carry them, and a revision history showing how it has moved. If you want the vocabulary for the objectives themselves, what CEFR levels actually mean covers the descriptors, and the method page describes the structure the plan personalizes within.