How to build a study plan for a competitive civil-service exam

You decided to sit a competitive civil-service exam, downloaded the syllabus, and did what almost every beginner does on day 1: you opened a spreadsheet and spread the subjects across the days of the week — Portuguese on Monday, constitutional law on Tuesday, "general review" on Saturday. A few weeks later, an extra shift, a bout of flu, or a harder chapter breaks the table apart, and the feeling that lingers is one of a lack of discipline. It isn't: this kind of plan has a factory defect, and it's measurable. This guide builds the plan a different way (each choice with an experiment behind it) and starts with the answer: the complete skeleton in 5 steps, now.

The plan skeleton in 5 steps

The whole skeleton, before the reason behind each piece:

1. Diagnose the syllabus. List the subjects, break each one into numbered topics, and note how much each subject weighs on the exam. That inventory is the map of everything that comes next.
2. Build a subject cycle, not a day-by-day grid. Define a sequence (subject A → B → C → A…) in which the highest-weight ones appear more often, and move through it in whatever time slot the day allows.
3. End every session by testing yourself. Studying a topic includes closing the material and answering questions — practice questions or your own — about it, the same day.
4. Schedule the review of each topic on the day you study it. Every topic leaves the session with the date of its next encounter set. Review takes up space on the calendar as an equal to new material.
5. Close the week with a short audit. Compare what you planned with what you did, look at where errors went up, and reorder the cycle. The plan absorbs the unexpected by design, instead of breaking with it.

Step 4 is what separates this skeleton from the typical exam-candidate plan, and it's where most schedules fail. Said bluntly: a plan that doesn't schedule review isn't a plan; it's a reading list with dates. The next sections show the evidence behind each step and the exact point where manual execution tends to die.

Why the typical exam study schedule dies in month 2

Take the classic grid: a table with the days of the week, each cell a subject, each week advancing to new content. It has a feature almost no one notices: 100% of the time is allocated to seeing material for the first time, and 0% to meeting material you've already seen again. It's a plan for making friends that never sets a second meeting: each topic gets a great first conversation and is then left to fend for itself.

If what you study in March needs to be alive on the exam in September, someone has to schedule the second encounters, and the typical grid has nowhere to write them. What happens to material that's never revisited has had a measured curve since 1885; the full diagnosis is in why you forget what you study.

This defect isn't ignorance on your part: it's the statistics of method. In a survey of 177 college students (Karpicke, Butler & Roediger, 2009, Memory), 83.6% listed rereading their notes or the textbook as a study strategy, and 54.8% ranked it as their number-one strategy. Only 11% (19 people out of 177) reported testing themselves while studying, and just 1% as their main strategy. The typical plan is that habit put into a table: read, mark as seen, move on.

There's also the resistance factor. Researchers who measured how long a new behavior takes to become automatic found an average of 66 days, ranging from 18 to 254 across participants (Lally et al., 2010, European Journal of Social Psychology). In other words: the plan doesn't need to impress in week 1; it needs to still be standing on day 66. A rigid grid demands daily heroism, and heroism doesn't last two months.

Rereading isn't reviewing: the encounter that sticks is a test

The landmark experiment here: Roediger & Karpicke (2006, Psychological Science) had 120 college students study texts and compared rereading with self-testing. Five minutes after studying, rereading won: 81% versus 75% recall. Two days later, the scoreboard had already flipped: 68% for those who tested themselves, 54% for those who reread. A week later, 56% versus 42%: a fourteen-point advantage for self-testing, which didn't even come with feedback.

The second experiment, with 180 students, measured the dose: those who read the text once and tested themselves three times recalled 61% after a week; those who only reread, across four sessions (14.2 readings total versus 3.4), recalled 40%. And the detail that explains the addiction: the rereading group was the most confident it would remember (4.8 versus 4.0 on a 7-point scale) and performed the worst. Rereading manufactures fluency, and fluency deceives its owner. It's the trap Bjork (1994) calls desirable difficulties: the effort that rereading removes is exactly what builds memory.

It's not an isolated finding. When Dunlosky et al. (2013) reviewed ten study techniques, only two came out as high-utility: practice testing and distributed practice. Rereading, highlighting, and summarizing (the standard exam-candidate kit) landed on the low-utility shelf. That's why step 3 of the skeleton: a session that doesn't end in a test is a half-session. The mechanism of self-testing, and how to turn any material into a question, is the subject of the active recall guide.

Spacing the reviews yields more, with the same study time

The obvious objection to step 4: "if I set aside time to review, I'll see less new material." The evidence answers that the total time doesn't even need to change, only how it's distributed. A 2006 meta-analysis (Cepeda et al., Psychological Bulletin), with 839 comparisons drawn from 317 experiments, measured it: with the same total study time, spaced practice produced 47.3% correct answers on the final test versus 36.7% for massed practice (about ten percentage points), across data totaling 14,811 participants.

And the date of the encounter matters. In a 2008 experiment with 1,354 participants (Cepeda et al., Psychological Science), with study time held fixed, getting the interval between study and review right increased final-test recall by 64% compared to reviewing the same day. The rule of thumb that came out of it: the first review falls nicely in the 10-20% range of the time remaining until the exam. And when in doubt, erring toward a longer interval costs much less than erring toward a shorter one. The full mechanism, with the manual method, is in the spaced repetition guide.

Keep the consequence for the plan: the review date is a performance variable, not a scheduling detail. Two people with the same study hours reach the exam with different memories because one of them chose the days better.

A subject cycle beats a rigid grid

Step 2 also has an experiment. Rohrer & Taylor (2007, Instructional Science) compared practicing math problems in blocks by type (the equivalent of the "one day, one subject" grid) with practicing them interleaved. During practice, the block looked better: 89% correct versus 60%. A week later, on the test, the order reversed violently: 63% for those who interleaved versus 20% for those who studied in blocks. The paper describes the gain as a 250% improvement. The honest caveat: small sample (18 participants) and a procedural task. But the direction is the same as everything else on this page: the format that looks worse during training is the one that delivers on the exam.

And the cycle has an advantage no experiment needs to prove: logistics. A rigid grid breaks on the first extra shift: "I missed Monday's Portuguese" turns into an unpayable debt. In the cycle there's no lost day: you pick up where you left off, and the sequence guarantees that no subject goes weeks without appearing.

The competitive-exam study plan in practice: from syllabus to the calendar of encounters

Step 1: Diagnose the syllabus

List the subjects and break each one into topics small enough to fit in a single session. Note how much each subject weighs on the exam and your current level in it. A simple self-assessment, from 1 to 3, does the job. The product is a numbered inventory. The entire plan is managing that inventory, not a timetable.

Step 2: Turn the inventory into a cycle

Order the subjects into a repeating sequence, giving more entries to the highest-weight ones and to your weakest ones. Each session takes the next topic from the current subject and advances the pointer. No fixed days: the cycle runs Monday morning or Saturday night, no matter which.

Step 3: End every session in a test

Studied the topic? Close the material and answer: practice questions on that topic, or questions you wrote yourself while reading. Question-and-answer cards work well here. The discomfort of answering without looking is the active ingredient: in Roediger & Karpicke's data, it's what held the fourteen-point advantage a week later.

Step 4: Schedule the encounter before closing the notebook

Every topic studied today leaves with the date of its first review set, as a fraction of the time remaining: exam in three months, first review around two weeks out; longer deadline, longer interval. On review, the same format as step 3: material closed, questions first, checking afterward. What comes easily waits a longer interval until the next round; what stalls comes back in a few days.

Step 5: Audit the week

Set aside a fixed time at the end of the week for three questions: what fell behind in the cycle? Where did errors go up? What unexpected thing is next week already announcing? Reorder the cycle with the answers. That's the valve that lets the plan survive month 2: it doesn't break when life happens, because it already expects life to happen.

Where the manual plan breaks: the review bookkeeping

Build this system on paper and it works. For a few weeks. Then the arithmetic collects its due: each topic studied generates a review date, each review generates the next, and easy and hard topics ask for different intervals. A syllabus has hundreds of topics; after two months of the cycle, your calendar of encounters is a creature of hundreds of lines that changes every day. No mincing words: managing a review schedule by hand is a second syllabus. And that one isn't on the exam.

That's why this layer (and only this one) migrated to algorithms. Flashcard apps with spaced repetition take over the bookkeeping of dates; the honest analysis of whether Anki is worth it shows what the category solves and the friction cost that the most famous tool charges. In Sift, the app we built, this is the part that Sprint takes over: the topics you turn into questions enter the queue with their own curve, and each one comes back on the date when review yields the most. The syllabus diagnosis, the cycle, and the hours stay yours. The bookkeeping ceases to exist.

What to do today

The whole plan isn't built in a day. These five moves are:

1. Turn the syllabus into an inventory: subjects, numbered topics, the weight of each.
2. Write your cycle on a single line (the highest-weight subjects appear more often) and keep it in sight.
3. Study the first topic of the cycle and end the session with the material closed, answering questions about it.
4. Set the first review of that topic as a fraction of the time until the exam, and write the date somewhere you look at every day.
5. Block a fixed audit time at the end of the week. That commitment is worth more than any hours target.

Frequently asked questions

How many hours a day should you study for a competitive exam?

The evidence says less about volume and more about format: in Roediger & Karpicke (2006), those who reread the material 14.2 times lost to those who read it 3.4 times and tested themselves; in Cepeda et al. (2006), the gain of about ten points came with the same total study time. Before adding hours, make sure the ones you already have end in a test and have a review scheduled.

Should I study more than one subject per day?

The interleaving data suggests yes: in Rohrer & Taylor (2007), those who interleaved problem types scored 63% on the test a week later, versus 20% for those who practiced in blocks (with the caveat of a small sample). And the cycle comes with the logistical bonus: no surprise takes down "the day" of a subject.

When should I do the first review of each topic?

As a fraction of the time until the exam: the 10-20% range is the rule of thumb from the Cepeda et al. (2008) experiment. Exam in three months, first review around two weeks out. When in doubt, pick the longer interval: erring long costs much less than erring short.

Does the 24-hour, 7-day, 30-day review schedule work?

It's a reasonable approximation for one- to two-month deadlines, not a law of memory: the Cepeda et al. (2008) data show that the right interval grows along with the time until the exam. A fixed schedule treats the September exam and next week's exam with the same ruler, and one of the two ends up losing out.

Do I need to finish the theory before starting on questions?

No. And waiting to "be ready" works against you. In Roediger & Karpicke (2006), self-testing beat rereading even with no feedback at all (56% versus 42% after a week); in Rohrer & Taylor (2007), the group that made the most errors during practice was the one that scored highest on the test. Getting it wrong while answering is training, not a symptom of being underprepared.

The skeleton is on the table: syllabus into inventory, inventory into cycle, a session that ends in a test, and a review with a date — a plan that doesn't depend on heroism to reach day 66. The one piece that doesn't fit into any amount of discipline is the bookkeeping of the encounters, and that's exactly the piece that Sift takes over: each topic becomes a question, each question comes back on the right day of the encounter, and the calendar of hundreds of lines never even gets born.