What "the plan" actually means, 13 weeks out
My CAT 2026 exam is on 29 November. I'm writing this on 25 August, which puts me at 13 weeks out. I've logged 340 study sessions and 23 full mocks since I started tracking, and this post is not a generic template - it's the actual shape that preparation took, month by month, plus what I'm changing for the stretch that's left. Where a claim needs a number, it comes from my own log, not from a formula.
The one honest caveat up front: a plan written in June looks tidy in hindsight because I already know which parts worked. It did not feel tidy in July.
May-July: building the base, and the mocks show it
I started full mocks in May, and the scores from that stretch look exactly like what you'd expect from someone still building fundamentals: 41, 65, 74, 63, 44, 68, 91, 63, 47, 54 across ten mocks between 10 May and 28 June. That's not a trend - it bounces from 41 to 91 and back down to 44 inside six weeks. If you're in this phase and your scores look like that, it isn't a sign you're doing something wrong. It's what building-phase mocks look like before technique catches up with content.
The one steady thing underneath the noise was the daily reading habit. I log it as its own subject - NEWSPAPER - separate from VARC, because reading for the habit and drilling RC sets are different activities and I wanted the heatmap to show both honestly. A subject that exists only to track a habit, with no test attached to it, is one of the more useful things Karma Yogi lets you do that a plain notebook doesn't.
August: the mocks stopped being polite about it
Here is the stretch that actually tested whether I trusted the process: 96 on 29 July, 40 on 30 July, 94 on 1 August. Three mocks, four days, a 56-mark swing down and a 54-mark swing back up. The 40 wasn't a fluke confined to one section either - VARC 9, DILR 2, QUANT 29 that day, all three down at once, which reads more like a bad day than a specific gap. Two days later I was back at 94.
I did not adjust my study plan on the strength of that one 40. That's the actual point of this section: a single mock, taken in isolation, tells you almost nothing about your level. Only the run of numbers around it does. The rest of August held up the case - 84 (3 Aug), 91 (8 Aug), 80 (15 Aug), 62 (22 Aug), 72 (24 Aug) - percentiles mostly in the 92-97 band even while the raw score kept moving. Percentile turned out to be the steadier number of the two, which is worth its own read if you haven't seen the data - see what 45 real score-to-percentile pairs actually show.
The number that's actually driving the next 13 weeks
My weekly study-time goal is 1,260 minutes - 21 hours a week - with a deadline of 8 September to be consistently hitting it. Across the log so far that's 20,100 total minutes over 135 active days, an average session of 59 minutes, and an average mood rating of 3.60 out of 5 across every session logged. None of those numbers are dramatic. That's the point - a sustainable 21-hour week for three months beats a heroic 40-hour week that lasts nine days.
If your own weekly average is well under your target with ten or more weeks to the exam, the fix is smaller than it feels: a slightly earlier wake-up or one weekend session added, sustained daily, closes the gap faster than trying to make up an entire week's shortfall on a Sunday.
What 21 hours a week actually looks like in sessions
1,260 minutes divided by my own 59-minute average session works out to roughly 21 sessions a week - about three a day, which is closer to how the week actually feels than "21 hours" is. Some of those sessions are long: my longest single logged session ran 190 minutes, over three hours, which happens on mock days when the test itself plus the same-day first pass of analysis get logged together. Most are not - a 20-30 minute VARC reading session before bed counts just as much toward the weekly total as a three-hour Saturday block does, and treating the two as interchangeable is what makes 21 hours achievable on a weekday that only has an hour of real slack in it.
September-November: what actually changes
Mock frequency goes up, but so does the gap between them
The May-August stretch was roughly one full mock every 4-5 days, which is closer to a diagnostic cadence than an exam-simulation one. From September the plan is three full mocks a week, spaced out rather than bunched, because the August lesson holds here too: back-to-back mocks with no analysis window in between just produce more scores, not more learning.
Every mock gets a second look, not just a first one
The mocks that moved my score weren't the ones I took - they were the ones I reviewed twice. Karma Yogi's revision round system tags each analysis session R1 through R7 against the mock it's reviewing, on a spaced schedule (R1 at 3 days, R2 at 7, and so on out to R7 at 60). One of my own sectional notes makes the case better than an abstract argument does: a DILR sectional I scored 18/66 on carries an R2 note reading "could have gotten 39 marks - Q10, Gold Standard for Games and Tournaments." That's more than double the original score, found on a second pass, weeks after the first one felt final. See how I run that second pass end to end.
DILR and VARC get named as the volatile ones, because the data says so
Across all 23 mocks, my DILR score has ranged from 2 to 30 and VARC from 3 to 39. QUANT has been the comparatively stable section, rarely below 19 and regularly in the high 20s and 30s. That asymmetry decides where September's marginal hour goes: not evenly across three sections, but weighted toward the two with the widest spread, because that's where a bad day costs the most and a good technique buys the most back. If your own section spread looks similar, the DILR practice strategy and the VARC guide go into what specifically to drill.
What I'm not changing
I'm not switching mock providers this late, and I'm not adding a fourth subject. Twenty-three mocks in, changing the instrument makes the trend line useless right when it starts being informative. The exam is not moving; the plan for the next 13 weeks is more mocks, spaced out, each one reviewed at least twice, on the same three sections I've been tracking since May.
The mistake worth naming explicitly
The closest I came to a real mistake in this stretch wasn't a bad mock - it was almost changing my plan because of one. The 40 on 30 July, read on its own, looks like a reason to rebuild an entire section's approach. Read next to 96 the mock before and 94 the mock after, it looks like exactly one bad day, and the section breakdown backs that up: VARC, DILR, and QUANT all dropped together that day rather than one section cratering while the others held - which points at something that day (poor sleep, a rough start) rather than a gap in any specific topic. Reacting to a single mock instead of a run of them is the single most avoidable planning error I can point to in my own log, and it's exactly the failure mode the analysis framework is built to catch before it changes a plan for the wrong reason.
See how Karma Yogi tracks a CAT prep timeline like this one, or start tracking free.
End of essay
- Anish Guruvelli