CAT exam day: rules, calculator and scribble pad

CAT 2026 gives each candidate one scribble pad, one pen and an on-screen calculator for the whole 120 minutes, and lets nobody bring a calculator or rough paper of their own. Each of those rules changes how you work: the pad has to last three sections, the calculator is a choice to make question by question, and a problem on the day already has an official answer.

This lesson walks through test day from the night before to the end of Section III with the FAQs’ rule at each stage, measures how far ten quick estimates land from the exact values to show when the calculator is worth its time, and sets out what a technical fault and a bio-break each cost.

What to carry, and what stays behind

The rules for what goes into the test centre are short, and every one of them is in the CAT 2026 FAQs:

Two of these need work weeks before the day. The admit card is shown only on the CAT website: print it, affix the same photograph you uploaded, and follow the instructions printed on it. A medicine or another restricted item that you need during the test is allowed only with a permission letter from the exam authority. You ask for it by sending the Authorization FormAuthorization Form (CAT): The CAT form a candidate sends to the CAT Helpdesk, with supporting documents and the admit card, to ask permission to carry an essential medicine or another restricted item into the test. The approval is shown at the centre., with supporting documents and your admit card, to the CAT Helpdesk. The form names you, your application number, shift, city and centre, lists the items with the reason, and has a part for the CAT authorities, with a designation and a stamp, that authorises you to carry them. At the centre you show that permission in addition to the routine screening.

Walk through the day, stage by stage, with the rule at each one:

A PwBD candidate who uses a scribe carries the Scribe Affidavit from the CAT website, and the scribe brings a valid proof of identity and a passport-size photograph and signs a declaration at the venue; the FAQs say a scribe’s qualification should be one step below the candidate’s.

One pad for three sections, and a calculator on the screen

The FAQs give every candidate one scribble pad and one pen, with no extra pad for anyone, and say that the invigilators collect the pad at the end. PwBD candidates with blindness or low vision may ask for an additional pad of the standard size. So a single pad has to hold the rough work of all three sections, 120 minutes of it, or 160 with PwBD time. Three habits make it last: write small and from the top of each page down, put the question’s number beside each block of working so that a second pass can find it, and keep the first lines of the first page for your plan’s three clock readings in each section.

The calculator is on the screen throughout the test, but every number has to be entered and every step takes seconds. A rough estimate on the pad is often faster, and the question is when it is safe. The program estimates ten calculations of the kinds CAT’s sections ask for, works out the exact values, and measures the relative errorrelative error: How far an estimate is from the exact value, as a share of the exact value: |estimate − exact| ÷ exact. An estimate of 32 for 34.70 is 7.8 % off. of each:

"""Calculator or estimate? Ten calculations of the kind a CAT section asks for, each estimated by hand and checked.

Each row gives the calculation, a quick estimate as you would make it on the scribble pad, and the exact value. The
relative error is |estimate - exact| / exact. An estimate can choose between options if its error is under half the
distance from the right option to the nearest wrong one. Add your own rows and run it again.
"""
from fractions import Fraction as F
from math import isqrt

# (what is asked, how it is estimated, the estimate, the exact value)
ROWS = [
    ("47.8 % of 3,260", "48 % of 3,300", F(48, 100) * 3300, F(478, 1000) * 3260),
    ("1,284 / 37", "1,280 / 40", F(1280, 40), F(1284, 37)),
    ("growth from 2,450 to 2,780", "round both: 2,400 to 2,800", F(2800 - 2400, 2400), F(2780 - 2450, 2450)),
    ("growth from 4,980 to 5,420", "subtract first: 440 / 5,000", F(440, 5000), F(5420 - 4980, 4980)),
    ("3.14 x 6.4 x 6.4", "3 x 6 x 6, both rounded down", F(3 * 6 * 6), F(314, 100) * F(64, 10) ** 2),
    ("3.14 x 6.4 x 6.4", "3 x 6.5 x 6.5, one down, one up", 3 * F(65, 10) ** 2, F(314, 100) * F(64, 10) ** 2),
    ("7 x 8,950 / 13", "7 x 9,000 / 13", F(7 * 9000, 13), F(7 * 8950, 13)),
    ("1.08 cubed", "1 + 3 x 0.08", 1 + 3 * F(8, 100), F(108, 100) ** 3),
    ("37.5 % of 1,688", "3/8 of 1,688", F(3, 8) * 1688, F(375, 1000) * 1688),
    ("square root of 1,720", "square root of 1,600", F(isqrt(1600)), None),
]
EXACT_ROOT = 1720 ** 0.5  # the one value that is not a fraction: shown to 4 decimal places

print(" #  Calculation                 Estimated as                       Estimate       Exact   Error")
errors = []
for i, (asked, how, est, exact) in enumerate(ROWS, 1):
    exact_value = EXACT_ROOT if exact is None else float(exact)
    error = abs(float(est) - exact_value) / exact_value
    errors.append(error)
    if asked.startswith("growth"):  # a growth rate reads best as a percentage
        shown = f"{float(est) * 100:>9.2f} % {exact_value * 100:>9.2f} %"
    else:
        shown = f"{float(est):>11.6g} {exact_value:>11.6g}"
    print(f"{i:>2}  {asked:<27} {how:<34} {shown} {error * 100:>6.1f} %")

print()
for gap in (0.20, 0.10, 0.05, 0.02):
    safe = [i for i, e in enumerate(errors, 1) if e < gap / 2]
    print(f"Closest options {gap * 100:.0f} % apart: the estimate decides {len(safe):>2} of {len(ROWS)}"
          f" ({', '.join(map(str, safe)) or 'none'})")

 #  Calculation                 Estimated as                       Estimate       Exact   Error
 1  47.8 % of 3,260             48 % of 3,300                             1584     1558.28    1.7 %
 2  1,284 / 37                  1,280 / 40                                  32     34.7027    7.8 %
 3  growth from 2,450 to 2,780  round both: 2,400 to 2,800             16.67 %     13.47 %   23.7 %
 4  growth from 4,980 to 5,420  subtract first: 440 / 5,000             8.80 %      8.84 %    0.4 %
 5  3.14 x 6.4 x 6.4            3 x 6 x 6, both rounded down               108     128.614   16.0 %
 6  3.14 x 6.4 x 6.4            3 x 6.5 x 6.5, one down, one up         126.75     128.614    1.4 %
 7  7 x 8,950 / 13              7 x 9,000 / 13                         4846.15     4819.23    0.6 %
 8  1.08 cubed                  1 + 3 x 0.08                              1.24     1.25971    1.6 %
 9  37.5 % of 1,688             3/8 of 1,688                               633         633    0.0 %
10  square root of 1,720        square root of 1,600                        40     41.4729    3.6 %

Closest options 20 % apart: the estimate decides  8 of 10 (1, 2, 4, 6, 7, 8, 9, 10)
Closest options 10 % apart: the estimate decides  7 of 10 (1, 4, 6, 7, 8, 9, 10)
Closest options 5 % apart: the estimate decides  6 of 10 (1, 4, 6, 7, 8, 9)
Closest options 2 % apart: the estimate decides  3 of 10 (4, 7, 9)

An estimate settles a question with options if it is off by less than half the distance from the right option to its nearest neighbour. With options 20 % apart, eight of the ten estimates decide the answer; with options 2 % apart, only three do. Three estimating habits account for most of the difference:

• Subtract before you round. Rounding both ends of a growth rate first, row 3, lands 23.7 % away from the true 13.47 %. Row 4 subtracts first, so the difference, 440, stays exact and only the base is rounded: 0.4 % away.
• Round one number up and one down. Row 6 rounds 3.14 down to 3 and 6.4 up to 6.5, so the two errors pull in opposite directions and the product lands 1.4 % off; row 5 rounds every number down, and its errors add up to 16 %.
• Use fractions you know. 37.5 % is 3/8, so row 9 is exact.

So estimate when the closest options are far apart, and use the calculator when they are close. A typed answer has no options to separate, so it is computed exactly, then given the size check and the substitution of the typed-answer lesson.

A technical fault costs no time; a bio-break does

Two things can interrupt a section, and the FAQs treat them in opposite ways:

A fault, then, is something to report rather than to fight: the paused timer means the minutes spent moving are not yours to lose. A break is the opposite. The timer counts down the section you are in, so a break of 6 minutes during DILR comes out of DILR’s 40 minutes, 15 % of the section, and no later section can give it back. The FAQs advise candidates to plan their day so that a break is not needed unless it is absolutely necessary.

If a break cannot be avoided, rebuild the section’s plan from the clock when you return: the time left decides what still fits, and the check keeps its minutes.

Practise: the rules of the day

Ten untimed questions on what to carry, the pad and the calculator, and faults and breaks. Your weakest part comes first:

Interview questions

Warm-up: can you bring your own calculator to CAT, and what do you use instead?

No. The CAT 2026 FAQs say candidates are prohibited from bringing calculators to the test centre, and that an on-screen calculator is available throughout the exam. Rough work goes on the one scribble pad and pen the centre gives you, since your own rough paper and pens are not allowed either. The on-screen calculator is worth using when an answer must be exact, as a typed answer must, or when the closest options are too near for an estimate to separate. When the options are far apart, a careful estimate on the pad is usually faster: subtract before you round, round one number up and the other down, and use fractions you know, such as 3/8 for 37.5 %.

Key takeaways

• Bring the printed admit card with the same photograph you uploaded, your photo ID and any permission letter, which is asked for in advance with the Authorization Form; leave calculators and rough paper at home.
• One scribble pad and one pen last the whole test and are collected at the end, and writing anywhere else, the admit card included, is unfair means.
• The on-screen calculator is there throughout; estimate when the closest options are more than twice your estimate’s error apart, and compute typed answers exactly.
• A technical fault pauses the timer and costs no time, while a bio-break is allowed only in an emergency or for a medical condition, and its minutes come out of the section you are in.