Relvora Prep

GMAT Data Insights lesson · Count Conditions

GMAT Data Insights: counting rows and bars that meet conditions

How to count on GMAT Table Analysis and Graphics Interpretation: strict vs inclusive thresholds, AND vs OR conditions, and conditions on derived values.

The core idea

Counting is a checklist, not a glance. Decide what happens at the threshold ('more than' excludes it), apply every condition to every row, and for derived conditions compute the value for each row first.

Core ideas

Write the condition down exactly ("> 21 doctors", "< 14% returns"), then go row by row. Count once, at the end.

Counting questions look easy, so people glance. Most wrong answers are off by one: a row at the threshold, a row that passed only one condition, a row skipped. A checklist fixes all three.

Write the conditionSettle the thresholdTick row by rowCount the ticks
Decide the edge case before you start ticking
  1. Write the condition in symbols: "more than 21" is > 21; "at least 21" is ≥ 21.
  2. Scan for any value exactly at the threshold and decide now whether it counts.
  3. Go down every row (or every bar) and tick the ones that pass every condition.
  4. Count the ticks. Check whether the question asks for those that pass or those that fail.

The words decide what happens at the threshold:

Wording Symbol Does 21 count?
more than 21, greater than 21, exceeds 21 > 21 no
at least 21, 21 or more, no fewer than 21 ≥ 21 yes
fewer than 21, less than 21 < 21 no
at most 21, no more than 21 ≤ 21 yes
more than 211516171819202122232425262721 is out
An open circle at 21: 'more than 21' starts just after 21

One threshold

Why it works. A threshold splits the rows into three groups: above it, below it and exactly at it. The wording only decides where the "exactly at" rows go. Settle that group first, and the rest of the count is a plain tally.

How to spot it: "The number of clinics with more than 21 doctors is ___".

Example. Doctors at seven clinics. How many clinics have more than 21 doctors?

Clinic Doctors
Ash 19
Birch 22
Cedar 21
Dale 9
Elm 12
Fir 24
Gorse 17
  1. The condition is > 21. Cedar has exactly 21, so it does not count.
  2. Tick the rows above 21: Birch (22) and Fir (24).
  3. Count: 2.
threshold 218101214161820222426
Two dots lie strictly to the right of 21

Trap: counting Cedar gives 3, the answer to "at least 21". Counting the clinics below 21 gives 4, the complement.

Try it. In the same table, how many clinics have at least 17 doctors?

Show the answer
  1. The condition is ≥ 17. Gorse has exactly 17, so it does count.
  2. Tick the rows: Ash (19), Birch (22), Cedar (21), Fir (24), Gorse (17).
  3. Count: 5. ("More than 17" would give 4.)

Edge cases.

  • "Between" needs a word about the ends. Between 12 and 21 inclusive counts Elm (12) and Cedar (21): 4 clinics. Strictly between leaves them out: 2.
  • The complement is a shortcut. Fewer than 21 = 7 − (at least 21) = 7 − 3 = 4.
  • No row at the threshold? Then "more than" and "at least" give the same count, and the wording doesn't matter.

GMAT-level example. Delays on six bus routes in March.

Route Delays
1 38
4 40
7 55
9 61
12 60
15 47

How many routes had between 40 and 60 delays, inclusive? How many would there be if "between" were strict?

  1. Sort the values: 38, 40, 47, 55, 60, 61.
  2. Two values sit exactly on the ends: 40 (Route 4) and 60 (Route 12).
  3. Inclusive: 40, 47, 55 and 60, so 4 routes.
  4. Strict: only 47 and 55, so 2 routes.
  5. Check: 4 inside + 1 below 40 (38) + 1 above 60 (61) = 6 routes ✓.
4060363840424446485052545658606264
Two dots sit exactly on the ends: inclusive counts them, strict leaves them out

Check yourself. In the clinic table, how many clinics have at most 21 doctors?

Check your answer

≤ 21 keeps Cedar (21): Ash, Cedar, Dale, Elm and Gorse, so 5.

Two conditions

Why it works. "And" asks for the overlap: a row counts only if it is inside both groups at once. Filtering on one condition first shrinks the list, so you test the second condition on fewer rows. And a row that passed only one condition is already gone, so it can't be counted by mistake.

How to spot it: "How many stores had more than 390 in 2024 sales and less than 14% returns?"

Example. Ten stores, 2024 sales (thousand units) and returns. How many meet both conditions?

Store Sales Returns Sales > 390? Returns < 14%? Both?
Alder 420 12% ✓ ✓ ✓
Birch 395 9% ✓ ✓ ✓
Cedar 510 13% ✓ ✓ ✓
Dover 390 10% ✗ (equal) ✓ ✗
Elm 455 11% ✓ ✓ ✓
Fern 402 16% ✓ ✗ ✗
Glen 360 15% ✗ ✗ ✗
Hale 470 8% ✓ ✓ ✓
Iris 600 13.5% ✓ ✓ ✓
Jade 300 20% ✗ ✗ ✗
  1. Filter on the first condition: 7 stores have sales above 390 (Dover's 390 is not above).
  2. Test those 7 on the second condition: Fern fails (16%).
  3. 6 stores pass both.
10 storesSales > 390Returns < 14%1 (Fern)1 (Dover)6neither2
Only the overlap counts for 'and': 6 stores

Trap: 8 stores pass at least one condition (6 + Fern + Dover). That 8 is the OR count, the tempting wrong answer. Counting Dover's 390 as "more than 390" gives 7.

Try it. In the same table, how many stores had sales above 400 and returns of at most 12%?

Show the answer
  1. Sales > 400: Alder, Cedar, Elm, Fern, Hale and Iris (Birch's 395 and Dover's 390 are out). That's 6 stores.
  2. Test those 6 on returns ≤ 12%: Alder (12%, exactly on the line, so it counts), Elm (11%), Hale (8%). Cedar, Fern and Iris fail.
  3. Count: 3. Reading "at most 12%" as "less than 12%" drops Alder and gives 2.

Edge cases.

  • OR = A + B − both. 7 stores pass the sales condition and 7 pass the returns condition. 6 pass both, so 7 + 7 − 6 = 8 pass at least one.
  • "Neither" is everything outside the OR: 10 − 8 = 2 stores (Glen and Jade).
  • "Not both" is not "neither". 10 − 6 = 4 stores fail at least one condition, but only 2 fail both.

GMAT-level example. Six laptop models.

Model Price ($) Battery (hours) Weight (kg)
K1 850 12 1.4
K2 920 11 1.2
K3 780 10 1.6
K4 899 9.5 1.3
K5 900 13 1.5
K6 640 10 1.5

How many models cost less than $900, have a battery life of at least 10 hours and weigh no more than 1.5 kg?

  1. Write the three conditions: price < 900, battery ≥ 10, weight ≤ 1.5.
  2. Price < 900: K1, K3, K4, K6. K5 costs exactly 900, so it is out.
  3. Battery ≥ 10 among those four: K1 (12), K3 (10), K6 (10). K4 (9.5) is out.
  4. Weight ≤ 1.5 among those three: K1 (1.4), K6 (1.5). K3 (1.6) is out.
  5. Count: 2 (K1 and K6).
  6. Check: every other model fails one named test. K2 on price, K3 on weight, K4 on battery, K5 on price. 2 pass + 4 fail = 6 ✓.

The traps are 3 (letting K5's $900 count as "less than 900") and 1 (reading "no more than 1.5 kg" as "under 1.5 kg", which drops K6).

Check yourself. In the store table, how many stores had sales above 390 and returns of 14% or more?

Check your answer

1 store: Fern (402, 16%). It's the "Sales > 390 only" region of the diagram.

Counting on a chart

Why it works. A combined condition is about a value the chart doesn't draw, like the total of the bars in a month. Each category gives one such value. Work it out for every category first, and the chart becomes a one-column table you can count like any other.

How to spot it: a chart with two series and "in how many months was the combined total more than 280".

Go category by category. For "the combined total was more than 280", add the series in each category first, and check whether any total equals 280 exactly.

Example. A shop's sales ($ thousand) by channel. In how many months were combined sales more than $280 thousand?

020406080100120140160180200Sales ($ thousand)JanFebMarAprMayOnlineIn-store140 + 140
Add the two bars in each month before comparing with 280
Jan120150270Feb140140280, not moreMar100190290 ✓Apr160130290 ✓May90170260
The same months stacked: only March and April pass 280
  1. Add each month: 270, 280, 290, 290, 260.
  2. Strictly more than 280: March and April.
  3. Count: 2.

Trap: February's total is exactly 280. Counting it gives 3. Reading one series only is the other slip: only one month has Online above 140, but that's a different question.

Try it. On the same chart, in how many months did In-store sales exceed Online sales by more than $20 thousand?

Show the answer
  1. Work out In-store − Online for each month: Jan 30, Feb 0, Mar 90, Apr −30, May 80.
  2. More than 20: January, March and May.
  3. Count: 3. April's −30 means Online led that month.

Edge cases.

  • "Each series" vs "combined". "Both channels sold more than 140" tests two bars separately; "combined sales above 280" tests their sum.
  • A bar ending exactly on a gridline equals that value. If the threshold is that gridline, the bar is not "more than" it.
  • "A exceeded B" is a condition on the difference A − B. Its sign decides, not how tall the bars are.

GMAT-level example. A company's units sold in three regions, in thousands. In how many months did at least two of the three regions sell more than 50 thousand units?

010203040506070Units sold (thousands)JanFebMarAprMayNorthSouthWest564860506242565258465054445240
The 50 gridline is the threshold; two bars end exactly on it
Month North South West Regions above 50
Jan 56 ✓ 42 50 (equal) 1
Feb 48 56 ✓ 54 ✓ 2
Mar 60 ✓ 52 ✓ 44 2
Apr 50 (equal) 58 ✓ 52 ✓ 2
May 62 ✓ 46 40 1
  1. This is a count inside a count. First, in each month, count the regions strictly above 50.
  2. January has 1 (West's 50 is not more than 50). April has 2 (North's 50 is out).
  3. Now count the months where that number is at least 2: February, March and April.
  4. The answer is 3 months.
  5. Check: the per-month counts 1 + 2 + 2 + 2 + 1 = 8 bars above 50 in all, and only the three middle months have 2 or more ✓.

The traps are 4 (counting the two 50s as passing, which lifts January to 2) and 8 (counting bars instead of months).

Check yourself. On the shop chart, in how many months did Online sales exceed In-store sales?

Check your answer

Only April (160 vs 130), so 1. February is a tie (140 and 140), so it doesn't count.

Conditions on derived values

Why it works. The condition is about a number that isn't printed, so there's nothing to tick until you make it. Computing the derived value for every row adds a new column. Then the question is a one-threshold count again.

How to spot it: the condition is about a value that isn't a column: growth, a rate, a difference.

Example. Members at five gyms. How many grew by more than 15% from 2024 to 2025?

Gym 2024 2025 Change Growth (÷ 2024) Wrong base (÷ 2025)
A 200 236 +36 18% ≈ 15.3%
B 400 464 +64 16% ≈ 13.8%
C 300 330 +30 10% ≈ 9.1%
D 250 290 +40 16% ≈ 13.8%
E 500 575 +75 15% ≈ 13.0%
  1. Compute (2025 − 2024) ÷ 2024 for every row.
  2. Compare with 15%, strictly: A, B and D pass; E is exactly 15%.
  3. Count: 3.
05101520Growth 2024–25 (%)ABCDE1816101615exactly 15
Three gyms are above the 15 gridline; E sits on it

Trap: dividing by 2025 instead understates growth and changes the count. With the wrong base only A passes, so you'd answer 1. Counting E (exactly 15%) gives 4.

Try it. In the same table, how many gyms added more than 35 members from 2024 to 2025?

Show the answer
  1. Use the Change column: +36, +64, +30, +40, +75.
  2. More than 35: A, B, D and E. C (+30) fails.
  3. Count: 4. E adds the most members but doesn't pass the 15% test, so "more members" and "faster growth" give different counts.

Edge cases.

  • Turn a percent test into a cutoff. Growth above 15% means 2025 > 1.15 × 2024. For Gym C the cutoff is 345, and 330 is below it, with no division needed.
  • A rate divides one column by another. "Loans per staff member" is loans ÷ staff, never staff ÷ loans.
  • Rounded displays can hide a tie. If a computed value shows as "15%", work it out exactly before you decide whether "more than 15%" passes.

GMAT-level example. Five public libraries.

Library Staff Loans 2024 Loans 2025
Ashton 8 11,000 12,400
Brook 12 16,000 18,000
Carver 5 8,000 8,800
Dunmore 10 14,000 16,100
Easton 5 7,200 7,800

How many libraries had more than 1,500 loans per staff member in 2025 and grew their loans by at least 10% from 2024 to 2025?

Library 2025 loans ÷ staff > 1,500? Growth (÷ 2024) ≥ 10%? Both?
Ashton 1,550 ✓ ≈ 12.7% ✓ ✓
Brook 1,500 ✗ (equal) 12.5% ✓ ✗
Carver 1,760 ✓ 10% ✓ (equal) ✓
Dunmore 1,610 ✓ 15% ✓ ✓
Easton 1,560 ✓ ≈ 8.3% ✗ ✗
  1. Make the first derived column: 2025 loans ÷ staff. Brook's is exactly 1,500, so it fails "more than 1,500".
  2. Make the second: (2025 − 2024) ÷ 2024. Carver's is exactly 10%, so it passes "at least 10%".
  3. Tick the rows that pass both: Ashton, Carver and Dunmore.
  4. Count: 3.
  5. Check with cutoffs: more than 1,500 per staff means loans above 1,500 × staff (Ashton 12,000, Carver 7,500, Dunmore 15,000; all beaten). At least 10% means 2025 ≥ 1.1 × 2024 (12,100, 8,800 and 15,400; all met) ✓.

The traps are 4 (counting Brook's 1,500), 2 (reading "at least 10%" as "more than 10%", which drops Carver) and 2 again by dividing growth by 2025, which puts Carver at about 9.1%.

Check yourself. A store's sales went from 80 to 92. Did they grow by more than 15%?

Check your answer

12 ÷ 80 = 0.15, exactly 15%. No: exactly 15% is not more than 15%.

How hard questions disguise it

For harder questions (L4–L5)

At L4 and L5, the count is the last step, and the hard part is everything before it. The condition may be a percent change or a ratio of two columns, so every row needs a calculation first. It may stack three conditions, use "at least" in one and "more than" in another, or put a value exactly on each threshold. On charts it may be a count inside a count ("months in which at least two regions…"). A sortable table helps: sort by the condition's column and the passing rows sit together. Write each condition with its symbol before you start, and tick row by row.

Practice

Practice counting rows and bars that meet conditions in timed sets on the counting conditions practice page.

Common traps

  • A value exactly at the threshold

    'More than 21' leaves out a row with exactly 21; 'at least 21' keeps it. Scan for the threshold value before counting.

  • AND read as OR, or one condition dropped

    A row counts for 'A and B' only if it passes both. Filter on one condition, then test the survivors on the other.

  • Miscounting or counting the wrong side

    Tick each row or category as you go and count the ticks; the complement (those that fail) is a classic wrong option.

  • Using the wrong base for a derived value

    Growth divides by the earlier year; a rate divides one column by another. Compute it for every row before counting.