Data Interpretation (DI) is not a new mathematical concept — it is the application of arithmetic you already know (percentages, ratios, averages) to a specific dataset presented in visual or tabular form. The question is not testing whether you know a formula; it is testing whether you can read a graph without losing time, extract only the numbers you need, and compute quickly without a calculator.
Think of it this way: a commanding officer is handed a logistics table at 0600 hours and needs to brief the unit in three minutes. She does not compute every cell — she scans for patterns, flags anomalies, and targets the specific figures her question demands. That is exactly the skill DI tests.
In CDS, DI questions appear in sets — typically 4 to 6 questions anchored to the same chart or table. This is both an opportunity and a trap. The opportunity: once you understand the source data, every question in that set costs less time. The trap: if you misread one number from the chart (a common error with closely spaced bar graphs or pie chart degree values), every answer in the set can be wrong.
The main formats you will encounter are:
value = (angle/360) × total.The core arithmetic skills that drive DI are: percentage calculation, ratio comparison, and average calculation. If any of those feel slow for you, fix them first — DI speed is downstream of those fundamentals.
Before you look at a single question in a DI set, spend 30-45 seconds surveying the chart. Note: What does each axis represent? What are the units (thousands? millions? per cent?)? Are there multiple charts linked together (like a three-chart pie problem)? Identifying these upfront prevents mid-solution rereads.
For a table-based set, your first move is to identify what the rows and columns represent and what the unit is. In CDS 2025, the newspaper circulation table had rows = newspapers (A–E) and columns = years (2019–2023), with values in thousands. Every question in that set reduces to one of three operations:
For "closest to average" questions, do not compute the difference for every row. Compute the average first, then visually bracket it: which values in the column are just above and just below the average? Those two are your candidates. Then compute the differences only for those two.
The master formula is:
If sectors are given as percentages:
When you have multiple linked pie charts (as in the CDS 2025 houses problem — three charts for houses, population, and families), draw a quick reference box:
| Category | Houses angle | Population angle | Families angle | |---|---|---|---| | Pucca | 110° | 60° | 70° | | Kutcha | 40° | 70° | 185° | | Semi-Pucca | 210° | 210° | 85° | | Houseless | — | 20° | 20° |
From this box, any question becomes a two-step lookup rather than a re-read of the chart.
Do not simplify this as a long division. Instead, note that 360 × (1/360) = 1, so you need 110 × (1\{,\}20\{,\}000 / 360). First compute 1\{,\}20\{,\}000 / 360 = 333.33. Then 110 × 333.33 ≈ 36\{,\}667. This two-step pattern works for every angle-to-value conversion — compute the "per-degree value" of the total once, then multiply by whatever angle you need.
These trip up many candidates. The setup: some units are added to one category, and you are asked how the angle in the pie chart changes. The steps are:
(original angle / 360) × original total.(new count / new total) × 360.Do not forget step 3 — the total changes too, not just the category count. This is the single most common error in pie chart update questions.
When a question asks "in how many cases does X exceed the overall average", the clean approach is:
In the CDS 2025 newspaper problem: grand total of all circulation values across 5 papers × 5 years = 378 (in thousands). Overall average = 378/25 = 15.12. Then you only need to check papers A and C — whose 5-year averages (20.8 and 16.0) are both above 15.12 — while B, D, and E fall below. Answer: 2.
Before solving any pie chart question, compute the "unit value per degree" once: divide the total by 360. Write it down. Every sector value is then a single multiplication. For the houses problem: 1,20,000 / 360 = 333.33 houses per degree. Pucca = 110 × 333.33 = 36,667. Kutcha = 40 × 333.33 = 13,333. You never do a long division again for that chart. Standard method: compute each fraction separately (3 divisions, ~20s each = 60s). This method: one division + multiplications (~25s total for all three).
For "which year is closest to the average" questions, first compute the mean, then immediately identify the two values it sits between in the list — one just above, one just below. Calculate only two differences, not five. In the newspaper D problem: average = 13. The series is 8, 12, 14, 15, 16. Average sits between 12 (2020) and 14 (2021). Check differences: |12-13| = 1, |14-13| = 1. Both tied — answer is both years. Standard method: compute 5 differences (~30s). Bracket method: spot the two neighbors + 2 subtractions (~10s).
When asked "how many individual averages exceed the overall average", do NOT compute individual averages first. Compute the grand total, get the overall average, then ask: which rows have a column-sum exceeding (overall average × number of years)? You compare sums, not averages. For the newspaper problem: overall average = 15.12, number of years = 5, so threshold sum = 75.6. Row sums: A=104, B=62, C=80, D=65, E=67. Only A (104) and C (80) exceed 75.6. Two subtractions replace five division-and-compare operations. Saves ~30s.
Stick this rule at the front of every "X more units added" pie question: the new angle uses the new total in the denominator, not the original total. Label the denominator explicitly before computing. In the Kutcha question: original total = 1,20,000, add 5,000 → new denominator = 1,25,000. Forgetting this gives angle ≈ 55° instead of 52.8°, a wrong answer. Writing the denominator takes 2 seconds; recomputing after a mistake costs 40 seconds.
For multi-chart DI sets (2 or 3 pie charts sharing the same categories), spend 20 seconds drawing a 4×3 grid with categories as rows and charts as columns, filling in the angles or percentages. Every subsequent question becomes a lookup rather than a re-read. Without the box, you re-scan the charts 2-3 times per question (adds ~15s per question across a 4-question set). With the box: zero re-scans. Net saving across the full set: ~60s.
When you open a DI set in the exam hall, follow this decision sequence:
Step 1 — Identify the chart type. Table? Single pie? Multiple linked pies? Bar or line graph?
Step 2 — Read the totals and units. Note units (thousands, millions, per cent) and all totals given. Write them down next to the chart.
Step 3 — For pie charts, compute the per-degree value immediately (total ÷ 360). For tables, scan for the largest and smallest values to calibrate your mental range.
Step 4 — Scan all questions in the set before solving any. Some questions feed into others. Compute shared values (like a grand average) once and use across questions.
Step 5 — For each question, classify the operation: Is it a direct value lookup? A ratio or percentage? An average comparison? A pie update? Apply the corresponding pattern from the tricks above.
Step 6 — Sanity check. Does the answer make physical sense? An average of persons per house between 1 and 30 is plausible; an answer of 950 is not.
Do not rework questions that took longer than 90 seconds. Mark and return. In a linked set, the next question may use the same intermediate value you already computed.
Why this question: The foundational operation of the entire newspaper set — computing the 5-year average for paper D — also underpins subsequent questions. Get this right first.
Solving path: Sum D's values across all five years: 8 + 12 + 14 + 15 + 16 = 65. Divide by 5: 65/5 = 13. Now bracket: the series contains 12 (2020) and 14 (2021) flanking the average of 13, each at distance 1. The next closest is 15 (2022) at distance 2. Answer: 2020 and 2021.
Why this question: This question requires computing a column average (average across all papers for one year) for each of the four years — more work, but the same underlying logic.
Solving path: Compute the yearly average for each year and compare it to D's value in that year. The official key gives 2021. Use the yearly totals and divide by 5 for each year, then find where the difference between that average and D's figure is smallest. Note: the explanation in the official key is accepted as the exam answer — always go with the official key in doubt.
Why this question: Introduces the "grand average vs. individual average" comparison — a pattern that recurs across CDS DI sets in multiple years.
Solving path: Grand total of all 25 cells = 104 + 62 + 80 + 65 + 67 = 378. Overall average = 378/25 = 15.12. Threshold sum for 5-year window: 15.12 × 5 = 75.6. Row sums: A = 104 (above), B = 62 (below), C = 80 (above), D = 65 (below), E = 67 (below). Two papers exceed the threshold.
Why this question: The classic linked-pie-chart question type. Tests whether you can pull the right angle from the right chart and apply the correct total.
Solving path: Pucca houses: (110/360) × 1,20,000 = 36,667. Population in pucca: (60/360) × 21,00,000 = 3,50,000. Average persons per pucca house: 3,50,000 / 36,667 ≈ 9.54.
Why this question: Tests the "update the pie chart" pattern — specifically whether you remember to update the total, not just the category count.
Solving path: Original Kutcha = (40/360) × 1,20,000 = 13,333. New Kutcha = 13,333 + 5,000 = 18,333. New total = 1,20,000 + 5,000 = 1,25,000. New angle = (18,333/1,25,000) × 360 = 52.8°. Change = 52.8 - 40 = 12.8° ≈ 13°.
Why this question: Combines the house count from Chart-I with the family count from Chart-III — a multi-chart lookup with a simple update.
Solving path: Kutcha houses (from Chart-I) = (40/360) × 1,20,000 = 13,333. Families currently in Kutcha (from Chart-III) = (185/360) × 3,60,000 = 1,85,000. After 300 houseless families shift: total Kutcha families = 1,85,300. Average = 1,85,300/13,333 ≈ 13.9.
Using the wrong total for pie chart computations. Each chart has its own total (houses, population, families). Applying the population total to a "number of houses" angle is a silent error — the computation looks clean but gives a nonsensical result. Always pair each angle with its own chart's total.
Forgetting to update the denominator in pie update questions. When new units are added, the grand total increases. The new angle must use the new total. Candidates who keep the old total in the denominator consistently get wrong answers on these questions.
Rounding intermediate values too aggressively. 1,20,000/360 = 333.33..., not 333. If you truncate to 333 and then multiply by 110, you get 36,630 instead of 36,667 — a 37-unit error that propagates. Keep at least one decimal in intermediate steps.
Treating "closest to average" as "equal to average." The question asks for the closest value, not an exact match. Average = 13, and neither 2020 (12) nor 2021 (14) equals 13, but both are closest. Candidates who scan only for an exact match miss the answer entirely.
In multi-row average comparison, computing all individual averages before checking the threshold. This is slower than the threshold-sum method and wastes 30-45 seconds. Compute the grand average once and convert it to a threshold sum.
Misidentifying which chart's angle to use in linked sets. In the three-chart houses problem, angle for "living in pucca" is from Chart-II (population, 60°), not Chart-I (houses, 110°) and not Chart-III (families, 70°). Confusing these three gives three different wrong answers. A reference box drawn before solving prevents this entirely.