Data Interpretation (DI) is not a math topic — it is a reading-under-pressure topic. The "data" part is just a vehicle. The actual test is whether you can extract the right numbers quickly, do simple arithmetic without errors, and avoid traps the question setter has laid out.
Here is the mental model that separates 90-percentilers from everyone else: treat every DI set like a newspaper table. You do not read every single number in a newspaper table — you scan for what the question asks. You look up, not across and down everything at once.
For UP Police Constable, DI questions almost always come from one of three formats:
The arithmetic required is almost never harder than SSC CGL. You will be doing percentages, averages, differences, and ratios. The difficulty is not calculation — it is misreading the graph and picking the wrong base.
Think of it this way: a DI set is like a closed-book exam where the "book" is right in front of you. You are not supposed to memorize the data. You are supposed to access it efficiently. The candidate who panics and reads all six bars before reading the question loses 40 seconds per question. You read the question, identify which year or category you need, look up only those values, calculate, and move.
A useful analogy: imagine you are a shopkeeper and someone asks, "How much did Branch C sell last year?" You do not re-read the entire ledger. You flip to last year, find Branch C, read the number. That is what DI expects from you.
The topics underneath DI — percentage change, simple average, ratio — you already know from Quantitative Aptitude. DI just wraps them inside a visual format to slow you down. Once you learn to unwrap the visual quickly, DI becomes your most reliable scoring topic.
Every DI set has a title, a Y-axis label, and (usually) a unit. These three things can change everything. Look —
Spend 15 seconds reading headers before you touch question 1. This is not wasted time — it saves you from choosing the wrong answer on every single question.
For a 4-5 question DI set, you will typically need:
Do not pre-calculate everything. Many candidates add all bars "just in case." That is a mistake. Wait for the question to tell you what to add.
This is where most marks are lost. The formula is:
The base is always the old value (previous year, or original value). For a decline, same formula — you get a negative result, and the question will say "percentage decline."
Common trap: the question asks "percentage increase in 2008 compared to 2001." Students sometimes use 2008 as the base. Don't. The reference year (2001) is always the denominator.
Formula is simple:
But the question often asks "in how many years was production more than average?" This requires you to:
Do not estimate here. Bars that visually "look close" to the average line will trip you up. Calculate the average first, then compare.
Some questions ask which pair of years has an average equal to a given pair. Here's the trick — instead of finding averages for all possible pairs, just find the target average once and scan for a pair that sums to exactly twice that average. Addition is faster than division.
For example, if the target average is 62.5, you need a pair summing to 125. Check each option by addition only.
When bars are not on round numbers (e.g., a bar falls between 140 and 160), use midpoints. If the gridlines are at 20-unit intervals and a bar appears exactly midway, read it as 150. If it is one-quarter above the lower gridline, add 5. Practice this visual interpolation — in the actual exam, you rarely get clean 100/200/300 values.
| Question Type | Approximate Calculation Time | |---|---| | Direct reading (one bar) | 5-10 seconds | | Difference between two bars | 10-15 seconds | | Combined total (2-3 values) | 15-20 seconds | | Percentage change | 25-35 seconds | | Average of all years | 35-50 seconds | | "How many years above average" | 50-70 seconds |
Tackle direct-reading and difference questions first within a set. Save average-based questions for last.
When comparing two bars visually, anchor on the smaller value and ask "how much gap is there?" rather than reading both bars independently and then subtracting. You read the gap directly by eye and verify with the scale. Standard method: read bar A (10s), read bar B (10s), subtract (10s) = 30 seconds total. Anchor method: read smaller bar (10s), read gap against gridline (10s) = 20 seconds total. Saves roughly 10 seconds per difference question, which across 4-5 questions in a set adds up to nearly a full question's worth of time.
For "years above average" questions, add all values first, divide by count to get the average, then use it as a filter. Do not calculate the average for each possible subset. Specifically: once you have the average = X, go through the list once more and simply tick values greater than X. This single-pass approach takes about 40 seconds total. The competing approach of trying to "eyeball" the average from the graph leads to errors on bars near the boundary — don't risk it. Standard guess-and-check method: 70-90 seconds. Sum-then-divide: 40-50 seconds.
When asked "which pair of years has the same average as years A and B," calculate the sum of A and B (not their average), then find the option pair whose sum matches. Dividing by 2 is an extra step you don't need. For example, if A = 70 and B = 55, sum = 125. Now check options: does 45 + 80 = 125? Yes. Done. This removes one division operation per check. Standard method (compute average then compare): 4 multiplications/divisions. Double-sum method: 4 additions only. Saves 15-20 seconds.
Convert percentage increase to a fraction immediately. If base = 30 and new = 85, increase = 55. Ask: what fraction is 55 of 30? That is roughly 55/30 = 11/6 ≈ 1.833. So 183.3%. This avoids multiplying by 100 early and then dealing with a large intermediate number. Specifically, express as a fraction, reduce if possible, then convert to decimal. Standard long-multiplication method: 45 seconds. Fraction shortcut: 20-25 seconds. Particularly useful when the denominator is a round number like 30, 40, 60.
In "which branch has highest combined sales" questions, eliminate obviously small branches before calculating. If one branch clearly has bars twice the height of all others, you only need to confirm its sum — you don't need to calculate every branch. In the spec PYQ: C2 has 150 + 200 = 350, while C1 = 150, C3 = 125, C4 = 196, C5 = 85. A quick visual scan shows C2 is the dominant bar in both years. You calculate C2 first (5 seconds) and stop. Standard full-calculation method: compute all 5 totals (60 seconds). Elimination method: 10-15 seconds total.
When you land on a DI set in the exam hall, follow this sequence without deviation:
Step 1 (15 sec): Read title, Y-axis label, unit, and legend. Note the unit — lakhs, thousands, crores — once, and don't re-read it.
Step 2 (10 sec): Scan question types across all 4-5 questions. Categorize each as: direct read / difference / percentage / average / pair-match.
Step 3: Solve in this order — direct reads first, then differences, then percentages, then averages. Average questions take the longest; don't start with them.
Step 4 (per question): Read the question, identify which years or categories it refers to, look up only those bars, calculate, match to options.
Step 5: If two options are very close (e.g., 183% vs 195%), go back and verify your reading — do not guess on close calls.
Decision rule for stuck questions: If a calculation takes more than 60 seconds and you are not close to an answer, mark and move. Come back. One DI question should not cost you three other questions.
Why this question: Tests your ability to visually compare two bars across multiple years and identify the maximum gap — a direct-reading question with a difference calculation.
Solving path: Read off bar values for both companies in each of the six years. Compute the absolute difference for each year. The year showing 110 vs 140 gives a difference of 30, which you verify is larger than all other year-gaps before confirming 2020. Do not stop at the first large gap — check all years before committing.
Why this question: Tests combined-total comparison across branches. The elimination method is the fastest approach here.
Solving path: C2 has 150 (2024) + 200 (2025) = 350 lakhs. A quick scan shows no other branch comes close — C4 = 196, C1 = 150, C3 = 125, C5 = 85. Confirm C2 = 350 is highest. Total calculation time: under 20 seconds using elimination.
Why this question: Classic "how many years above average" question. Tests whether you compute the average correctly before counting.
Solving path: Sum all 8 years: 30 + 45 + 65 + 50 + 70 + 55 + 80 + 85 = 480. Average = 480 ÷ 8 = 60. Now scan: which years have production > 60? Years with 65, 70, 80, 85 — that is 4 years. Answer: 4.
Why this question: Percentage increase from previous year — you must compute for each candidate year and compare, not just eyeball the steepest-looking bar.
Solving path: Check each option year. 2002: (45−30)/30 × 100 = 50%. 2003: (65−45)/45 × 100 ≈ 44.4%. 2005: (70−50)/50 × 100 = 40%. 2007: (80−55)/55 × 100 ≈ 45.5%. Maximum is 2002 at 50%. Always compute for all candidate years — the steepest-looking bar is not always the highest percentage increase.
Why this question: Percentage decline — tests whether you use the correct base (2003, not 2004) and whether your arithmetic lands on the right option.
Solving path: Production fell from 65 (2003) to 50 (2004). Decline = 15. Base = 65 (the old year). Percentage = 15/65 × 100 = 1500/65 ≈ 23.07% ≈ 23%. Among options 23%, 29%, 21%, 27% — match to 23%. Note: base is 2003, not 2004.
Why this question: Pair-average matching — a question type that rewards the double-sum shortcut over brute-force average computation.
Solving path: Average of 2005 and 2006 = (70 + 55)/2 = 62.5. Target sum for any matching pair = 125. Check options: 2002 + 2007 = 45 + 80 = 125. Match found. Stop — don't check the others unless you want to verify.
Why this question: Long-range percentage increase — large numbers, easy to mis-identify the base year.
Solving path: 2001 = 30 lakhs (base), 2008 = 85 lakhs (new value). Increase = 55. Fraction = 55/30 = 11/6. Convert: 11 ÷ 6 = 1.8333... → 183.33%. Match with option 183.33%.
Why this question: Pure direct-reading from a bar graph followed by a subtraction — the most basic DI question type, but tests accuracy of bar reading.
Solving path: Read the bar height for Department B, read the bar height for Department C, subtract the smaller from the larger. The difference is 250. No formula needed — only accurate visual reading.
Using the wrong base for percentage change. The base is always the earlier/original value, never the new value. Writing the formula on rough paper before substituting forces you to identify the base explicitly.
Misreading the Y-axis unit. If the Y-axis says "in lakhs" and the question asks for an answer in thousands, you must convert. Many candidates skip this and get an answer that is 10x or 100x off the mark.
Calculating averages for all option pairs instead of using the target sum. This wastes 30-40 seconds per question and increases arithmetic error risk.
Counting the average year itself as "above average." If production in a year equals exactly the average, that year is not "more than average" — strictly greater than means it does not count. Read the question word carefully: "more than" vs "at least."
Reading the bar for the wrong company in dual-bar graphs. When two bars are side by side for two companies, always confirm which colour/pattern corresponds to which entity before reading. Look back at the legend every time, not just the first time.
Stopping calculation early in maximum-percentage-increase questions. The bar that shows the largest absolute jump (biggest raw increase) is not necessarily the bar with the highest percentage increase, because the base values differ. Compute all candidate years; don't guess from the graph's steepness.