Read the row, column, scale and denominator first.
Practise row selection, scaled units, weighted means and pooled percentages. Fresh tables show why counts and rates can rank groups differently.
Each of six original items contributes one correct answer. The count describes this edition’s fictional quantities and assumptions, not a speed, ability or job-readiness band. All six worked answers and this full guide remain free at every count.
Write a data address before a sum
The six items provide small fictional tables in words. Locate the required row, period and measure before using a value. A whole-day total needs every included shift, while a question about a change needs two comparable periods. Copy the relevant figures with their row and column labels. This avoids accidentally combining a morning count from one day with an evening count from another, or subtracting costs from a quantity that was already a net figure.
Read the scale printed in the heading
If a fresh table is headed “hundreds of units” and lists 4.2 and 3.8, the sum is eight hundreds, or 800 individual units. The decimal values are not themselves individual items. Mark the scale once before calculating and apply it to the final requested unit. If the question asks for a percentage between values sharing the same scale, the common factor cancels; if it asks for an absolute count, that factor must remain. These are different output requirements.
Rebuild totals before averaging subgroup means
For a new example, ten records have mean 12 and thirty records have mean 20. Their reconstructed totals are 120 and 600, so the combined mean is 720 ÷ 40 = 18. Averaging 12 and 20 directly gives 16 and incorrectly gives the smaller group equal weight. This method assumes the groups are disjoint and the same measure is recorded in compatible units. If either condition is missing, do not invent a defensible overall mean from the available labels alone.
Pool numerators and denominators together
Suppose one fictional cohort has 120 records with 40% meeting a criterion, and another has 80 with 25%. Their matching counts are 48 and 20. Pooling gives 68 out of 200, or 34%, rather than the simple mean of the percentages. The word “disjoint” matters because overlapping records would be double-counted. Read whether the supplied percentages are exact or rounded as well; these checked examples use exact shares and do not infer hidden original counts.
A larger count can have a lower rate
An example line has four marked items among 80, a five-percent share. Another has six among 150, a four-percent share. The second line has more marked items but a lower marked rate. Neither summary is inherently wrong; they answer different questions. Keep both numerator and denominator beside a comparison, and state the requested unit. A rate difference does not by itself explain why the groups differ or prove that one operational method caused the difference.
Perform a whole-table reasonableness check
After calculating, compare your answer with the relevant totals. A combined mean of nonempty groups with positive weights lies between the subgroup means. A part-of-whole share cannot exceed the whole under the stated exclusive-count setup. A change in receipts-minus-costs should reflect changes in both components. If a result breaks such a relationship, revisit the selected cells and units before assuming the table is wrong. The page provides all six worked answers so that you can locate the earliest misread value.
A table exercise does not validate a real dataset
These fictional tables are fully specified to support a single answer in this practice bank. A real report can have missing rows, different definitions, rounded shares or delayed updates, none of which this count assesses. There is no timer, analyst certification, population percentile or employer pass threshold. Treat a missed answer as a specific selection, weighting or unit error to investigate. Retaking known tables may measure memory; use fresh compatible data and explain the complete calculation before making claims about improved independent analysis.