Business Analytics Quiz: Data Types And Summaries
10 questions · exam conditions
0:00
Data Types And SummariesQuestion 1 of 10

A subscription company is building a descriptive dashboard from four fields: a 12-digit customer account number, a service tier labeled Bronze, Silver, or Gold, monthly spending, and a transaction timestamp. Monthly spending is strongly right-skewed.

Which treatment of the fields is most appropriate for the dashboard?

Treat account number as numeric, tier as nominal, spending as numeric, and timestamp as a categorical label.
Treat account number as an identifier, tier as ordered categorical, spending as numeric, and timestamp as a time-series index.
Treat account number as ordered categorical, tier as numeric, spending as time series, and timestamp as an identifier.
Treat account number as an identifier, tier as numeric, spending as categorical, and timestamp as a continuous outcome.
← Back to quizzes

Business Analytics Quiz

Business Analytics Quiz: Data Types And Summaries

Practice Data Types And Summaries in Business Analytics with focused quiz questions that help you check what you know, review explanations, and build confidence with test-style prompts.

What this quiz covers

This quiz focuses on Data Types And Summaries, giving you a quick way to practice the rules, question types, and explanations that matter most for Business Analytics.

How to use this quiz

Try each quiz question before looking at the correct answer. Use the explanations to review missed ideas, then come back to similar questions until the pattern feels familiar.

All questions

Question 1

A subscription company is building a descriptive dashboard from four fields: a 12-digit customer account number, a service tier labeled Bronze, Silver, or Gold, monthly spending, and a transaction timestamp. Monthly spending is strongly right-skewed.

Which treatment of the fields is most appropriate for the dashboard?

  1. Treat account number as numeric, tier as nominal, spending as numeric, and timestamp as a categorical label.
  2. Treat account number as an identifier, tier as ordered categorical, spending as numeric, and timestamp as a time-series index. (correct answer)
  3. Treat account number as ordered categorical, tier as numeric, spending as time series, and timestamp as an identifier.
  4. Treat account number as an identifier, tier as numeric, spending as categorical, and timestamp as a continuous outcome.
Explanation: When building a descriptive dashboard, your first task is classifying each field correctly before choosing any visualization or aggregation method. The key question is: what does each field represent, and what operations are meaningful to perform on it? A 12-digit account number looks numeric but carries no quantitative meaning — adding or averaging account numbers is nonsensical. It's purely an identifier. A service tier labeled Bronze, Silver, and Gold has a natural ranking (Gold > Silver > Bronze), making it ordered categorical (also called ordinal), not just nominal. Monthly spending is a true numeric measure, so it should stay numeric even though it's right-skewed — skewness is handled with transformations or median-based summaries, not by changing its type. Finally, a transaction timestamp represents a moment in time, which makes it a time-series index — the natural backbone for plotting trends over time. That reasoning maps exactly to answer B, which correctly identifies all four field types. A is wrong because it treats account number as numeric (meaningless arithmetic) and demotes timestamp to a mere categorical label, destroying its temporal structure. C assigns an ordering to account numbers (nonsensical) and treats spending as a time series (spending is a measure, not a time dimension) while calling the timestamp an identifier — nearly every field is misclassified. D repeats the identifier treatment for account number correctly but then calls spending categorical (you can't compute averages on categories) and labels timestamp a continuous outcome, which implies it's something you'd predict rather than a structural dimension. Study tip: On field-classification questions, always ask three things — does the field identify, measure, or describe order/category? Numbers that label (IDs, zip codes) are never truly numeric.

Question 2

A retailer records weekly order values, in thousands of dollars, for 11 weeks: 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, and 208. The final value resulted from a one-time promotion. Management wants summaries of a typical week and routine variability that are resistant to this unusual week.

Which pair of summaries best meets management's objective?

  1. A median of 2323 and an interquartile range of 66 (correct answer)
  2. A median of 2323 and a full range of 190190
  3. A mean of about 39.439.4 and an interquartile range of 66
  4. A mean of about 39.439.4 and a full range of 190190
Explanation: When a dataset contains an outlier, your choice of summary statistics determines whether that unusual value distorts your picture of "typical." This question tests whether you can match the right measures to the goal of resistance to outliers. Here, the value 208208 is a clear outlier — a one-time promotion far above the routine range of 1818–2727. Management wants summaries unaffected by it. The two resistant measures are the median and the interquartile range (IQR). With 11 values in order, the median is the 6th value: 2323. The IQR, measuring spread from the 25th to 75th percentile (roughly Q1 = 2020 to Q3 = 2626), equals 66. Neither calculation is pulled by the extreme value of 208208. That makes A the correct answer — it gives management an accurate, stable summary of routine weeks. Choice B pairs the correct median of 2323 with the full range (208−18=190208 - 18 = 190). The range uses only the minimum and maximum, so the outlier completely dominates it — this measure is not resistant and misrepresents routine variability. Choice C pairs the mean with the IQR. The mean of all 11 values is approximately 43311≈39.4\frac{433}{11} \approx 39.4, which is pulled far above the typical week by the 208208 outlier. The mean is non-resistant, so it fails management's objective even though the IQR is appropriate. Choice D combines both non-resistant measures — the inflated mean and the outlier-driven range — making it the worst option for this goal. Study tip: Whenever a question mentions outliers and asks for "typical" or "resistant" summaries, immediately pair median with IQR — they are always the resistant duo.

Question 3

A software company reports monthly churn rates of 4.8 percent in January 2024, 5.0 percent in February 2024, 5.4 percent in January 2025, and 5.1 percent in February 2025. Churn is known to vary by calendar month.

To assess whether the latest February result changed meaningfully while accounting for seasonality, which summary is most appropriate?

  1. Compare February 2025 with January 2025 and report a decline of 0.30.3 percentage point.
  2. Compare February 2025 with February 2024 and report an increase of 0.10.1 percentage point. (correct answer)
  3. Compare the two-month 2025 average with the two-month 2024 average and report an increase of 0.350.35 percentage point.
  4. Compare February 2025 with the four-month average and report an increase of 0.0250.025 percentage point.
Explanation: When data is known to vary by calendar month, comparing the same month across different years is the standard way to strip out seasonal noise and isolate true year-over-year change. This is the core concept being tested: proper like-for-like comparison in the presence of seasonality. That logic points directly to B. February 2025 (5.1%5.1\%) versus February 2024 (5.0%5.0\%) gives you an apples-to-apples read: both observations share the same seasonal forces, so the +0.1+0.1 percentage point difference reflects genuine change rather than calendar effects. Here's why the other options fall short. A compares February 2025 to January 2025 — two consecutive months in the same year. Since churn varies by calendar month, any difference you observe blends seasonal movement with underlying trend, making it impossible to tell whether the February result truly "changed meaningfully." The reported −0.3-0.3 pp decline could be entirely seasonal. C averages January and February across both years ((2025 avg: 5.25%5.25\%; 2024 avg: 4.9%4.9\%; difference: +0.35+0.35 pp)), which dilutes the February-specific signal you're trying to evaluate — you're no longer isolating the latest February result as the question requires. D compares February 2025 against the four-month grand average (5.075%)(5.075\%), mixing different months and different years into a single benchmark, which is neither seasonally controlled nor conceptually clean. Study tip: Whenever a question mentions that a metric "varies by calendar month," treat that as a direct signal to match months year-over-year. Mixing months — whether consecutively or through blended averages — is the trap the question is designed to set.

Question 4

In an A/B test, variant A receives 900 desktop visitors with 90 conversions and 100 mobile visitors with 30 conversions. Variant B receives 100 desktop visitors with 15 conversions and 900 mobile visitors with 99 conversions. The future channel mix has not yet been selected.

Which summary is most defensible before management chooses a variant?

  1. Use pooled conversion rates of 12.0%12.0\% for A and 11.4%11.4\% for B, then select A.
  2. Average the two channel rates equally for each variant, then select the variant with the larger average.
  3. Report channel-specific rates and calculate standardized overall rates using an agreed future channel mix. (correct answer)
  4. Compare total conversion counts of 120120 for A and 114114 for B, then select A.
Explanation: Whenever an A/B test involves Simpson's Paradox — where subgroup trends reverse at the aggregate level — you must think carefully about how you combine data across segments before drawing conclusions. Here, desktop converts at 10%10\% for A vs. 15%15\% for B, and mobile converts at 30%30\% for A vs. 11%11\% for B. Variant B actually wins on desktop, while A wins on mobile. The problem is that each variant was exposed to very different channel mixes: A saw mostly desktop traffic, B saw mostly mobile. Any aggregate number you compute will silently encode today's traffic mix — which hasn't been chosen yet for deployment. The only defensible approach is C: report the channel-specific rates transparently, then compute a standardized overall rate by weighting those channel rates according to the agreed future channel mix. That way management compares apples to apples and can revisit the numbers if their channel strategy changes. A falls into the Simpson's Paradox trap directly. Pooling all visitors regardless of channel gives A 1201000=12%\frac{120}{1000}=12\% and B 1141000=11.4%\frac{114}{1000}=11.4\%, but this result is an artifact of the lopsided sampling — not evidence of A's true superiority. B sounds more balanced, but averaging channel rates equally (desktop 50%, mobile 50%) imposes an arbitrary mix that may not reflect reality. It's marginally better than A but still sidesteps the need for an agreed future mix. D compares raw counts (120 vs. 114) while ignoring that both variants had identical total sample sizes of 1,000 — making this equivalent to answer A and equally misleading. Study tip: Whenever subgroups are unevenly distributed across variants, flag Simpson's Paradox as a risk. Always ask, "Does the overall rate reflect the mix I actually care about?"

Question 5

A company reviews 500 orders. Refund status is Yes for 40 orders, No for 410 orders, and missing for 50 orders. The missing values have not been verified as either refunded or not refunded.

Which summary most transparently describes the categorical refund variable without making an unsupported assumption?

  1. Report an 8.9%8.9\% refund rate among 450 known records and a 10%10\% missing rate. (correct answer)
  2. Report an 8.0%8.0\% refund rate among 500 records by treating missing records as not refunded.
  3. Report an 18.0%18.0\% refund rate among 500 records by treating missing records as refunded.
  4. Report an 8.9%8.9\% refund rate among known records and omit missingness from the dashboard.
Explanation: When summarizing categorical data with missing values, your job is to report what you actually know — without quietly burying uncertainty or inflating it into a false certainty. The key question to ask yourself: does this summary make an assumption about the missing records that hasn't been verified? Answer A is correct because it separates verified data from unverified data. Among the 450 known orders, 40450≈8.9%\frac{40}{450} \approx 8.9\% were refunded — that's a defensible calculation. Separately flagging the 50500=10%\frac{50}{500} = 10\% missing rate tells stakeholders exactly how much uncertainty exists, without pretending it doesn't. Answer B reports 40500=8.0%\frac{40}{500} = 8.0\% by treating the 50 missing records as "No refund." That's an assumption — those orders were never verified. Silently absorbing them into the denominator as non-refunds is a form of imputation without disclosure, which distorts the picture. Answer C goes the opposite direction, treating all 50 missing records as refunded: 40+50500=18.0%\frac{40+50}{500} = 18.0\%. This is equally unsupported and would dramatically overstate the refund rate — a critical error if someone used this figure to make financial decisions. Answer D shares the correct 8.9%8.9\% figure but omits the missing-data disclosure entirely. Transparency requires reporting both the rate and the missingness. Hiding the 10%10\% gap from the dashboard is exactly the kind of quiet omission that misleads decision-makers. The study tip here: whenever you see missing values in a data summary question, immediately check whether the answer choice acknowledges the missingness explicitly. Transparency means showing both what you know and what you don't.

Question 6

Three stores report annual sales and selling space. Store A has 1.2 million dollars in sales and 20,000 square feet; Store B has 0.9 million dollars and 10,000 square feet; Store C has 1.5 million dollars and 30,000 square feet. Management wants to compare space productivity and report a typical store without allowing the largest store to dominate.

Which summary best addresses both parts of management's request?

  1. Rank C, A, B by total sales and report median sales of 1.21.2 million dollars.
  2. Report portfolio sales per square foot of 6060 dollars and omit individual store rankings.
  3. Rank B, A, C by sales per square foot and report a mean rate of about 66.766.7 dollars.
  4. Rank B, A, C by sales per square foot and report a median rate of 6060 dollars. (correct answer)
Explanation: When a question asks you to compare productivity across stores and report a typical value that resists distortion from outliers, you need two things: the right metric (sales per square foot) and the right average (median, not mean). Start by calculating sales per square foot for each store. Store A: $1,200,00020,000=$60/sq ft\frac{\$1{,}200{,}000}{20{,}000} = \$60/\text{sq ft}. Store B: $900,00010,000=$90/sq ft\frac{\$900{,}000}{10{,}000} = \$90/\text{sq ft}. Store C: $1,500,00030,000=$50/sq ft\frac{\$1{,}500{,}000}{30{,}000} = \$50/\text{sq ft}. Ranked by space productivity: B ($90), A ($60), C ($50). The median of these three values is A's rate — $60\$60 per square foot — because it's the middle value when ordered. That's exactly what D reports: correct ranking by the right metric, and a median that prevents the extreme values from skewing the result. Answer A ranks stores by raw sales, which ignores how efficiently space is used — larger stores naturally generate more revenue, so this doesn't measure productivity. Answer B reports a blended portfolio rate (total sales ÷ total space = $3,600,00060,000=$60\frac{\$3{,}600{,}000}{60{,}000} = \$60) but omits individual rankings entirely, failing half of management's request. Answer C gets the ranking right but uses the mean — 90+60+503≈$66.7\frac{90+60+50}{3} \approx \$66.7 — which is pulled upward by Store B's high rate and doesn't represent a "typical" store as robustly as the median does. Your takeaway: whenever management wants a "typical" value that won't be dominated by extremes, default to the median. Pair it with the appropriate rate-based metric (here, sales per square foot) rather than raw totals.

Question 7

A manager must set a daily inventory level using ten equally relevant historical daily demand values: 20, 22, 24, 25, 26, 28, 30, 35, 50, and 100 units. Assume no trend or seasonality, and ignore ordering costs. The manager wants the smallest inventory level that would have met all demand on at least 80 percent of those days.

Which summary should drive this prescriptive decision, and what inventory level does it imply?

  1. Use the empirical 80th percentile and stock 3535 units. (correct answer)
  2. Use the arithmetic mean and stock 3636 units.
  3. Use the sample median and stock 2727 units.
  4. Use the observed maximum and stock 100100 units.
Explanation: When a manager needs to set an inventory level that satisfies demand on at least a certain percentage of days, the right tool is the empirical percentile — not a measure of central tendency. This is a prescriptive (decision-focused) problem asking: "What threshold covers 80% of observed demand values?" With ten equally weighted historical values sorted in ascending order — 20, 22, 24, 25, 26, 28, 30, 35, 50, 100 — the 80th percentile corresponds to the value at position 0.80×10=80.80 \times 10 = 8, which is 35 units. Stocking 35 units would have met demand on 8 out of 10 days (80%), satisfying the stated service-level criterion exactly. Answer A is correct. Answer B uses the arithmetic mean: (20+22+24+25+26+28+30+35+50+100)/10=360/10=36(20+22+24+25+26+28+30+35+50+100)/10 = 360/10 = 36. The mean is a measure of central tendency, not a service-level threshold. It's also distorted upward by the outlier of 100, making it a poor prescriptive tool here — it doesn't guarantee any specific coverage percentage. Answer C uses the median, which is the average of the 5th and 6th values: (26+28)/2=27(26+28)/2 = 27. The median only ensures 50% coverage by definition, falling far short of the 80% target. Answer D stocks the maximum (100 units), which guarantees 100% coverage but ignores the cost of overstocking — the problem explicitly says to minimize inventory while hitting the 80% threshold. Study tip: Whenever a question specifies a "service level" or asks for coverage on X% of days, immediately think percentiles — not means or medians.

Question 8

During a 30-day billing period, a business account has a balance of 100 thousand dollars for the first 10 days. The balance then changes to 160 thousand dollars and remains there for the next 20 days. No other balance changes occur.

If the company wants the average daily balance, which summary should it report?

  1. A simple average of the two observed balance levels, equal to 130130 thousand dollars
  2. The change between the two balance levels, equal to 6060 thousand dollars
  3. The ending balance for the billing period, equal to 160160 thousand dollars
  4. A time-weighted average of the balance levels, equal to 140140 thousand dollars (correct answer)
Explanation: When a balance stays at different levels for different lengths of time, a simple average ignores how long each level was actually in effect. The right tool here is a time-weighted average, which weights each balance by the fraction of the total period it occupied. Here, the 30-day period breaks into two segments: 10 days at $100K and 20 days at $160K. The time-weighted average is: (10×100)+(20×160)30=1000+320030=420030=140 thousand dollars\frac{(10 \times 100) + (20 \times 160)}{30} = \frac{1000 + 3200}{30} = \frac{4200}{30} = 140 \text{ thousand dollars} This is exactly what answer D reports — and it's the correct average daily balance because it reflects how much money was actually in the account each day across the full billing period. A is tempting but wrong: averaging the two balance levels as 100+1602=130\frac{100 + 160}{2} = 130 treats each level as equally important, ignoring that the $160K balance lasted twice as long as the $100K balance. This underestimates the true average. B reports the change in balance ($60K), which is a measure of movement, not a summary of the account's typical state. It doesn't answer "what was the average balance?" C reports only the ending balance ($160K), which describes where the account finished — not what it averaged over time. When you see "average daily balance" on the exam, always ask: did the balance stay the same the whole time? If not, weight each level by its duration — never just average the distinct values.

Question 9

One hundred customers rate service using the ordered categories Poor, Fair, Good, Very Good, and Excellent. The respective response counts are 8, 17, 46, 21, and 8.

Which dashboard summary best represents both the center and the distribution without assuming equal numeric distances between categories?

  1. Report a coded mean of 3.043.04 and the standard deviation of the assigned category numbers.
  2. Report the median category as Good and the percentage in each ordered response category. (correct answer)
  3. Report the modal category as Good and combine every other response into a not-Good category.
  4. Report the top-two-category percentage of 29%29\% as the sole measure of central response.
Explanation: When data uses ordered categories without numeric meaning — like satisfaction scales — you're dealing with an ordinal variable. The core rule: never treat the gaps between categories as equal distances, because "Poor to Fair" isn't necessarily the same psychological jump as "Good to Very Good." This rules out any arithmetic that assumes equal spacing. Choice B is correct because it respects the ordinal nature of the data on both fronts. The median is the appropriate center measure for ordinal data — it identifies the middle observation without assuming numeric distances. With 100 responses, the median falls at observation 50–51. Cumulating the counts: Poor + Fair = 25, plus Good's 46 brings you to 71, so the 50th and 51st observations both land in Good, confirming the median. Pairing this with the full percentage distribution (8%, 17%, 46%, 21%, 8%) preserves the shape of responses without distortion — students see where the mass concentrates across all categories. Choice A is a classic trap: assigning codes 1–5 and computing a mean (304100=3.04\frac{304}{100} = 3.04) treats the gaps between categories as numerically equal, which ordinal data does not support. Choice C reports the mode, which only captures the most frequent category and discards all distributional information — collapsing everything else into "not-Good" destroys the ordered structure entirely. Choice D's top-two-box percentage (Very Good + Excellent = 21+8=29%21+8=29\%) cherry-picks the upper tail and ignores the center and lower responses, giving an incomplete and potentially misleading picture. Study tip: On any question involving ordered-but-non-numeric categories, default to median + frequency distribution — and flag any answer that performs arithmetic (mean, standard deviation) on category labels as automatically suspect.

Question 10

A predictive customer-retention dataset contains five-digit postal codes stored as integers. Some codes begin with zero, several codes are rare, and missing postal codes occur in 4 percent of records. An analyst is conducting an initial data-quality summary before feature engineering.

Which summary best reflects the data type and the analyst's objective?

  1. Calculate the mean, standard deviation, and correlation of postal-code integers with retention.
  2. Create equal-width numeric intervals and summarize retention by low-to-high postal-code range.
  3. Report distinct-code count, category frequencies, rare-code prevalence, and the missing-code percentage. (correct answer)
  4. Order postal codes by record arrival time and summarize month-to-month changes in their average.
Explanation: Whenever you encounter a data-quality question, your first move should be to identify the data type. Postal codes are nominal categorical variables — they label geographic regions but carry no numeric meaning. The fact that they're stored as integers is a storage decision, not a statistical one. Treating them as numbers would be like averaging Social Security numbers. Given that framing, option C is the correct summary. Reporting the distinct-code count tells you how many unique categories exist, frequency distributions reveal which codes dominate, rare-code prevalence flags sparsity that may hurt model generalization, and the 4% missing rate is exactly the kind of data-quality metric an analyst needs before feature engineering. This summary respects the categorical nature of the variable and directly serves a pre-modeling audit. A is the classic trap here — it applies numeric statistics (mean, standard deviation, correlation) to a variable that has no meaningful numeric scale. A postal code of 90210 is not "larger" than 10001, so averaging them or correlating them with retention produces meaningless results. Leading zeros (e.g., 02134) being dropped from integer storage is a red flag that the field was mistyped. B creates equal-width numeric intervals, which again imposes an ordinal/numeric structure that doesn't exist. Geographic proximity doesn't follow the integer number line. D tracks temporal trends in postal-code averages — doubly wrong because it both performs numeric arithmetic on a categorical field and shifts focus from data quality to time-series analysis, which isn't the analyst's current objective. Study tip: When a variable looks numeric but represents a label or identity (ZIP codes, phone numbers, IDs), always default to categorical summaries — frequencies, counts, and missing rates — not arithmetic ones.