Skip to content
Tech Interview Prep

Top 100 Data Analyst Practice MCQs

Every question on one page: open one, choose an answer, and see why at once. Signed in, your latest answer to each question counts, and you can answer again after 24 hours.

Curated: · Written: · Reviewed:

Reviewed 100Review pending 0

Answers to reviewed questions count toward this role and the overall leaderboard; your latest answer to each one is the one that counts.

100 questions

MCQ-1An analyst is asked for March revenue. The `orders` table holds 5,000 rows for March, one per order, and `SUM(order_total)` over those rows is 1,250,000. The `order_lines` table holds 15,000 rows for the same orders: each March order has exactly 3 line rows. The analyst runs: ```sql SELECT SUM(o.order_total) FROM orders o JOIN order_lines l ON l.order_id = o.order_id WHERE o.order_month = '2024-03'; ``` What does this query return?
MCQ-14A finance dashboard query totals paid order revenue by customer segment. orders has one row per order (order_id, customer_id, amount, status); order_items has one row per item (order_id, sku, quantity); customers has one row per customer (customer_id, segment). The query is: WITH paid_orders AS ( SELECT order_id, customer_id, amount FROM orders WHERE status = 'paid' ), customer_totals AS ( SELECT po.customer_id, sum(po.amount) AS total FROM paid_orders po JOIN order_items oi ON oi.order_id = po.order_id GROUP BY po.customer_id ) SELECT c.segment, sum(ct.total) AS revenue FROM customer_totals ct JOIN customers c ON c.customer_id = ct.customer_id GROUP BY c.segment; The sum of orders.amount where status = 'paid' is 1,240,000, but the query returns 1,488,000. Which change should the team make?
MCQ-43A daily revenue report joins `orders` (12,000 rows, one per order, $440K) to `order_items` on order_id to add each order's total item discount, summed from the per-item discount column. The report now returns 38,400 rows and sums order_total to $1.42M. It must stay at one row per order and reconcile to $440K. Which explanation and fix are correct?
MCQ-94A weekly revenue dashboard sums `orders.total_amount` after joining `orders` to `order_items` on `order_id`. For the week of March 3 it reports $5.2M. Finance's ledger for the same week shows $2.6M. The `orders` table has 50,000 rows for that week; the join output has 100,000 rows covering 50,000 distinct `order_id` values. What is the most likely cause, and what is the corrected figure?
MCQ-2A weekly retention report runs `SELECT c.id, o.order_date FROM customers c LEFT JOIN orders o ON o.customer_id = c.id WHERE o.order_date >= '2026-01-01'`. The customers table holds 122,000 rows and the query returns 84,000. The report shows 0% of customers as inactive, though the analyst expected about 31%. An engineer moves the `o.order_date >= '2026-01-01'` predicate out of the WHERE clause and into the ON clause, then reruns the query. What does the rerun return?
MCQ-13A refunds table allows NULL customer_id for guest refunds. A query `SELECT id FROM customers WHERE id NOT IN (SELECT customer_id FROM refunds)` returns 0 rows. Which change is defensible?
MCQ-3A signup conversion rate is defined as the share of all sessions that end in a signup. The sessions table has one row per session and a nullable email column: 50,000 rows, 44,000 with an email, 31,500 distinct emails. The analyst's draft uses COUNT(email) as the denominator. Which denominator should the reviewer require?
MCQ-22An orders table holds 10 rows in last month's reporting window. Four have amount NULL because those orders are still being priced; the other six hold 120, 80, 150, 90, 60 and 100. The dashboard runs SELECT AVG(amount) FROM orders WHERE created_at >= ... and shows $100.00, while finance divides total revenue by the order count and gets $60.00. Which statement about the two figures is correct?
MCQ-4A nightly query groups 9 million orders by status and filters status = 'refunded' in HAVING. Only 210,000 rows are refunded, and the query takes 3 minutes. The result must not change. Which change holds up?
MCQ-5A monthly bonus goes to every rep whose `RANK() OVER (ORDER BY closed DESC)` is at most 3. The `deals` table holds: Ana 120, Ben 120, Cara 110, Dan 110, Eve 110, Fay 95. The query is `SELECT rep FROM (SELECT rep, RANK() OVER (ORDER BY closed DESC) AS rnk FROM deals) t WHERE rnk <= 3;`. Which reps receive the bonus?
MCQ-6A `sales` table holds one row per transaction: (Sep 1, 100), (Sep 2, 200), (Sep 2, 300), (Sep 3, 400). On PostgreSQL 16 this query runs: ```sql SELECT day, amount, SUM(amount) OVER (ORDER BY day) AS running FROM sales ORDER BY day; ``` What does it return for the two Sep 2 rows?
MCQ-25An analyst needs each order's share of its customer's total spend, counting only orders above $30, and runs this on PostgreSQL 16: ```sql SELECT order_id, 100.0 * amount / SUM(amount) OVER (PARTITION BY customer_id) AS pct FROM orders WHERE amount > 30; ``` The orders table holds (order_id, customer_id, amount): (1, A, 40), (2, A, 60), (3, B, 25). What does the query return?
MCQ-9An HR report must list every department with its second-highest distinct salary, showing NULL for a department with fewer than two distinct salaries. The `employees` table holds: | dept | salary | | --- | --- | | Sales | 90 | | Sales | 90 | | Sales | 70 | | Ops | 60 | | Tech | 150 | | Tech | 120 | | Tech | 120 | The analyst starts from: ```sql SELECT dept, salary FROM ( SELECT dept, salary, ROW_NUMBER() OVER (PARTITION BY dept ORDER BY salary DESC) AS rn FROM employees) t WHERE rn = 2; ``` Which change makes the query produce the required report?
MCQ-8An orders table has order_id (UUID), customer_id, updated_at, seq (a per-customer sequence that increases with each order), amount and item_count. A nightly job collapses it to one row per customer with: ```sql SELECT customer_id, MAX(updated_at) AS updated_at, MAX(amount) AS amount, MAX(item_count) AS item_count FROM orders GROUP BY customer_id; ``` Customer 88 has two orders: (2026-03-03 09:00, amount 40.00, item_count 3) and (2026-03-04 11:20, amount 25.00, item_count 1). The job's output row for customer 88 is (2026-03-04 11:20, 40.00, 3). The later timestamp carrying the earlier order's amount and item_count. 2,300 of 90,000 customers have an amount or item_count that never co-occurred with the reported updated_at, and 180 customers have two orders with the same updated_at. Which change makes the job emit one row per customer whose amount and item_count come from the same order as its updated_at?
MCQ-7A `monthly_revenue` table has one row per month that had sales: Jan 100,000, Feb 120,000, Apr 150,000, May 165,000. March has no row. A report computes `revenue / LAG(revenue) OVER (ORDER BY month) - 1` and prints April as 25.0%. Which diagnosis of that 25.0% is correct?
MCQ-10A retention dashboard reports a size-weighted month-6 retention across the last four weekly cohorts. The query left-joins activity to the cohort table and fills months with no activity using COALESCE(..., 0). Cohort sizes and month-6 retention: W1 2,000 users, 30%; W2 1,000 users, 24%; W3 1,500 users, 3 months old; W4 500 users, 2 months old. The dashboard shows 17%. What is the correct month-6 figure?
MCQ-15`logins(user_id, login_date)` holds one row per login event. User 42's rows are Sep 1, Sep 1, Sep 2, Sep 4, Sep 5, Sep 5, Sep 6 (2024). A dashboard runs this query (PostgreSQL 16) to report each user's longest streak: ```sql SELECT user_id, MAX(cnt) AS longest_streak FROM ( SELECT user_id, COUNT(*) AS cnt FROM ( SELECT user_id, login_date - (ROW_NUMBER() OVER (PARTITION BY user_id ORDER BY login_date))::int AS grp FROM logins ) x GROUP BY user_id, grp ) y GROUP BY user_id; ``` What does it return as `longest_streak` for user 42?
MCQ-17A `revenue(product, month, amount)` table holds one row per product per month it was live. Basic has Jan 12,000, Feb 13,500, Mar 14,100; Pro launched in March and has one row, Mar 9,000. An analyst pivots with `SUM(CASE WHEN month = 'Jan' THEN amount ELSE 0 END) AS jan`, the same for Feb and Mar, groups by product, and computes `(jan + feb + mar) / 3.0 AS avg_monthly`. Pro's `avg_monthly` comes out as 3,000. Which statement about that figure is correct?
MCQ-20A nightly job stacks two click tables with `SELECT user_id, event_ts FROM clicks_web UNION SELECT user_id, event_ts FROM clicks_app`. clicks_web holds 400,000 rows and clicks_app 300,000. Both tables carry a unique event_id, no event_id appears in both tables, and user_id and event_ts contain no NULLs. The query returns 640,000 rows, 60,000 short of the 700,000 the team expected. What is the correct diagnosis and fix?
MCQ-21A join between a CRM export and billing matched 58% of 50,000 customers; the expected match is about 99%. CRM stores the customer ID as text with trailing spaces and mixed case, billing as an integer. Leadership will act on the churn figure the join produces. Which step is defensible?
MCQ-19`events` is partitioned by day on `event_date`; each partition holds 6 GB and the table spans 365 days (2.19 TB). This query runs in 22 minutes and its plan reports 2.19 TB scanned: ```sql SELECT user_id, COUNT(*) FROM events WHERE created_at >= '2024-03-01' AND created_at < '2024-03-04' GROUP BY user_id; ``` `created_at` is a timestamp column; `event_date` is the partition key. Which change cuts the bytes scanned, and to what?
MCQ-12Daily revenue on the orders dashboard runs 6–9% below the finance ledger every day, while the monthly totals agree within 0.2%. The orders that disagree are the ones placed between 20:00 and midnight Eastern. Timestamps are stored in UTC and the dashboard groups orders by the date part of that timestamp. Finance's ledger is the source of record and closes its day in Eastern. Which change do you make?
MCQ-99A monthly churn report runs in a workbook: the analyst pastes a 1,200,000-row CSV export into a data tab, runs VLOOKUPs against a mapping tab, and copies the summary into a slide. This month the workbook reports 3.1% churn and the warehouse reports 4.4%. The data tab's last populated row is 1,048,576, and the export has 1,200,000 rows. The report is due again next month. Which action should the team take?
MCQ-81A sales-ops analyst owns a commission sheet. Rates live on a `Rates` sheet: column A is Rep ID, sorted ascending, column B is the per-unit rate. Rows 2–4 are 1044 → $0.031, 1046 → $0.028, 1047 → $0.026. The commission sheet has one row per rep-day; column A is Rep ID, column B is units sold, and column C is `=B2*VLOOKUP(A2,Rates!$A$2:$B$500,2)`. Rep 1045 left the company and was removed from `Rates`, but 214 rows for 1045 remain in the commission sheet, each with 1,000 units. Payroll has already run. What did column C show on those 214 rows?
MCQ-82A weekly revenue report is a pivot table on the Orders sheet, with its source range hard-coded as Orders!$A$1:$H$48000. In week 1 the sheet held 48,000 order rows and the pivot's grand total matched the source Amount SUM at $1.24M. By week 5 the sheet held 61,000 rows and the source SUM had risen to $1.58M, but the pivot's grand total was still $1.24M. Which change fixes the report?
MCQ-83A reviewer is checking a notebook that merges a 2M-row sessions table to a purchases table on user_id. Both tables have many rows per user, and the merged frame came out at 37M rows with spend per session 18× too high. The reviewer wants the merge to fail loudly at run time rather than ship the inflated totals. Which requirement should the reviewer accept?
MCQ-84A nightly job builds a `region_code` column for 8,000,000 rows. The current line is `df['region_code'] = df.apply(lambda r: REGION_CODES.get(r['region'], 0), axis=1)` and takes 11 minutes. `REGION_CODES` maps 40 region strings to ints, and 1,204 rows carry a region that is not in the dict. A teammate rewrites the line as `df['region_code'] = df['region'].map(REGION_CODES)`, which runs in 0.6 s; the two versions agree on a 2,000-row sample. What should you do?
MCQ-85A weekly business review has to answer one question: which of 9 marketing channels brought the most signups last quarter? The analyst brings a stacked bar chart, with 9 channels stacked inside each of 13 weekly bars, and the top two channels finish the quarter at 22% and 21% of total signups. Reviewers cannot agree which of the two is larger. Which change makes the required comparison decidable?
MCQ-86An analyst plots conversion rate for two checkout variants as a bar chart. Variant A converts at 4.02%, variant B at 4.11%. The y-axis starts at 4.0%, so B's bar is about 5.5 times the height of A's. The analyst writes 'B is a clear winner.' Which assessment is correct?
MCQ-93A retail analytics team must publish a per-domain customer count for a marketing review, and leadership wants the total row included. The warehouse query returns: | domain | customers | | --- | --- | | gmail.com | 8,412 | | yahoo.com | 1,203 | | acme.com | 7 | | total | 9,622 | The privacy policy requires that no published cell covers fewer than 10 customers and that no suppressed cell can be recovered by subtraction from the published table. Which table may be published?
MCQ-16One pilot day in the orders table holds ten order values, in dollars: 12, 18, 25, 30, 44, 60, 85, 140, 320, 900. Marketing proposes a free-shipping threshold at the mean, 163.40, describing it as 'what a typical customer spends'. The analyst wants the figure that leaves half of the day's orders below it and half above. Which reply should go back?
MCQ-38A streaming team must forecast next month's bandwidth cost, which is proportional to total minutes streamed. Last month there were 4 million sessions: mean session length 14 minutes, median 3 minutes, 90th percentile 11 minutes, and 2% of sessions ran over two hours. Which statistic should drive the forecast?
MCQ-39An analyst is finalizing Q3 average order value (AOV) for a marketplace. The BI dashboard shows $6,003,400 across 50,000 orders. Three records stand out: order 88 for $1,000,000, placed by an internal QA account; order 412 for $98,000, from a customer with a signed enterprise contract; and order 97 for $5,400, whose unit price field reads 5400 where the product's list price is $54. The other 49,997 orders total $4,900,000. Which AOV should the analyst report?
MCQ-64A pricing-page test assigns 4,000 visitors to each arm. Revenue per visitor has a median of $0, a mean of $9.40 in control and $10.25 in treatment, and the 20 largest purchases make up 55% of total revenue. A Welch t-test on the difference in means gives p = 0.036; a bootstrap with 10,000 resamples gives a 95% interval of [-$0.30, $2.10]. Which conclusion should the analyst report?
MCQ-67A team estimates mean monthly revenue per account from a random sample of 50 accounts: mean $120, standard deviation $500. One account bills $4,200; every other account bills under $300. Their 95% normal-approximation interval is −$19 to $259. Which conclusion is correct?
MCQ-68A marketplace team estimates median order value from 50,000 orders placed by 2,000 customers. Bootstrapping the orders gives a 95% interval of $41.80–$43.20 (width $1.40); bootstrapping the customers gives $40.10–$44.30 (width $4.20). Which interval belongs in the report?
MCQ-66A weekly report compares average order value for two campaigns. Each campaign has 40,000 orders. Mean A = $52.00, mean B = $55.50. The standard deviation of order value is $60 in both. The chart uses error bars of ±1 standard deviation, so the bars overlap heavily and the analyst writes that the $3.50 difference is noise. What should the analyst report instead?
MCQ-71A checkout service is sized from a plan that assumed hourly order counts follow a Poisson distribution with mean 40. Over 30 days of production data the hourly counts have a sample mean of 40.2 and a sample variance of 180, and capacity was exceeded on 9 of the 30 days against the roughly 1 day the plan predicted. Which conclusion is best supported?
MCQ-70A payments team runs a fraud model over 1,000,000 orders a day. 0.1% of orders are fraudulent. The model catches 99% of fraudulent orders and flags 1% of legitimate ones. The team sizes its manual review queue to the number of alerts. Which statement about the queue is correct?
MCQ-65A team validates its experiment pipeline with 40 A/A tests - the same variant served on both arms of 40 independent metrics - each run at alpha = 0.05 with no multiplicity adjustment. Three of the 40 come back significant (p < 0.05), and the metrics lead asks what the count says about the pipeline. Which reading is right?
MCQ-51A team tests 100 product changes this quarter. About 10% of tested changes are real improvements; the rest do nothing. Their A/B tests have 80% power at the 0.05 significance level. One change comes back at p = 0.04. Of the changes that come back significant, about what fraction are real improvements?
MCQ-55A team ran an A/B test. They checked the dashboard daily and stopped the test on day 12 when the p-value first fell below 0.05. In the review, they need to justify the early stop. Which justification is sound?
MCQ-56A team A/B tested a checkout redesign against 24 independent metrics, each at α = 0.05. Only 'time on page' came back significant, at p = 0.03, and the team shipped on that basis. A quarter later the effect was gone. What does the arithmetic say about that single result?
MCQ-53A checkout change shows a +0.2% conversion lift (p < 0.001) across 40 million monthly sessions. Shipping it requires a new pricing service at $90,000/year; the extra margin from the lift is $25,000/year. Which method is defensible for the ship/no-ship call?
MCQ-52A two-week test of a new pricing page moved revenue per user by +3.2%, with a 95% confidence interval of [+0.4%, +6.0%]. The change costs 1.0% of revenue per user to maintain, and the team ships a change only when it clears that cost. The team must decide this week whether to keep the change. What should they do?
MCQ-54A checkout A/B test has a 4.0% baseline conversion. The team wants 80% power to detect a 15% relative lift (4.0% to 4.6%) with a two-sided test at α = 0.05. The site sends 4,000 users per day, split evenly between the two arms. How long must the test run to reach the required sample?
MCQ-57A team tests a new pricing banner by randomizing at the session level, 50/50, over two weeks. The metric is revenue per user, and 38% of users saw both the banner and the old page. The treatment arm shows +3.1% revenue per user (p = 0.04). What should the team conclude?
MCQ-58A checkout A/B test planned a 50/50 split of 200,000 users. The experiment log shows 103,000 in control and 97,000 in treatment, and the treatment arm's conversion rate comes out 4% higher. The team is about to ship the variant. What is the right next step?
MCQ-59A 12-week A/B test of a redesign shows a topline engagement lift of +9% in week 1, +4% in week 2, +1% in week 3, and -1% in week 12, and the effect is still declining at week 12. The redesign will be rolled out to the existing user base, which is 96% of traffic. Which analysis should the reviewer accept?
MCQ-60A B2B SaaS team ships Bulk Export. Of 12,000 accounts active in Q1, the 1,800 that used Bulk Export at least once retained at 71% over the next two quarters, against 46% for the other 10,200: a 25-point gap. The team wants 'Bulk Export drives retention' in the board deck, and the analyst has time for one analysis before it ships. Which analysis decides whether the claim holds?
MCQ-61A retention report tells leadership that onboarding drives retention: 90% of the customers still active at one year had completed onboarding. The report's sample is customers still active at the one-year mark. Leadership plans to fund an onboarding program on that basis. Which analysis should the decision rest on?
MCQ-63An analyst reports a 3% order lift from a delivery-policy launch, computed as the treated region's 12% before-after rise minus a control region's 9% rise. The reviewer asks what evidence supports using the control region's change as the counterfactual. Which justification is sound?
MCQ-62A team fits a linear regression of weekly sales in units on price in dollars and advertising in thousands of dollars: sales = 900 - 110 * price + 25 * advertising + 8 * price * advertising. The fitted data cover prices from 8 to 12 dollars and advertising from 5 to 20 thousand dollars. The team asks for the predicted change in weekly sales from raising price from 10 to 11 dollars while advertising stays at 10 thousand dollars. Which value is the correct predicted change?
MCQ-73A grocery chain's 200 stores show a raw regression slope of +$48,000 in annual sales per additional staff member. Store square footage correlates with both staffing and sales, and within stores of similar size the slope is +$9,500. A regional VP plans to add 2 staff to each of 40 stores and asks what sales gain to expect. What should the analyst report?
MCQ-30A subscription product's overall monthly churn rose from 3.0% to 3.6% quarter over quarter. Both segments improved: self-serve went from 6.0% to 5.0%, enterprise from 2.0% to 1.5%. Self-serve's share of the base grew from 25% to 60%. Leadership wants one number for the board deck that explains the topline move. Which analysis should the team run?
MCQ-69A retailer coached its 20 lowest-selling stores for a quarter. Their sales rose 15% the next quarter, and leadership wants to roll the program out to all 400 stores. Which analysis is defensible?
MCQ-72A retention campaign applies a $10 credit to every targeted customer's account when the campaign starts, before any renewal decision. A pilot targeted 20,000 predicted churners; 2% changed their decision versus a held-out control group; each saved customer is worth $240. The team recommended scaling to all 400,000 predicted churners, and a retrospective found that recommendation wrong. Which change to the analysis addresses the cause?
MCQ-45Leadership will act on an analysis that depends on analyzing a price change. Which method is defensible?
MCQ-46A retailer emails a promotion to 900,000 of its 1,000,000 eligible customers, choosing the 100,000 who receive nothing by random draw. Over the 14-day window, the emailed group spends $52.00 per customer and the holdout spends $50.00 per customer. A separate dashboard compares the emailed group with customers who were excluded from the campaign because they had lapsed for over a year; those customers spent $44.00 per customer. Which figure is the campaign's incremental revenue?
MCQ-48Daily signups for a consumer app run 30-40% lower on Sundays than on weekdays; apart from that weekly cycle the metric is stable, with weekday-to-weekday swings of a few percent. A rule that alerted whenever today was more than 10% below yesterday fired 23 times last quarter, 19 of them on Sundays. The team muted it, and three weeks later missed a Tuesday when Android signups (40% of the total) fell 90% and the topline fell 36%. Which change to the alert rule addresses the cause?
MCQ-47A team forecasts weekly signups for the coming quarter, and the number feeds a capacity plan. Backtesting over the last 12 weeks, a gradient-boosted model has a mean absolute error of 3.1% and a seasonal naive baseline has 3.4%. The model's stated 90% prediction interval contained the actual value in only 7 of those 12 weeks. One forecast ships Monday. What should the team ship?
MCQ-49An analyst sizing the daily coffee market for a city of 8 million people builds a top-down estimate: 8M population × 50% who drink coffee × 1.5 cups per drinker per day × 30% of cups bought out rather than made at home = 1.8M cups sold per day. A bottom-up estimate from 2,000 cafes at 300 cups each per day gives 600k cups sold per day. The two differ by 3×. What should the analyst do next?
MCQ-50You are finishing a 4-hour take-home for a payments company. The prompt asks whether to launch a new checkout flow to all users, and the company's launch decision is based on external-user conversion. Your query returns 12.0% conversion for the new flow and 10.0% for the old flow. You then notice that 10% of the new-flow rows are internal test accounts, and those accounts convert at 40%. The reviewer's rubric allots 5 minutes to find a recommendation, the decision metric, and one query that reproduces the headline number. You have 45 minutes left. What should you do?
MCQ-40A retailer's December 2026 revenue was $1.40M, up 40% from November 2026's $1.00M, and the team called it a record month. December 2025 revenue was $1.49M. December 2026 had 27 selling days and November 2026 had 25, against December 2025's 26. The board asks whether the business actually grew, adjusted for the different number of selling days. Which comparison answers the question?
MCQ-41An analyst scores 27,000 customers on recency, frequency, and monetary value in quintiles, producing 125 RFM cells. The cell sizes are: 92 cells have fewer than 50 customers (4,000 customers total), 21 cells have 50–499 customers (8,000 total), 10 cells have 500–999 customers (9,000 total), and 2 cells have 1,000–3,100 customers (6,000 total). Marketing can run at most six campaigns next quarter, and each segment must contain at least 1,000 customers. The analyst must keep all three RFM dimensions because marketing uses spend to prioritise offers, and the segments must be explainable and reproducible from the existing quintile scores. Which deliverable should the analyst produce?
MCQ-42A weekly channel report credits branded search with 45% of sales under last-click attribution, but doubling branded search budget last quarter left total sales flat. The report sets next quarter's budget split, and the team can run one experiment per quarter. Which change should the team make to the report?
MCQ-44A grocery delivery service charges a $5.40 delivery fee. Orders under $25 average $18.00 in revenue (fee included) against $23.16 in variable cost: product $12.60, picking $3.10, delivery $5.40, payment fee $0.82, packaging $0.70, refunds $0.54. Orders of $25–$60 contribute $9.10 each and orders over $60 contribute $25.80 each. The growth team proposes waiving the delivery fee on orders under $25 to lift that band's order count 30%. Which conclusion does the unit economics support?
MCQ-33A subscription product bills $50 per customer per month at a 40% gross margin. Finance's LTV model applies a flat 4% monthly churn, giving an LTV of $500. A cohort acquired 60 months ago has produced $310 of cumulative gross margin per acquired customer, and only 2% of it is still active. What LTV should the analyst report?
MCQ-34A SaaS company spent $300,000 on paid acquisition in Q3 and added 1,500 new customers, 600 of whom signed up organically. Monthly gross margin per customer is $30 and monthly revenue per customer is $100. The board slide reports CAC of $200 and payback of 2 months. Which statement about the slide is correct?
MCQ-35A subscription team's north star is weekly orders delivered. Its guardrails are refund rate (max 3%) and margin per order (min $3.00). They ship a discount change and read the weekly result below. | Metric | Control | Variant | Guardrail | | --- | --- | --- | --- | | weekly orders | 100,000 | 106,000 | north star | | refund rate | 2.0% | 7.0% | max 3% | | margin per order | $4.10 | -$0.60 | min $3.00 | Which conclusion does the result support?
MCQ-36You run growth analytics at a B2B SaaS company. Revenue is recognised at contract signature and the sales cycle runs five to nine months, so a deal booked this quarter was sourced up to three quarters ago. You have two years of weekly history, twelve candidate indicators, and three weeks to put three of them on the weekly exec dashboard. Which method do you use to choose them?
MCQ-37A subscription app's growth dashboard shows cumulative signups rising from 380,000 to 440,000 to 500,000 across three quarters, and leadership plans to double paid acquisition spend on that basis. Over the same quarters monthly active users fell from 60,000 to 52,000 to 41,000. What should the analyst present?
MCQ-32A subscription product had 10,000 active subscribers on September 1. During September, 600 of those subscribers cancelled. A promotion added 3,000 new subscribers during September, and 450 of those cancelled before October 1. The dashboard reports September churn as 8.8%. The team wants to know whether retention of the base that was already there on September 1 got worse. Which figure measures that?
MCQ-29A checkout experiment moved conversion from 4.0% to 5.0%. The analyst's slide reads "conversion up 1%, so expect about 1% more orders." Finance is sizing next quarter's inventory from that line. Which correction should the analyst make?
MCQ-31A retail chain's weekly dashboard reports a company conversion rate of 11.7%, the mean of its three stores' rates. The stores logged 10,000, 2,000 and 500 sessions and converted 500, 200 and 100 of them. The COO asks for the company's true conversion rate. Which figure should be reported?
MCQ-11A growth team's signup-to-purchase funnel query returns 64%. The query counts distinct users with any purchase event in the 7-day window divided by distinct users with a signup event in the window, with no requirement that the purchase follow the signup. For the same window: 50,000 users visited, 6,000 users signed up, 1,260 of those signups purchased within 7 days of their own signup, and 2,580 users purchased but had signed up before the window opened. What should the reviewer conclude?
MCQ-18A consumer app's March report shows stickiness (DAU/MAU) of 0.31. The query defines an active user as anyone who opens the app, averages DAU over the 31 days of March, and counts MAU as distinct users who opened the app at least once in March. A review finds that 90,000 of the 200,000 monthly users opened the app exactly once, all on the day of a push campaign, and never returned; average DAU under that definition was 62,000. Recomputing with 'active' meaning a completed core action gives 36,000 average daily actives and 110,000 distinct monthly users. What is the March stickiness under the core-action definition?
MCQ-23An analyst explored a 4 TB events table, partitioned by day and spanning three years, by sampling to estimate per-user funnel counts, then reported the numbers. Asked to justify the sampling method, which justification is sound?
MCQ-24An orders dashboard compares yesterday with the same weekday a week earlier. Yesterday stands at 8,200 orders; the report runs 8 hours after the day closes. Last week's same weekday finished at 10,500, so the dashboard shows a 22% drop and the growth team is about to pause a campaign. Measured completeness of a day's orders by age: 8 hours 78%, 1 day 94%, 3 days 99.5%, 7 days 100%. Which reading does the data support?
MCQ-26You own the weekly growth report. The VP of Finance asks for 'active customers' for a board slide titled 'Paid accounts up for renewal next quarter'; finance will tie the number to subscription revenue. Your warehouse returns three counts for last quarter: 131,000 customers logged in at least once, 74,000 placed at least one order, and 48,000 held a paid subscription on the last day of the quarter. What should go on the slide?
MCQ-27Two analysts reported the same week's conversion rate as 2.9% and 3.4% (one excluded internal traffic, the other did not) and the definition has changed twice this quarter with no record of when or why. Finance reads the number without reading SQL. Which definition should the reviewer accept?
MCQ-28Orders are down 15% week over week and the hiring manager asks how you would diagnose it. Which approach holds up?
MCQ-95Checkout conversion on the weekly dashboard reads 4.8% this week against 3.0% last week, a 60% lift, and the growth team wants to announce it. The analyst pulls the underlying counts: sessions fell from 100,000 to 63,000, while orders in the payment system went from 3,000 to 3,030. Which conclusion is supported?
MCQ-98A checkout dashboard reports 30,000 `purchase_completed` events for last week. The payment provider's settlement report shows 24,000 successful charges for the same period. The web client sends `purchase_completed` when the user clicks Pay, and about a third of transactions finish through a redirect to the provider's 3-D Secure page. Which change makes the dashboard match the settlement report?
MCQ-92A payments team's dashboard is 38 tiles. Its refund tile shows 2.1%, the cumulative refund rate since launch 14 months ago; last completed week's refund rate was 2.7%, and the failed-payment rate for that week was 4.3%. Over the last 60 days the dashboard drew 190 opens from 4 of the team's 11 engineers, with a median session of 30 seconds, and the analyst still fields 22 requests a month, 20 of which ask for the refund and failed-payment rates for the week that just ended. The lead has budget for one small change and wants the request count down. Which change removes the requests?
MCQ-89An analyst's EMEA revenue dashboard shows Q3 down 8% QoQ. The EMEA sales director disputes it, saying her team grew. The dashboard query joins orders to `accounts.region`, and channel-partner orders are booked under accounts tagged `region = 'GLOBAL'`. Last quarter, channel partners were 28% of the director's EMEA revenue and direct sales were the other 72%. This quarter, channel-partner revenue grew 40% QoQ while direct sales fell 8%. The director's team owns both the direct and channel business in EMEA. What is the correct QoQ revenue change for the EMEA business the director owns?
MCQ-90One analyst covers the ad hoc queue this week with two days of unbooked capacity. Requests arrived Monday in this order: (1) a dashboard colour change, no decision attached, 1 hour; (2) a board slide figure the VP wants Thursday, no decision attached, 3 days; (3) a pricing-test readout feeding a $2M decision due Friday, 1 day; (4) a churn-by-region analysis feeding a $300k retention decision due at month end, 2 days. The team works requests strictly in arrival order. What is the outcome this week?
MCQ-91An analyst's summary of a 4-week free-shipping test (120,000 sessions per arm) reads: 'Free shipping lifted orders per session by 6% (95% CI +3% to +9%). Roll it out for all orders over $30.' The full analysis also has average order value −4% (95% CI −7% to −1%) and revenue per session +1.8% (95% CI −0.5% to +4.1%); the analyst's recommendation is to hold and extend the test. Which statement about the summary is correct?
MCQ-75A ride-hailing marketplace is deciding whether to launch a dispatch rule that shows a driver the next ride request before the current drop-off. For two weeks, 40,000 drivers across the company's 12 city markets are randomized individually within each city, half on the new rule and half on the old one. Treated drivers complete 9% more trips per driver-hour and reach their riders 0.4 minutes faster than control drivers. Across the 12 cities, weekly trips stay at 1.02 million, the same as the four-week pre-period; median rider wait is 4.1 minutes, unchanged; drivers online per day are unchanged. The launch decision is due Friday. Which conclusion should the readout carry?
MCQ-74A weekly report is due today and feeds a roadmap meeting tomorrow. It compares a feature live for six days (9 conversions from 45 users) against the existing flow (7 from 42). The meeting has to decide whether to act on the gap now or defer the call. What should the analyst do?
MCQ-76You are validating a completed-orders extract for a revenue readout. The warehouse reports 1,284,000 completed orders from 2026-01-01 through 2026-09-10. Your extract has 962,000 rows, min order_date 2026-01-01, max order_date 2026-07-20, zero duplicate order_id values, and 0.4% NULL customer_id. The dashboard shows July revenue 38% below June, and a teammate is drafting a 'demand softening' section. What should you do next?
MCQ-77A subscription business computes average revenue per account (ARPA) from a billing export. 800 of its 1,000 accounts have an `mrr` value; the other 200 are blank because their contracts are still being migrated from a legacy system. The 800 with values average $100. A random sample of 50 of the 200 blank accounts, read from the legacy system, averages $250. Leadership is about to act on the ARPA figure. Which number should the analyst report?
MCQ-78A retailer's loyalty table holds 1,000,000 signup rows, and the distinct-customer count sets each member's reward budget. Three dedup rules were run: exact email match leaves 940,000 rows, lowercased and trimmed email leaves 918,000, and lowercased and trimmed email or normalized phone leaves 902,000. A hand review of 200 pairs that the phone rule merged on phone alone found 60 were two different people sharing a household phone. Which rule should the team adopt?
MCQ-79March closes with analytics reporting $1,000,000 gross revenue and finance reporting $960,000 net revenue. Finance will book a separate adjusting entry for each cause of the gap before it closes the month. Which justification of the reconciliation is sound?
MCQ-80A churn analysis quoted 4.2% in a board deck. Three months later, a rerun from the same notebook gives 5.1%. The notebook is in a Git repo and was committed weekly, but the deck slide only says 'Churn: 4.2%' and the dashboard now shows 5.1%. The analyst who edited the query has left. Which change best addresses the cause?
MCQ-87You have a ten-minute slot on the CFO's agenda to get a decision on cutting paid channel X. The evidence: incremental sales attributed to X are 2% of its $2M spend, cutting X frees $400k a year, and brand-search volume may dip about 3% for a quarter. Your draft deck runs 14 slides: six on attribution method, four on data cleaning, two on the recommendation, two on risk. What change to the deck gives the best chance of a decision in the slot?
MCQ-88You must give the exec team next quarter's revenue forecast; they will use it to decide whether to commit $1.2M to a hiring plan that pays back only if revenue clears $11.5M. The model's point estimate is $12.4M. Its backtest over the last 8 quarters has a mean absolute percentage error of 9%, and the error is dominated by the Q4 renewal rate, which ranged from 88% to 96% over that window. Which presentation gives the exec team what it needs to make the call?
MCQ-96You are interviewing for a role posted as "Data Analyst". In the recruiter screen you learn the team runs about 40 A/B tests a quarter, analysts own the experiment readouts, and the last two people hired under this title shipped a churn model into the CRM. Your prepared answer to "what does a data analyst do?" describes building dashboards and weekly reporting, and the interviewer pushes back. What should you do?
MCQ-97You have a 4-minute behavioral slot. Your story: churn rose from 4% to 6% in Q3; you split accounts by plan tier and found the annual-plan price change drove it; the company reverted the price; churn was back to 4% by Q1. Your draft runs 120 s on the ingestion stack, 25 s on the situation, 20 s on the task, 55 s on the action ('I ran a cohort analysis'), and 20 s on the result ('churn improved'). The rubric: situation and task together get at most 45 s; the action is scored on the specific choices you made; the result scores only if it names a number and the decision it changed. Which revision is most likely to score?
MCQ-100An interviewer gives you a one-line prompt: "Our retail client's revenue is down 15% year over year; what's going on?": with no data attached, and will only answer questions you ask. You have 30 minutes. Which opening move holds up?