Business Intelligence with Power BI

PF · Practice Final

Practice Final — Concept Rehearsal

What this is. Twelve questions in the final's own shape: cumulative, weighted toward Modules 3 and 4, every one situational. These are fresh questions — not the exam bank — so treat your score as a diagnosis, not a prediction. Answers, with the reasoning that matters more than the letter, are at the bottom.

How the real final behaves

The final is a concept exam: no dataset, no Desktop, delivered in the quiz environment. It is cumulative, but deliberately back-weighted — Modules 3 and 4 carry most of it, because by Week 15 the modeling skills have been examined twice already (midterm, Project 1) and the visualization-and-judgment skills have only been examined by assignment. Expect situations, not definitions: every question below hands you a scenario and asks what a careful analyst does, which is exactly how the bank is written.

Questions

  1. A stakeholder asks for "revenue by the 23 states we operate in, at a glance, ranked." A colleague reaches for a map. Make the case for a sorted bar chart instead — and name the one question a map answers that the bar cannot.
  2. A regional sales chart starts its axis at $5M "to make the differences visible." The bars now suggest the West does triple the Southeast's business; the true gap is 34%. Name the technique, the honest alternative that still shows small differences, and the reason "the numbers are labeled anyway" is not a defense.
  3. Your filled map of revenue by store state looks complete and correct. The company does 44.7% of its business online. What just happened, and what is the fix that does not involve deleting the map?
  4. One line chart shows five product categories as five lines; nothing can be read where they cross. Name the visual that fixes this, and the property that makes the five panels honestly comparable.
  5. A scatter of 240 products plots units sold against average selling price. A cluster sits low-volume/high-price; one point sits far right at high volume. What business question does each region of that chart answer, and why would a bar chart of the same data answer neither?
  6. A drillthrough page shows "Category Detail" and works perfectly — for you. A colleague opens the report and cannot find it. What did the report author forget about how drillthrough pages are reached, and which design habit fixes it?
  7. Two visuals sit on one page. Selecting a bar in the first should highlight the second on one page, but filter it on another. Where does that behavior live, and what is the difference a reader sees?
  8. A report ships to executives who read it on phones in the elevator. Name the two design deliverables that decision imposes beyond the desktop layout, and the failure mode of skipping each.
  9. Row-level security filters the Midwest role to its region. A Midwest manager opens the report and the total revenue card still shows the whole company's number. Diagnose: what kind of gap is this, and where do you look first?
  10. You publish a report and the numbers are right; a week later they are stale. The data sits in files on a shared drive. Name the two pieces of Service machinery involved in keeping it fresh, and which one the shared-drive detail forces.
  11. An AI feature drafts a DAX measure for you and it returns a number. The course's verification standard says the number is not yet an answer. What are the two checks that turn a generated measure into one you would sign?
  12. A junior analyst says: "The dashboard is done — every chart the stakeholders asked for is on it." Give the course's two-part rebuttal: what a report is for, and the one-page written artifact this course made you produce to prove it.
Answers and reasoning — open after your attempt
  1. Ranking is a length judgment; humans compare lengths far better than shaded areas, and 23 sorted bars scan in seconds. The map's sole advantage is geography itself — adjacency, clustering, "is the Southwest soft as a region?" If the question has no where in it, the map is decoration.
  2. Axis truncation. The honest alternative: keep the zero baseline and show a second visual (or a delta/variance chart) for the differences. Labels don't help because the preattentive read — bar length — happens before anyone reads a number, and the first impression is the one that travels.
  3. The map silently dropped every row with no state — Online has no geography, so nearly half the business vanished without an error. Fix: caption the map with what it excludes and pair it with a card showing the true total ("the caption is the map"). A visual that can't show a category must confess it.
  4. Small multiples: one panel per category. The property that keeps the comparison honest is shared axes — let each panel autoscale and every panel's story becomes "up and to the right" regardless of magnitude.
  5. Low-volume/high-price answers "what do we sell rarely but profitably?" (candidates for promotion or quiet discontinuation); the far-right point answers "what carries the volume?" (protect it). A bar collapses two dimensions into one, and the whole point is the relationship between the two.
  6. Drillthrough is reached by right-click on a data point — which nobody discovers on their own. The habit: put a visible back button on the target page (Power BI adds one) and a line of on-page instruction, or surface the same path with a button. Interaction no one can find is interaction that does not exist.
  7. Edit interactions, set per source visual, per page. Highlight dims the non-matching portion but keeps context (both totals still visible); filter removes non-matching data entirely and the second visual recalculates. The reader's tell: highlight keeps ghost bars, filter changes the numbers.
  8. A mobile layout (portrait canvas, cards and one key visual, thumb-sized targets) and ruthless prioritization of what appears there. Skip the layout and phones get a postage-stamp desktop page; skip the prioritization and the layout is a scroll of everything, which is the same failure with better formatting.
  9. A filter-flow gap: the role's filter sits on a dimension table that does not reach the fact table the card reads — wrong direction, missing relationship, or a measure built on an unrelated table. Look first in Model view and trace the path from the filtered dimension to the fact; RLS is just a filter, and filters only travel existing roads.
  10. Scheduled refresh, and a data gateway. Files on a shared drive are on-premises from the Service's point of view, so refresh cannot reach them without the gateway — the exact detail the deployment-plan assignment graded.
  11. First: recompute the number by an independent route (a manual filter, a pivot on the raw table, a second measure written yourself) and match it. Second: read the generated DAX and explain every function's role in your own words. A measure you cannot explain is a measure you cannot defend when it's challenged in a meeting.
  12. A report is for a decision, not a display; charts are evidence, not product. The artifact: the capstone's one-page insight memo — the three findings and what the business should do next. "Done" means someone can act, not that every request rendered.

Reading your result

Missed questions cluster by module, and that is the point of a rehearsal: 1–5 send you to Lesson 7 and Lesson 8, 6–8 to Lesson 9 and Lesson 10, 9–10 to Lesson 11, 11 to Lesson 12, and 12 to the capstone brief. Reread the lesson, not just the answer — the bank does not repeat these questions, but it asks from the same places.