Week 7 Practice: Iron Viz on Video Game Sales

BUS220 - Business Intelligence and Analytics

What You’ll Need

  • Tableau Public on your laptop
  • video-game-sales.csv (download from Moodle, ~19K rows)
  • Pen and paper for sketching

The Setup

The lecture opened with the Iron Viz analogy: three finalists, the same dataset, 20 minutes to build. The winner won not on analysis depth, not on polish — but on the story they told with the data.

Today is a forgiving version of that. You have a fresh dataset, two sessions, and a simple brief:

Find a story in this data. Sketch what it looks like. Build the dashboard that tells it. Write the script you would say if you were presenting it.

You own the question, the angle, and the Big Idea — nothing is prescribed.

Submission Checklist

2 points (satisfactory completion) | Deadline: 3 days after the practice session

Submit in Moodle:

1. One Tableau workbook (.twbx) containing:

2. Moodle submission text — a short presentation synopsis (100-150 words):

The arc of what you would say if you were presenting your dashboard to a publisher, analyst, or journalist. Not the whole script — just the core, in full sentences:

  • Setup (1 sentence) — what the data covers and what question you took on
  • Conflict (2-4 sentences) — what did the data show that was unexpected, or that cut against a common assumption?
  • Resolution (2-3 sentences) — your Big Idea delivered as a claim, plus what you would recommend the audience do with it

Write it the way you would say it. If you would not read this sentence to a friend, rewrite it.

The Dataset

Video game sales from Maven Analytics, sourced from vgchartz.com. One row per (title, console) release. Sales are in millions of copies sold.

Columns

Column Description
title Game title
console Platform the game was released for
genre Genre (20 unique values)
publisher Publisher (many values — expect to top-N or filter)
developer Developer (thousands of values — useful as drill-down, not top-level)
critic_score Metacritic score out of 10
total_sales Global sales, millions of copies
na_sales North America sales, millions
jp_sales Japan sales, millions
pal_sales Europe + Africa sales, millions
other_sales Rest of world sales, millions
release_date Release date

Known gotchas — read these before you start

  • Critic score is sparse — ~90% of rows have no score. The ~10% that do have one can support a “do critics predict sales?” angle, or a “who gets rated vs who doesn’t” angle. But most of your analyses will not use it.
  • Regional sales are sparse. total_sales is populated for every row, but the four regional columns are fully populated for only a small subset (~2.2K rows). If you want a “NA vs EU vs JP” story, filter to rows where all four regions are non-null first.
  • Time coverage is usable 1990-2019. Data before 1990 has very few rows; data after 2020 is trivially small.
  • Sales are in millions of copies, not dollars. Format your axes and tooltips accordingly.

Shape hint

The data is in wide shape — four regional sales columns side by side. If your story involves comparing regions directly (e.g., a bar chart of sales by region), you will probably want to unpivot using a script or spreadsheet tool, the same way you did in Week 3. If your story doesn’t need region as a single dimension, wide shape is fine.

Session 1: Explore → Sketch → Start Building

Timings on each phase below are approximate — use them as rough guides, not a stopwatch. Move on when the phase has done its job for you.

Explore (~20 min)

Open the CSV in Tableau. Build a few throwaway views just to see what’s in there — drag fields onto rows, columns, color. Don’t polish anything. You are looking for patterns that surprise you, measures that don’t move the way you expect, subsets that behave differently from the whole.

Take notes on paper as you go. Write down at least 2-3 candidate questions the data could answer — questions that would be worth a publisher’s, analyst’s, or journalist’s attention.

Do not skip this phase. Students who open Tableau and start building immediately usually build the wrong thing.

Sketch on paper (~20 min)

Pick one question from your list.

Write your Big Idea candidate — one sentence that states what the data tells you about that question, in a way the audience could agree or disagree with. (The “so what?” test from the lecture applies.)

Sketch your dashboard on paper. Not in Tableau, on paper. Decide:

  • What is the headline view — the chart the audience sees first?
  • Which KPIs (single numbers) anchor the story? Revenue? Title count? Regional share? Peak year?
  • What supporting charts help the headline land?
  • Where does interactivity sit, and what question does it answer?
  • What’s the reading order — top to bottom, left to right?

Roughly plan for 4 charts + 2 KPIs — that is the submission bar. A sketch that fits on a single A4 page is probably the right scope.

Sketch review (~5 min)

When your sketch is done, flag a TA or instructor. They’ll spend a minute with you on the sketch:

  • Is your question clear enough to build around?
  • Is the scope right for the time you have left?
  • Does the reading order tell the story you wrote down?

Help with implementation — which Tableau feature, which chart type — is also available; raise your hand whenever you’re stuck.

Start building (rest of the session)

Back to Tableau. Build.

Goal by end of Session 1: at least two working worksheets that correspond to your sketch. Full dashboard assembly happens in Session 2.

Session 2: Build → Polish → Write the Story

Same rule — timings are approximate. The phase order matters more than the clock.

Continue building (~40 min)

Finish the remaining worksheets — you’re aiming for 4 charts + 2 KPIs contributing to a single dashboard page. Assemble them on one dashboard in the reading order you sketched.

Add your interactivity — whichever element you sketched. Test that it actually serves the story (a filter that lets the audience answer a follow-up question, a parameter that changes what the headline shows, a highlight action that connects two views). Interactivity for its own sake is a design failure.

Revise your sketch as you build. Dashboards almost always diverge from the sketch once real data hits real chart types — that’s fine. But stay anchored to the Big Idea you wrote. If a worksheet isn’t serving the Big Idea, cut it.

Polish (~15 min)

  • Dashboard title — put your Big Idea here, one sentence. This is the first thing the audience reads and the last thing they remember.
  • Worksheet titles — describe what the view shows, not what type of chart it is. “Sports overtook Action after 2005” beats “Bar chart of genre sales.”
  • Tooltips — remove the default field dump. Keep 3-4 fields that a viewer would actually want.
  • Colors — one consistent palette. If you use color to encode a dimension, use the same colors everywhere that dimension appears.
  • Number formatting — sales in millions with one decimal, percentages as percentages, dates formatted readably.
  • Hide worksheet tabs — right-click each worksheet tab → Hide Sheet. Keep the dashboard tab visible. The audience should open your .twbx and land on the dashboard.

Write the story (~15 min)

Write a short synopsis of the presentation arc — 100-150 words total, in full sentences. Not the whole script, just the core:

  • Setup (1 sentence) — what the data covers and what question you took on.
  • Conflict (2-4 sentences) — what is the tension in the data? What would an audience likely assume, and what does the data show instead? If there is no tension, there is no story — go back to your candidate questions from exploration.
  • Resolution (2-3 sentences) — your Big Idea delivered as a claim, plus what the audience should do with it.

Write in full sentences, the way you would say it out loud. No bullet points, no “as previously mentioned.” If you would not say it to a friend, rewrite it.

Paste this into the Moodle submission text box.

Wrap

Save as .twbx with data embedded. Verify the worksheet tabs are hidden. Submit both the workbook and the script.

Takeaways

  • The lecture’s 63%/5% claim applies here. A dashboard without a Big Idea is a collection of charts someone has to figure out. A dashboard with one is a story they remember.
  • Pick a question, not a dataset angle. “What’s interesting about video game sales?” is a dead end. “Did Japan ever lead the global market?” is a story.
  • The sketch matters more than it feels like it should. Fifteen minutes on paper saves an hour of building the wrong dashboard.
  • Interactivity serves the story or cuts itself. Every filter, parameter, and action is a claim that the audience will want to ask this follow-up. If they won’t, it shouldn’t be there.
  • Writing the synopsis tests the dashboard. If the arc collapses when you try to put it on paper, the dashboard was not telling a story — you just noticed before your audience did.
  • This is the method you apply to the group project next week. Same explore → sketch → build → write-the-story arc, same Big Idea discipline, same story-over-tool-mechanics priority.