05 · Formatting & Storytelling with Data

A technically correct chart can still fail to communicate. This module covers formatting discipline and Tableau's Story Points feature, using the Orders table's Region totals (East 2210, West 3750, Central 920) as the running example.

1. Number formatting that matches the audience

1. Right-click a measure pill → Format → set Numbers to Currency (Custom) with 0 decimal places for an executive audience: $2,210 reads faster than $2,210.00 or the raw 2210.
2. For percentages (e.g. Pct of Region Sales from Level 2 Module 1), format as Percentage with 1 decimal: East's Bookcase share (950/2210 = 0.4299...) should display as 43.0%, not 0.43 or 42.99% — 1 decimal is the standard precision for a share-of-total metric at this scale.
3. Use Unit abbreviation (thousands K / millions M) only when numbers are large enough to need it; forcing $2.2K for a $2,210 value on a small dashboard often reads as less precise, not more — a design judgment worth stating explicitly to students who over-apply this.

2. Color: sequential vs. diverging vs. categorical

1. Categorical palette (distinct hues) for Region: East, West, Central get three visually distinct, non-ordered colors — correct because Region has no inherent order.
2. Sequential palette (single hue, light→dark) for SUM(Sales) itself: darker = higher sales. Central (920) would be the lightest shade, West (3750) the darkest, correctly encoding magnitude.
3. Diverging palette (two hues meeting at a midpoint) for Profit, centered at 0: the Bookcase order's Profit (-40) should render in the "loss" hue while every other order (all positive Profit) renders in the "gain" hue — diverging color is the only palette type that visually flags the one negative row in this dataset.

3. Titles, subtitles, and annotations that carry the point

1. A worksheet title should state the finding, not just the field name: prefer "West leads at $3,750 — 55% of total Sales" over a generic "Sales by Region". Verify the 55% claim: 3750/6880 = 0.545 ≈ 55%.
2. Add a mark annotation (right-click the Bookcase mark → Annotate → Mark) reading "Only loss: -$40 profit" — annotations should call out exceptions the audience would otherwise have to hunt for.
3. Avoid over-annotating: with only 8 rows, annotate at most the 1–2 marks that matter (the single loss, the single largest order) rather than labeling every mark, which recreates a dense table and defeats the purpose of a chart.

4. Story Points

1. Story is a Tableau sheet type (right-click the sheet tabs → New Story) that sequences multiple existing worksheets/dashboards into a guided narrative, each step called a story point.
2. Build a 3-point story on the Orders data: Point 1 = "Overview" (bar chart of the three region totals: 2210 / 3750 / 920). Point 2 = "What's driving West" (West's 3 orders: Monitor 450, Laptop 2200, Desk 1100 — the Laptop alone is 2200/3750 = 58.7% of West's total). Point 3 = "The one exception" (Bookcase, Profit -40, the only loss in the dataset).
3. Each story point can carry its own caption (below the point navigator) — use short, declarative captions ("West's $3,750 is driven by one $2,200 laptop order") rather than restating the chart title.

5. Layout discipline: the "5-second test"

1. A dashboard should communicate its headline number within 5 seconds of viewing — for this dataset, that's the $6,880 grand total and the $3,750 West-leads-Region finding, both of which should be the largest, top-left elements, not buried below a legend or filter panel.
2. Group secondary detail (the full 8-row order table) below or behind a Show/Hide toggle (Level 2 Module 2) rather than competing for the same visual weight as the headline chart.

How It Actually Works

Formatting and color are almost entirely a rendering-layer concern applied after VizQL's query has already returned its numbers — but two mechanisms in this module do interact with the query/calc engine directly:

1. Number formatting never changes underlying values — it's a display transform applied to the same returned aggregate. East's Sales stays the exact integer 2210 in the query result regardless of whether it's rendered as 2210, $2,210, or $2.2K; verify this distinction matters by noting a calculated field referencing [Sales] in a formula always sees the raw 2210, never a rounded or abbreviated version, even if every visible label on the sheet is abbreviated.
2. Diverging color's midpoint is a genuine value comparison, not purely visual: Tableau's diverging palette assigns each mark's color by comparing its numeric value against a computed or user-set center (default: the field's midpoint, or 0 if the field can meaningfully cross zero) — the Bookcase order's Profit (-40) is colored differently only because -40 < 0 is evaluated per-mark before rendering; the color assignment is itself a small per-mark classification computed from the same aggregate the axis already shows, not a separate stored field.
3. Story Points don't create new queries — each story point is a captured reference to an existing sheet/dashboard's current filter and parameter state (a saved "bookmark" of shelf/filter configuration), so navigating between points re-applies whichever filter/parameter state was captured, potentially re-triggering that sheet's query if the state differs from what's currently cached, but building no new calculation logic of its own — this is why a story point can "break" if the underlying sheet's fields are later renamed or removed: the story only holds a reference, not an independent copy.

Cheat sheet

Element	Rule of thumb
Currency formatting	0 decimals unless sub-dollar precision matters
Percentage formatting	1 decimal place is standard
Categorical color	Unordered dimensions (Region, Category)
Sequential color	Ordered magnitude, one direction (Sales)
Diverging color	Values that cross a meaningful zero (Profit)
Title	State the finding, not the field name
Story point	One narrative beat per point; short caption

Exercise

Write three story-point titles (not generic chart titles) for a story built from the Orders data that would each pass the "states a finding, not a field name" test — one about the grand total, one about the West region, and one about the Bookcase loss — and hand-verify each numeric claim you use.