Beyond the Bar Chart

Why this matters

The VP looks at your Claims page and says the thing every analyst eventually hears: "Can I see it on a map?" And here is what makes this lesson necessary: the map you would build in thirty seconds — State from stores, revenue as the value — renders beautifully, colors twenty-three states, and silently omits $29.89M: 44.7% of the company, because Online has no state. Nothing errors. The specialty visuals in this lesson are all like that: genuinely powerful, each with a failure mode sharper than anything the core four can manage. The craft is knowing both halves.

Maps: powerful, and twice deceptive

Build it: a filled map, State from stores, [Total Revenue] as color saturation. California leads at $2.91M, New York and Texas just behind ($2.84M, $2.82M), Indiana trails the mapped states at $0.93M. Geographic patterns that a bar chart of 23 states would bury — coasts versus interior, region clusters — appear at a glance. Maps earn their popularity.

Now the two deceptions:

• The missing-geography problem. Our biggest "region" — Online, $29.89M, nearly forty-five cents of every dollar — appears nowhere, because it has no state. A reader comparing the map's total to the revenue card comes up short by almost half. The rule: a map's title or caption states what it excludes. "Retail revenue by state (excludes Online, 44.7% of total)" is honest; "Revenue by state" is a silent lie of omission.
• The area confound. Filled maps color land, and land is not revenue. Texas at $2.82M paints a vast area; New Jersey-sized states vanish at the same revenue. Readers integrate color × area whether you want them to or not, so big empty states read as important. Mitigations: a bubble map (size encodes the value, position encodes place), or accepting the filled map as a pattern tool and pairing it with the sorted bar that carries the ranking honestly.

Scatterplots: the positioning chart

A scatter claims: "each point is a thing, positioned by two measures." Build the product scatter — 240 points, X = units sold, Y = average selling price (DIVIDE([Total Revenue], [Total Units]) per product), color by category, and read its landmarks like a resident:

• Top-left: Aurora Signature Bed Frame — $941 average price, 922 units. Premium-low-volume, and quietly the company's top revenue product at $867.6K. The scatter is the only chart where that strategy is visible as a location.
• Bottom-right: Cedarline Petite Floor Lamp — 1,399 units, the volume king, producing $163.4K. Same company, opposite corner, opposite business.
• The category colors show Furniture living up-left and Lighting down-right — five categories, five point-clouds, one picture of the catalog's structure.

Scatter craft: label the landmark points (not all 240); consider log scales only when you can defend them aloud; and resist trend lines until you have a claim the line supports.

Conditional formatting: color as evidence

Take the matrix of [Return Rate], Category rows × Region columns, and apply a sequential color scale. The grid runs from 2.44% (Decor–Northeast) to 3.82% (Furniture–Midwest) — and the heat map's honest headline is how narrow that band is. No cell doubles another; there is no crisis cell. That is a finding: return behavior is stable across the business, and a manager hunting for a "returns problem region" should be told the data does not show one.

The craft rules: sequential scale for magnitude (light→dark of one hue — the Iris theme's ramp), diverging scales only when a meaningful midpoint exists (attainment around 100% qualifies), and always numbers in the cells — color ranks, numbers prove. Data bars do the same job inside a table column: length, the honest encoding, smuggled into a matrix.

Small multiples: the stacked column's cure

Lesson 7 showed why a stacked column cannot support "Lighting's momentum continues" — every segment floats on the ones below. The cure is small multiples: one line chart of monthly revenue, split into five tiles, one per category, sharing axes. Each category gets its own baseline; Outdoor's summer wave, Decor's December spike, and Lighting's steady climb become five readable sentences instead of one floating stack. Shared axes are the integrity rule — per-tile axes turn comparison into deception.

Tooltips: the reader's next question

Default tooltips show the point's values — fine. The upgrade is adding the measures a reader would ask for next: hovering a category bar that shows revenue should also offer margin %, units, and share of total in the tooltip fields. The full move — report-page tooltips, a miniature page that appears on hover — belongs in your toolkit with a warning label: build one when hover-context genuinely saves a click, not because it demos well. A tooltip nobody hovers is invisible work.

KPI cards, gauges, waterfalls, funnels — and skepticism

• KPI visuals (value + target + trend) earn their place the moment a target exists — attainment against the 100% line is the natural fit.
• Gauges spend half their pixels on decoration to show one number against one target — a KPI card usually does it better. Boardrooms like them; know the cost.
• Waterfalls claim "here is how we got from A to B in named steps" — year-over-year revenue change decomposed by category is the honest use.
• Funnels claim ordered stage-shrinkage; without a real staged process behind them they are a sorted bar wearing a costume.
• Custom visuals from the marketplace: check certification (certified visuals can't call home with your data), check the publisher, and ask whether a core visual does the job — every custom visual is a dependency someone must maintain after you leave.

Common mistakes

• Maps that don't say what they exclude. Rows without geography vanish silently — here, 44.7% of the company.
• Filled maps asked to rank. Area confounds color; rankings belong to the bar beside the map.
• Scatters with every point labeled. 240 labels is a word cloud. Label landmarks; tooltip the rest.
• Diverging color scales without a meaningful midpoint. Red-to-green implies good and bad; a magnitude question wants one hue, light to dark.
• Custom visuals as first resort. Certification, publisher, and "could a core visual do this?" — in that order, before installing.

Summary