Lecture video · L08 Beyond the Bar Chart

[Slide 1] Tonight starts with the sentence every analyst eventually hears: can I see it on a map? And the map you would build in thirty seconds — states colored by revenue — renders beautifully, and silently omits twenty-nine point nine million dollars. Forty-four point seven percent of this company sells online, and online has no state. Nothing errors. Tonight is the specialty visuals: maps, scatterplots, heat grids, small multiples — each genuinely powerful, each with a failure mode sharper than anything the core four can manage. The craft is knowing both halves, and by the end of the session you will.

[Slide 2] The map first, because it is the most requested and most dangerous specialty visual. Build it: states colored by revenue. California leads at two point nine one million, New York and Texas just behind, and genuine geographic pattern appears at a glance — that is the power. Now the two deceptions. First: rows with no geography simply do not render, and our largest region — Online, twenty-nine point nine million, nearly forty-five cents of every dollar — has no state. A reader comparing the map to the revenue card comes up half short with no way to know why. So the rule: a map's title or caption states what it excludes, every time. Second: filled maps color land, and land is not revenue — Texas paints a vast area at the same revenue a small state paints a sliver, and readers integrate color times area whether you want them to or not. The pairing that fixes it: the map shows pattern, and a sorted bar beside it carries the ranking honestly.

[Slide 3] The scatterplot claims something no other chart can: each point is a thing, positioned by two measures at once. Two hundred forty products — units sold across, average selling price up, category as color — and the catalog becomes a map of itself. Upper left: the Aurora Signature Bed Frame, nine hundred forty-one dollars average, nine hundred twenty-two units — premium, low volume, and quietly the company's top revenue product at eight hundred sixty-seven thousand dollars. Lower right: the Cedarline Petite Floor Lamp, fourteen hundred units at a hundred seventeen dollars — the volume business, in the opposite corner. No bar chart shows how an outcome is achieved; the scatter shows strategy as location. Craft note: label those two landmarks and let tooltips carry the other two hundred thirty-eight — every point labeled is a word cloud.

[Slide 4] Prediction beat. A colleague builds the scatter with total revenue on the vertical axis instead of average price, gets a gorgeous upward sweep, and concludes volume drives revenue. One minute: what is wrong? Here is the trap: revenue is units times price — the vertical axis contains the horizontal one. The correlation is real, and it is empty: a quantity plotted against something built from itself must trend, by arithmetic alone. Real is not the same as informative. Swap the axis to average selling price and the sweep dissolves into the honest structure — the category clouds and the two corners. You will build both versions in the Medium practice, because the trap is much easier to recognize after your own hands have set it.

[Slide 5] Conditional formatting turns a matrix into a heat grid, and the return-rate grid is our worked example: category by region, sequential color scale, numbers visible in every cell. It runs from two point four four percent — Decor in the Northeast — to three point eight two — Furniture in the Midwest. And the honest headline is the narrowness: no cell doubles another, no crisis cell exists, and a manager hunting a returns problem region should be told this data does not show one. A finding of stability is a finding. The craft rules: sequential scales — one hue, light to dark — for magnitude; diverging scales only when a true midpoint exists, like attainment around one hundred percent; and numbers always visible, because color ranks and numbers prove. One warning with teeth: the scale's endpoints are adjustable, and stretching them across a narrow band manufactures drama the values do not support. That is the truncated axis, generalized — and you will build it once, on purpose, to learn its face.

[Slide 6] Rapid tour of the rest of the bench. Small multiples cure the stacked column from last week: one line per category, five tiles, each with its own zero baseline — Outdoor's summer wave and Decor's December spike finally readable — and shared axes are the integrity rule, because tiles beg comparison and unshared scales betray it. Tooltips: the default shows the point's values; the upgrade adds what the reader asks next — hovering revenue should offer margin, units, and share. Build a full report-page tooltip only when hover-context genuinely saves a click. The specialty bench: KPI visuals earn their place wherever a target exists; gauges spend half their pixels decorating one number — know the cost before the boardroom asks; waterfalls decompose A-to-B in named steps; funnels assert a staged pipeline and are a costume without one. And custom visuals from the marketplace get three questions in order: is it certified — certified visuals cannot phone home with your data — who publishes it, and could a core visual do the job? Every custom visual is a dependency someone maintains after you leave.

[Slide 7] Tonight in five lines. Maps are pattern instruments that must confess their exclusions — one caption stood between your reader and a missing thirty million dollars. Scatters position things by two measures at once, and strategy becomes location — Aurora in one corner, Cedarline in the other. Color scales are evidence: sequential for magnitude, diverging around true midpoints, numbers always visible, and endpoints never stretched for drama. Small multiples give every category its own baseline with one shared scale. And the specialty bench — KPIs, waterfalls, funnels, custom visuals — earns its place one honest claim at a time. This week's Hot practice hands you two executive questions, one of which the data can only half answer — and saying which half is the deliverable. Next week: interactivity — slicers, drillthrough, bookmarks — where the reader stops reading your page and starts driving it.