Choosing the chart first manufactures questions to fit it

What it is

The healthy flow is question → data → chart; the inverted flow is "I want to build a Sankey / word cloud / flashy 3D," followed by a reverse hunt for data and questions that fill the mould. The product is distinctive: the chart is legal and the data real, but the question is tailored to the graphic — the conclusion usually restates the graphic's built-in structure (a Sankey's manufactured "insight" is "flows move from source to target"; a word cloud's is "frequent words are bigger"). The analysis advances nothing; it merely wraps a demonstration.

Why it happens

Every chart carries a preset question set (its affordances): Sankeys natively answer "where does the flow go," treemaps natively answer "shares and hierarchy." Fixing the chart first locks in the answer type, and the hunt for data becomes confirmation-bias harvesting — whatever fits the mould stays, whatever does not is dropped; the manufactured question always "holds," because it is the chart's definition. The detection signal comes from the same place: swap the chart for another — does the "conclusion" survive? If conclusion and chart live and die together, it is a restatement of the graphic, not a finding about the data.

Where it stops holding

Two legitimate ways a chart can come first must be separated out: exploratory sketching (using a chart as a lens on the data, no conclusion owed — conclusions come from the data) and template reuse (fixed report layouts whose questions were settled long ago, the chart being convention). Inversion specifically means delivering a "finding" that the graphic's structure generated. It also differs from deliberate misleading (weaponizing charts — the ethics family); here the process has run off the rails, often without the author noticing. Innovation-driven settings (wanting to use a novel chart) are the most vulnerable and need a human gate on the question.

Applying it

• At review, ask each chart three questions: what was the question before delivery? did it exist before the chart? does the conclusion survive a different chart? A "no" on the third means a restatement-type conclusion — withdraw or redo.
• Run new charts through a "lens" protocol: declare "exploratory sketch, no predetermined conclusion," and require conclusions to be stateable independently of that chart.
• Verification: sample report conclusions and mark whether each one's evidence is visible only in a single chart type; more than a handful of cross-chart-failing conclusions signals inversion.

Related

• Same group: U2.17.1 A mismatched chart forces readers to compute the answer mentally · U2.17.2 One question, several viable charts — they differ in reading cost · U2.17.3 Automatic recommendation reads data types, not analytical intent · U2.17.5 As data magnitude shifts, a once-fitting chart stops working
• Nearby: U2.01.2 Define the question before choosing the chart · U10.01.1 Recognizing misleading intent
• Search terms: chart-first design · confirmation bias · affordance · question-driven design