A chart can show when a number changed, but only an annotation can say why it changed

What it is

What a data graphic can show directly is "when something changed and by how much" (time, value, direction), but not "why"—the causes (a release, a channel adjustment, a holiday, a competitor's move) and the context (policy environment, industry cycle) live outside the data and can enter the chart only through text. Annotations carrying this explanation ("promotional season," "competitor launched in August") connect the patterns on the graphic to events in the reader's world, completing the crossing from "data description" to "business understanding." Without them, the reader sees a shape; with them, a story with causes.

Why it happens

The cognitive mechanism behind contextual annotation is the activation of causal scripts: humans naturally seek causes when interpreting events, and seeing a sudden drop automatically generates candidate explanations ("did we ship a bug?" "did the campaign end?"). If the chart offers no explanatory cues, readers search their own experience for explanations—and the retrieved cause may have nothing to do with the real one (the reader remembers changing prices last week and attributes a drop from two weeks ago to that price change), forming a wrong attribution that may drive subsequent decisions. An explicit contextual annotation interrupts this uncontrolled attribution process: it supplies the explanation the author already knows ("the drop began March 8 when the competitor cut prices"), steering the reader's attribution toward the real event; even if the reader disputes the explanation ("the price cut was only part of it"), the discussion moves from "what caused it" to "how much does this cause explain"—the annotation raises the discussion's starting point from guessing to verifying. The correct position for an event annotation is on the timeline itself (a vertical reference line + event name), letting readers directly check whether "event time" aligns with "change time"—which is itself a testable causal hint.

Where it stops holding

Contextual annotation carries "explanation," not "proof": marking "release shipped" on the timeline hints at a relation to the change but is not causal evidence (other events may have coincided), and the wording must preserve that distinction ("release shipped (channel adjustment ran concurrently)" is more honest than "the release caused the growth"). The information source of annotations needs a boundary too—only background the author knows should enter the annotation; speculative explanations ("possibly a competitor effect") must be explicitly marked as speculation (prefixed with "possibly" or moved to discussion notes), or readers will treat conjecture as confirmed fact. Annotation density spirals out of control in event-heavy periods (major sales, frequent releases); the fix is annotating only events directly relevant to the current narrative, with the full event timeline in an appendix or an interactive layer.

Applying it

• On time-series charts, drop vertical reference lines for key business events (releases, campaigns, incidents, policy changes) with event names on the axis.
• Distinguish fact from speculation in annotation text: state confirmed events directly, prefix speculative explanations with "possibly" or "conjectured."
• When event time and change time do not align, do not force an annotation implying causality; note the two timelines in the footnote instead.
• Verification: ask readers to explain the cause of the chart's largest change and count how many cite an annotated event; a high proportion inventing other causes means contextual annotation is missing or insufficient.

Related

• Same group: U9.03.1 Annotations point to where readers should look · U9.03.3 Unannotated charts are interpreted idiosyncratically
• Nearby: U9.03.1 Annotations point to where readers should look · U7.05.5 The comparison period's statistical scope must match the current period's
• Search terms: event annotation · contextual annotation · reference line