The same metric can read as an all-time high or a three-week slide, depending only on the baseline

What it is

The same metric value can be "an all-time high" and "down three weeks running" at once — depending on which point it compares to: rising against last year, falling against two weeks ago. Choosing the baseline is choosing the second point of the comparison, and freedom in the second point means freedom in the conclusion's direction. It is the most direct narrative lever in metric presentation, and selective presentation's favourite tool.

Why it happens

Rise and fall are the outputs of a differencing operation, and differencing needs two points: the current point is fixed, the baseline point is optional, and where the baseline sits in history decides the difference's sign. When a metric is declining recently but sits on a low base further back, "up year-over-year" and "down three weeks straight" are simultaneously true — each states facts, with completely opposite frames. Turning that freedom into misleading requires three conditions: multiple baselines available, opaque selection, and single-baseline presentation. Breaking any one defends against it: a governed primary baseline (fixed by metric semantics), visible baselines (readers know the comparison), and parallel baselines (rise and fall both shown for readers to see whole). The relationship to the misleading-graphics line is clean: the narrativity of baseline choice exists objectively; cherry-picking the favourable one is the manipulation.

Where it stops holding

Governing the primary baseline costs some flexibility: different scenes genuinely need different baselines (weekly reports read year-over-year, daily watch reads period-over-period), and governance means "default plus transparency," not lock-down — switching is allowed but its state visible. Changing a metric's primary baseline is a narrative change and should be logged and announced like a numerical change (reusing the metric-definition change notification). Reporting contexts are the most sensitive: every period-over-period versus year-over-year choice in front of management should survive the question "why this baseline" — the answer must be task fit, not merely favouring the conclusion.

Applying it

• Define a primary baseline per metric in the metric dictionary ("GMV primary baseline = year-over-year") and default all views to it.
• Additional comparisons may be shown in parallel; never hide the one contradicting the headline.
• Verification: audit the baseline choices in recent reporting material against the dictionary; favourable deviations without task justification go back for correction.

Related

• Same group: U7.05.1 A single value means something only against a baseline · U7.05.2 Period-over-period suffers cyclic interference; year-over-year removes cycles but reacts late · U7.05.4 With a tiny base, percentage change inflates into distortion · U7.05.5 The baseline period's statistical scope must match the current period's
• Nearby: U10.02.1 The chosen time window decides the trend's direction · U4.03.2 Midpoint placement changes the conclusion
• Search terms: baseline framing · cherry picking · metric dictionary