A line drawn straight through a missing data point quietly invents a trend that was never measured

What it is

Many chart libraries by default connect the data on either side of a missing point with a straight line or curve—the line appears continuous and smooth, implying "the data transitioned steadily through this period." In reality there were no observations at all during that time, and every point in the middle of the segment is fabricated: the line is an interpolation, not data. Readers cannot distinguish "continuously observed change" from "an interpolated line bridging a hole" because the two look identical. This is more harmful than the missing data itself: a gap at least leaves a visible hole, while interpolation erases the hole and leaves behind a false impression of continuity.

Why it happens

The fabrication works through the visual grammar of line charts: a continuous line segment conventionally means "a genuine connection between adjacent observations," and readers assume every point on the line is supported by data. Interpolation violates this contract—the middle of the segment has no observations yet is drawn exactly like an observed segment. The harm's severity depends on how the true change behaved during the gap: if the real change happened to be nearly linear (a slow daily growth trend), the interpolation lands close to truth and misleads little; if a sudden change occurred during the gap (a crash-time plunge, a post-restart recovery), the smooth interpolated slope diverges completely from the true path—it shows neither the plunge nor the recovery, just a calm "passing through" line. The standard honest treatments are: break the line across the missing interval (leaving a gap), or draw the interpolated segment as dashed/lighter and explicitly label it "interpolated." A gap says "we don't know here"; a dashed line says "this is a guess"—both are honest. A solid line crossing is the only dishonest option.

Where it stops holding

Interpolation should not be banned outright: in regular sampling with an occasional single missing point and highly stable neighbors (per-minute server metrics with one minute missing), linear interpolation's error is tiny and a visual break's cost (interrupting the reading flow) may exceed its benefit—light interpolation with a subtle visual cue (a lighter segment) is a reasonable compromise. The criterion is "whether interpolation error could exceed the reader's interpretation precision for the line": when error is far below interpretation precision, interpolation is harmless; when it could flip a trend direction or mask a sudden change, the line must break. Another boundary is interactive compensation: if the tooltip in the interpolated segment explicitly says "no data in this interval; the segment is interpolated," the solid line's misleading power drops substantially—dual visual-plus-text annotation partially compensates for a single visual encoding's limits, but never fully replaces the honesty of a break.

Applying it

• Default line charts to break (gap) across missing intervals; disable automatic crossing behavior.
• When interpolation must be shown (e.g., preserving visual flow), draw the interpolated segment dashed or at 50% opacity and label it "interpolated" in the tooltip.
• Enforce system-level settings that disable chart libraries' default interpolation (e.g., ECharts's connectNulls defaults to off).
• Verification: remove a known interval containing a sudden change from test data and check whether the chart shows a break or a smooth line crossing it; a smooth crossing is a fabricated-continuity instance.

Related

• Same group: U8.04.1 Missing must be visually distinguished from zero · U8.04.3 Missing proportion affects conclusion credibility
• Nearby: U8.04.1 Missing must be visually distinguished from zero · U2.03.2 Categorical x-axes should not be connected with lines
• Search terms: line interpolation · gap in line chart · connectNulls