Gridlines only earn their place when a reader actually needs to read an exact value off the chart

What it is

A gridline's only legitimate duty is reading support: it extends the axis scale into the plot area so "point answers value" can be estimated visually. When the reader's task is trend, shape comparison, or anomaly spotting — no specific values need to be read — the gridline loses its duty and remains pure noise: it densifies the canvas, slices the data shapes, and competes with reference lines for the background-structure role.

Why it happens

The keep-or-drop criterion is the task, not the style: value-reading tasks ("what is March's value?") need the grid because visual interpolation demands an internal reference; trend tasks ("which month grew fastest?") do not, since trend judgement runs on shape and slope, and the grid joins no computation — only interference. The interference has three paths: added visual density lowers data contrast; the horizontal-vertical lattice cuts shapes into fragments; and the grid competes with reference lines for classification as background structure. Interactive charts can further outsource reading to hover, dropping the grid's necessity again.

Where it stops holding

"Drop the grid" does not mean "zero scaffolding": a chart with no internal reference at all is equally hard to estimate against, and pale major-tick marks (short, extremely light, at major ticks only) are the common compromise. Print cannot interact, so genuine reading demands keep the grid; dashboards built for trend monitoring drop it first. Axes and tick labels are baseline configuration outside this rule — what goes is the reference structure inside the plot area, not the coordinate system itself.

Applying it

• Fix the reader's task before drawing: reading values → keep a sparse major grid; comparing trends → drop the grid, keep axes and tick labels.
• When kept, draw major ticks only, one set horizontally or vertically, never both plus minor grids.
• Verification: build a grid-free version and run the readers' task; a clear drop in estimate accuracy means the grid is necessary; unchanged performance means it can be removed for good.

Related

• Same group: U5.03.1 Reference lines mark external benchmarks such as thresholds, means, or targets · U5.03.2 Reference lines must be labelled or they read as data · U5.03.3 Reference-line visual weight belongs between data and gridlines · U5.03.4 Multiple reference lines blur into the data series
• Nearby: U5.01.2 Direct labelling eliminates the round-trip · U5.06.4 High-density charts keep overall shape and lose individual readability
• Search terms: gridlines · chart minimalism · value reading task