A trend chart's job is showing which way a variable is heading and how fast, not just its value now
Aliases: trend display for rate and direction · process control display
What it is
A trend chart makes time an explicit axis so an operator can judge which direction a variable is heading, how fast it is moving, whether it is oscillating, and how the process responded after a control action was issued. It fills in a layer of information that a current value or an alarm state cannot provide on its own — a single number or an alarm lamp never tells you whether things are getting better or worse right now, or how quickly, and that judgment only comes from a continuous historical trace.
Why it happens
Without a trend chart, judging a rate of change requires the operator to hold several readings in working memory across time and subtract them mentally — slow and error-prone, especially while watching several variables at once. A trend chart externalizes that comparison onto the graphic itself: slope reads directly as speed of change, an inflection reads directly as a reversal, and the operator gets the answer by scanning a shape instead of computing it in their head. There is an easily overlooked premise here, though: the slope and inflection on screen are not the raw data itself but the product of three layers of processing — sampling interval, smoothing, and pixel resolution. Too long a sampling interval flattens a real fast fluctuation out of existence; smoothing suppresses noise but also delays when a genuine inflection becomes visible; pixel resolution sets the smallest slope change the eye can still distinguish. If these three layers are not disclosed, what the operator reads as "the trend" is actually a product of the processing pipeline, not the process itself, and that gap is exactly where a wrong judgment hides.
Where it stops holding
The same real change, plotted at a different aspect ratio, can look several times steeper or flatter — automatic axis scaling, a common default in charting tools, stretches the range to fill the canvas and can visually inflate a trivial fluctuation into an apparently significant change on its own. A trend chart shows association — two variables moving together — not which one causes the other, and reading synchronized movement as causation is the most common misreading. Very fast transients shorter than the sampling period, and events that are not fundamentally sequential in time, are also poorly served by this representation and call for a different view.
Applying it
When comparing two interface designs on whether operators judge trends better with one or the other, lock both to the exact same aspect ratio first — otherwise a measured improvement in "judgment accuracy" may simply be because one version happened to draw the curve steeper, not because the design is actually better. This is the most common trap in trend-chart usability testing. Co-display the chart with the current value, unit, the currently selected time window, and event markers on the same screen, and surface sampling gaps and smoothing artifacts rather than quietly smoothing them away. Test with three real process behavior patterns — slow drift, periodic oscillation, and step response — to check judgment accuracy for direction, rate, and inflection separately.