The chart of change over time
A line graph connects values in time order so that slopes become visible: steep segments are rapid change, flat ones are stalls, direction changes are turning points. It is the default for series that arrive over time: monthly sales, daily temperatures, weekly weights. The connecting line is the claim: it asserts the value travelled between the measured points, which is why the points must be genuinely ordered and, ideally, evenly spaced in time.
Monthly savings across half a year: Jan 42, Feb 45, Mar 51, Apr 49, May 57, Jun 63 (in dollars). The graph climbs from 42 to 63, a gain of 21, or +50% across the six months, with one visible dip in April. The slope tells the story a table of six numbers cannot say as quickly: steady progress, one bad month, recovery.
Axis discipline for time series
Zero matters less here than on bar charts (you judge position and slope, not bar length), but the axis still steers the drama: a cramped vertical range turns a wobble into a crisis. Choose a range that fits the data's real variability, label it plainly, and keep it constant across charts you intend to be compared. Dates belong in order, at their true spacing when gaps are irregular: a missing month should leave a visible gap, not a quietly tightened axis.
Smoothing away the story. Rolling averages and trend curves have their place, but drawing only the smoothed line hides volatility that may be the most important feature (a savings habit that collapses every April). Plot the raw points; add smoothing as an extra, lighter line if you must.
Line or bars for time data?
Bars emphasize individual periods (this month vs that month); a line emphasizes the trajectory. Reports on performance over time usually want the trajectory. If the audience must look up exact period values, a table beats both, with the line as the summary picture.
Frequently asked questions
What goes on the horizontal axis of a line graph?
Time, in order: months, days, years. The vertical axis carries the measured value. Points are connected in time order, which is what makes slope readable as rate of change.
Does a line graph axis have to start at zero?
Not strictly: you read positions and slopes, not lengths. But the range chooses the drama, so pick it honestly, keep it labelled, and hold it fixed across charts meant to be compared.
How do I handle missing months?
Leave a gap in the line or mark the missing point explicitly. Quietly connecting across a missing period fabricates a journey the data never took.
Can I put two series on one line graph?
Yes, with clearly distinguished lines and a legend. Two is comfortable; past three or four, small multiples (one small chart per series, shared axes) read better.