Line chart

Best for

Showing how one or a few numbers change over a continuous span of time, such as daily visits or monthly revenue.

Avoid when

The x-axis holds unordered categories (use bars), or you have more than about five series that all matter equally.

Data you need

A time or date column, one numeric column per series, and values recorded at regular intervals.

A line chart plots values in time order and connects them, so the reader sees the shape of a change: rising, falling, flat, seasonal, or sudden. Use it whenever the x-axis is time and there are enough points for a trend to exist, from a dozen monthly totals to a year of daily visits.

The line itself implies continuity. It tells the reader that something happened between the dots, which is true for temperature or follower counts and false for, say, sales by product. Most line chart mistakes come from forgetting what the line promises.

When to use a line chart

Lines are the default for time series. Website sessions per day, a stock price, weekly signups, rainfall by month, a team's points over a season. If you have one value per period and you care about the direction of travel, draw a line.

They also handle comparison well, as long as the number of series stays small. Two or three lines on shared axes let the reader see which grew faster and where they crossed. A bar chart is better when there are only a handful of periods and the exact amount in each one matters more than the trend, such as revenue for four quarters.

Avoid lines for categories with no order. Connecting "Shoes", "Bags" and "Hats" with a line invents a slope between products, and the slope means nothing.

Highlight one line, gray the rest

Put six series on one chart, give each a color, and readers stop reading the data and start decoding the legend. The chart below has a story (Tampa is pulling away), but on the left you have to find the right shade of gray-blue to see it.

The fix is cheap. Draw every series in a light neutral color, then draw the one you want to talk about in a strong color on top, and label it at the end of the line. The gray lines still give context ("Tampa is doing better than all the others"), while the colored line carries the message. If different readers care about different series, make one small chart per series with the others in gray behind it.

Avoid dual axes

A chart with one y-axis on the left and another on the right lets you put revenue in dollars and conversion rate in percent on the same plot. It looks efficient. The problem is that both scales are arbitrary, so the chart maker decides where the lines sit relative to each other. Move one axis and the lines cross somewhere else, or not at all.

People treat crossings and gaps as meaningful, and with dual axes they never are. Better options:

• Index both series to 100 at a common starting point and plot them on one axis, as on the right above.
• Stack two charts vertically with a shared x-axis, each with its own y-axis.
• Plot the ratio itself (revenue per dollar of ad spend) if that is the real question.

Show missing data as missing

Real time series have holes. A sensor goes offline, a tracking script breaks for a weekend, a store closes for renovation. If your charting tool draws a straight line across the gap, it fills those days with made-up values that look exactly like real ones.

Break the line where data is missing, and say why if you know. A light shaded band with a short label ("tracking outage") is enough. Never fill a gap with zeros unless the true value was zero. A drop to zero on a traffic chart reads as "nobody visited", which is a different claim from "we don't know".

Irregular intervals are a related trap. If you have readings for January, February and June, spacing them evenly on the x-axis squeezes four months into one step. Use a real date axis so the spacing matches the calendar.

Choosing the y-axis range

Unlike bars, lines do not need a zero baseline. The reader judges slope and position, not length, so a y-axis from 90 to 130 is fine when all the values sit in that range. Starting at zero there would flatten every movement into a near-straight line.

That freedom can be abused. Zoom in far enough and a 0.5% wobble looks like a cliff. A reasonable test: would someone who knows this metric agree the visible movement is about the size it should feel? For a share price, a 3% daily move is news. For body temperature, a 3% rise is a fever. Let the domain set the scale, and when you zoom in, keep the axis labels visible so readers can check.

Rule of thumb: label lines at their right end instead of using a legend. The eye is already there when it reaches the end of the line.

Too many points, too much noise

Daily data over several years turns into a jagged band where the trend hides behind day-to-day noise. Two common fixes:

• Aggregate to weekly or monthly totals when the daily detail does not matter.
• Draw a rolling average (7-day for web traffic with weekly cycles, 28-day for slower metrics) as a solid line, with the raw values in a light color behind it.

Say what you did in the subtitle ("7-day average"), because a smoothed line hides single-day spikes. If one of those spikes is the story, such as a viral post or an outage, annotate it on the chart. The labels and annotations guide has examples, and the follower growth guide applies the same ideas to social media data.

Common mistakes

A finished line chart usually has fewer elements than the first draft: one strong line, a few quiet ones, labels at the ends, and a title that states what changed.