To explain a graph in a video, never show the finished chart first. Reveal it in three steps: the axes (what is measured, in what units, over what range), then the data, drawn in the order you talk about it, then the one point the graph proves, stated in words with a number. Describe the overall trend before any detail, quote at least two values with units, and name any anomaly. Then say why the pattern happens, which is a separate step from describing it. A viewer who sees a full chart at once reads it alone, at their own speed, and stops listening to you.

Why a finished graph fails on screen

Put a completed line chart on screen and the viewer does three things in about a second: looks for the highest point, looks at the right-hand end, and tries to read the axis labels. None of those happen in the order you are narrating. By the time you say "on the vertical axis we have", they have already decided what the graph means, sometimes wrongly.

A graph built piece by piece avoids this. The viewer only sees what you are talking about, so their eyes and your words stay together. It is the same principle as a map in a video, where every label arrives on the word that names it; how to explain a map in a video covers that in detail.

Pick the chart that answers the question

Before revealing anything, write the question the graph answers in one line. Then check you have the right kind of graph for it.

If the honest answer to "what does this prove" is "nothing in particular, it is interesting", the graph belongs in the description, not the video.

Reveal in three steps: axes, data, point

Step one: the axes. Draw the horizontal axis and say what it measures and its range. Then the vertical axis, with units. This takes one or two sentences and is the part almost everyone skips. Skipping it is how viewers end up reading a graph of rate as a graph of total.

Step two: the data. Draw the line or the bars in the order you talk about them. For a time series, draw left to right as the narration moves forward in time. For a bar chart, add bars one at a time, and say each value as it appears. Two or three values spoken aloud are enough; the rest the viewer can see.

Step three: the point. One sentence that says what the graph shows, with a number in it. "Carbon dioxide rose from about 315 to about 430 parts per million." Then, as a separate step, why. Description and explanation are different jobs, and many exam mark schemes treat them separately.

Worked example: the Keeling curve

The Keeling curve is the record of carbon dioxide in the air measured at Mauna Loa Observatory in Hawaii. Charles David Keeling of the Scripps Institution of Oceanography began the measurements in March 1958, and they now run alongside NOAA's own measurements at the same site. The data are freely published by NOAA's Global Monitoring Laboratory and by Scripps, which makes it a good graph to practise on. Here is a script for a two-minute explanation, scene by scene.

Three things in that script are worth copying for any graph.

The axes get their own scenes. Scenes 2 and 3 are the shortest and they prevent most misreadings.

Numbers come with units every time. "315 parts per million", not "315". A number without a unit on a graph is unreadable.

The detail gets its own frame. The seasonal wobble is about six parts per million in a year, tiny on a sixty-year axis. Rather than ask the viewer to squint, scene 7 cuts to a one-year graph. That is the graph equivalent of a map zoom: show where you are going, then go there.

Check the latest figures before you use them. NOAA publishes the monthly and annual values, and they change every year.

The honest axis

The vertical axis in that example starts at 300, not zero. That is a choice, and you should say so. For carbon dioxide it is reasonable: no one is asking whether the level is near zero, and starting at zero would flatten a real, important change into a gentle slope. For other graphs, a truncated axis can make a small difference look huge, which is the oldest trick in misleading charts.

Rules that keep you honest:

• Bar charts start at zero. The length of a bar is the value. Cut the bottom off and the lengths lie.
• Line graphs may start elsewhere, but say so. "Notice the scale starts at 300" costs four seconds.
• Do not change the scale between two graphs you compare. If one runs to 100 and the other to 50, put them on the same scale or say clearly that they differ.
• Keep gaps in the data visible. In late 2022 an eruption of Mauna Loa cut power to the observatory, and measurements were moved temporarily to nearby Maunakea. A drawn line that smooths over a gap without saying so hides something.

Describing a graph the way exams want

Many geography, science and economics exams ask you to "describe" a graph and then "explain" it. A video is a good place to rehearse both, because the narration forces you to say it in order.

A strong description has four parts:

1. The overall trend. "Carbon dioxide increases throughout the period."
2. Two values with units. "From about 315 ppm in 1958 to about 430 ppm in 2025."
3. The rate or shape. "The rate of increase gets faster over time."
4. Any anomaly or pattern within it. "There is a regular seasonal cycle of about 6 ppm each year."

An explanation then gives the cause of each part. Mark schemes usually do not credit an explanation when the question only says "describe", so in your script keep the two in separate scenes.

Checklist before you export

• The question the graph answers is written in one line.
• The axes are drawn and named, with units, before any data appears.
• Data appears in the order the narration mentions it.
• At least two values are spoken with units.
• Any truncated axis or change of scale is mentioned aloud.
• The point of the graph is a single sentence with a number in it.
• Every number on screen matches your source. Drafted drawings get values and years wrong; check each label.

Make it

1. 1

   Write the question and find the source data

   One line for what the graph proves, and the published source you will take values from. Note the two or three values you will say aloud.

2. 2

   Script axes, data, point as separate scenes

   One scene for each axis, then the data in the order you narrate it, then the point with a number, then the cause in its own scene.

3. 3

   Paste the script into openCanviz

   Choose Keep my wording so every number is spoken exactly as you checked it, and pick whiteboard or poster style, which draw clean axes and flat labels that build as you speak.

4. 4

   Check every value and label against the source

   Pause on each graph scene. Make sure years, units and values are right and that the line goes the right way. Fix any scene that drifted.

5. 5

   Watch it with the sound off

   If you can still tell what each axis measures and what the point is, the graph is doing its job. If not, the scene has too much on it or the labels are too small.

Common questions

How long should a graph stay on screen? As long as the narration about it, usually 20 to 60 seconds across several scenes. If one graph needs two minutes, split it into a main view and one or two zoomed details.

Can I put a real chart image in instead of drawing it? Sometimes, with the source credited. But a finished image shows everything at once, which is the problem this post is about. A drawn version you can reveal in steps teaches better.

What about two lines on one graph? Draw the first line fully and say what it shows, then draw the second. The comparison works when the viewer already understands one line. For more on contrasts, see how to explain a comparison in a video.

How do I explain a graph that changes over a long period? Mark key dates on the time axis as the line reaches them, the way a timeline builds event by event. How to explain a timeline in a video covers pacing and gaps.