A chart is a sentence, not a decoration
Most workplace charts fail in the same quiet way. They are not wrong. They are just not saying anything. Twelve coloured lines cross each other, the legend is in alphabetical order, the title says Sales Data, and everyone in the meeting nods politely and moves on.
The fix is one sentence long. Before you choose a chart type, write down what you want the reader to conclude. "The South region has been falling since March" is a conclusion. "Sales by region" is a filing label. Once you have the sentence, the chart almost picks itself, and the AI can pick it for you.
In plain English
- Series:
- One set of values being drawn. One line, or one colour of bar. Three regions means three series.
- Axis:
- The two edges of the chart. Usually categories or dates along the bottom, values up the side.
- Legend:
- The little key that says which colour means what.
- Data label:
- The actual number printed on or beside a bar or point, so nobody has to squint at the axis.
- Baseline:
- Where the value axis starts. Bars must start at zero or the comparison lies.
Match the chart to the question
You do not need chart theory. You need four patterns, and this is all of them.
Comparing things that are not in time order, such as regions, products or people, use a bar chart, and sort it by size rather than alphabetically. Sorting is the single cheapest improvement available to any chart.
Showing something over time, use a line chart, with time running left to right and no gaps in the dates.
Showing parts of a whole, use a bar chart anyway. Pie charts work only when there are two or three slices and the differences are obvious. A twelve-slice pie is a colour swatch, not an argument.
Showing whether two things move together, such as discount and order size, use a scatter plot. This is the only one of the four that most people never reach for and often should.
I want a reader to conclude: [write the sentence here].
My data has columns [list them] and covers [time period or categories].
The audience is [my manager / the board / the ops team]
and they will look at this for about [10] seconds.
Which chart type says that fastest, and why?
Give me one recommendation, not a list of options.
Tell me what to put on each axis, how to sort it,
and a title that states the conclusion rather than the topic.
Checkpoint
Write the conclusion you want the reader to reach before you choose the chart. The sentence decides the chart type, the sort order and the title.
Titles that do the work
Your chart title is the most-read text in any deck and it is almost always wasted. Sales by Region tells the reader what they can already see from the axis labels. Sales fell in two of five regions, both in the South, tells them what to think.
โ Weak prompt
Prompt
Make a chart of my sales data.
Output
A line chart with nine coloured series, an alphabetical legend and the title Sales Data.
No conclusion, so no point. Nine lines is a plate of spaghetti and the reader gives up in two seconds.
โ Good prompt
Prompt
I want my manager to see that two regions caused nearly all of our Q2 drop. I have quarterly revenue for five regions. Recommend one chart, tell me the sort order, and write a title that states the finding. Keep it readable when printed in black and white.
Output
A sorted bar chart of change by region, biggest fall first, titled with the actual finding, using shading rather than colour to highlight the two regions that matter.
Conclusion, audience and constraint all stated. The reader gets the point before they have finished reading the title.
Let AI critique the chart you already made
This is the underused half of the skill. Most modern chat tools accept an image, so take a screenshot of the chart you have already built and ask for the criticism you would never get from a colleague.
Here is a screenshot of a chart I made. The point I want it to make is:
[the sentence].
The audience is [who] and they have about [10] seconds.
Tell me:
- What conclusion a reader would actually reach in 10 seconds.
- Anything misleading: a value axis that does not start at zero,
uneven time gaps, or a scale that exaggerates the change.
- Three specific changes, ranked by how much each one helps.
- A better title, stated as the finding.
Be blunt. Do not tell me it looks good.
The first answer is the valuable one. If the conclusion a stranger reaches in ten seconds is not the conclusion you intended, the chart has failed, however handsome it looks.
The most common way a workplace chart misleads by accident is a value axis that starts at 90 rather than zero. It turns a 2 percent wobble into a cliff edge. On a bar chart, start at zero. Always.
Practical rules that survive contact with real meetings
Cut the series. If you have nine lines, ask which three carry the message and group the rest as Other. Nobody has ever complained that a chart was too easy to read.
Label the numbers you want people to quote. If a figure is going to be repeated out loud, print it on the chart rather than making people estimate it against gridlines.
Highlight rather than colour everything. One bar in a strong colour and the rest in grey directs the eye better than a rainbow ever will, and it survives being printed in black and white.
Say what the data does not cover. If the last month is partial, put it in the title or leave it out. A half-finished month plotted next to full ones creates a dramatic downward cliff that is pure fiction.
Once you know what you want, getting the software to do it is the easy half, as long as you say which software you are in.
I am in [Excel 365 / older desktop Excel / Google Sheets].
My data is in columns [A and D], rows [2 to 40].
I want a [bar] chart, sorted [largest to smallest],
with data labels on, gridlines off, one bar highlighted
and everything else in grey.
Give me the exact menu steps in order. No formulas, just clicks.
If a step differs between versions, say which version you are describing.
Ask Gemini or ChatGPT for the exact menu steps in your version of Excel or Google Sheets, and mention which one you are using. Following the clicks is much quicker than hunting through ribbon tabs for the option to sort a bar chart.
๐ Quiz
Question 1 of 4What should you decide before choosing a chart type?