The concept
A chart is a tool for making a comparison visible. Every useful chart answers one question: is this bigger than that, is this rising or falling, is this related to that, how is this made up. If you cannot say which comparison a chart is showing, the chart is decoration. Data visualization is the practice of choosing the comparison that matters, encoding the numbers so the eye can read it instantly, and removing everything else.
Why it matters
Operators look at charts to make decisions: cut a product line, change a price, hire, stop a campaign. A chart that takes two minutes to decode will be skimmed and misread. A chart that shows the comparison in two seconds gets acted on. The cost of a bad chart is a bad decision made confidently.
Message first, chart second
Before opening any tool, write one sentence that states what you want the reader to take away. Examples:
- “Online orders overtook in-store orders in March.”
- “Three of our eight products generate 80% of margin.”
- “Refund rate climbs sharply when delivery takes more than five days.”
That sentence is your chart title. It also tells you the chart type: the first is a comparison over time (line), the second is a ranking with part-to-whole (sorted bar), the third is a relationship (scatter or binned bar).
Worked example
You have monthly revenue for six products over two years. The lazy approach is one line chart with six lines. Nobody can read it. Ask the question first.
| Question | Chart |
| Which products are growing? | Small multiples: six tiny line charts, same axis |
| Which product is biggest this year? | Single sorted bar chart of this year’s totals |
| Did the mix shift? | Two stacked 100% bars, last year vs this year |
Three questions, three different charts from the same data. The data did not choose the chart; the question did.
Exploratory vs explanatory
When you are exploring, make many fast, ugly charts to find the story. When you are explaining, make one clean chart that tells it. Do not show exploratory charts to decision-makers; they will see fifteen lines and no point.
Common mistakes
- Starting with “what chart types does my tool offer” instead of “what is my message”.
- Putting every available series on one chart because the data was there.
- Titles that name the data (“Revenue by Month”) instead of the finding (“Revenue fell 12% after the price change”).
- Making one chart try to answer three questions.
Action Step
Take one dataset you actually use (sales, expenses, web traffic, orders). Write three different one-sentence findings from it. For each, name the comparison type (over time, ranking, part-to-whole, relationship, distribution) and the chart you would use. Keep this sheet; you will build these three charts over the course.