The final challenge: three questions and a chart to fix
Go back to the three questions from the beginning with new charts, fix a chart with three problems on your own, and choose what comes next.
Free course. Your Maestro, an AI, .
On the first page you answered three questions before learning anything. Here they are again, with other charts and other numbers. Answer them without going back to look: the idea is for you to notice how much your way of reading has changed.
After that comes the challenge: a chart with three problems that you'll fix yourself, without step-by-step instructions.
First question
The annual report has this ticket chart. Read it with the four moves before you answer: what's measured, axes and baseline, scale, what to compare.
Check yourself: According to this chart, how many times Tonalá's tickets does Chapultepec have?
Second question
Check yourself: Another chain reports that the average WhatsApp order is 300 pesos at both of its stores. The manager concludes: "Both stores have the same kind of WhatsApp customer." What do you tell her?
Third question
Check yourself: Your boss asks you: "Make me a chart of which branch grew fastest for its size." What do you do first?
Compare with what you answered on the first page. If it came easier now, it's not luck: you already look at where the axis starts, you already distrust an average on its own, and you already start with the question.
Does one of the three still make you hesitate? Tell your Maestro which one and why you chose what you chose.
The challenge: fix it yourself
The regional manager asks: "Which branch sold the most in 2025, and which the least?" Someone sent him the chart below. It has three problems for that question. This time there are no step-by-step instructions: read the chart with the four moves, name each problem and fix it in the JSON.
You're done when the chart answers the manager's question at a glance and its length tells the truth about each amount. It takes three changes. If you get stuck, ask your Maestro for a hint, not the answer.
Fix the chart for the manager's question: "Which branch sold the most in 2025, and which the least?" It takes three changes in the JSON. Hints, if you need them: is the question about change over time or about comparing? In what order does the reader find the biggest one? Where d
Your turn
Fix the chart for the manager's question: "Which branch sold the most in 2025, and which the least?" It takes three changes in the JSON. Hints, if you need them: is the question about change over time or about comparing? In what order does the reader find the biggest one? Where does the axis start?
You're done when you see bars, from the branch that sold the most to the one that sold the least, that start at zero.
Edit the JSON; the preview updates when you pause.
How this maps to Altair
Altair builds this same JSON: chart.to_json() prints it. "mark": "point" is alt.Chart(df).mark_point(). Each encoding channel is an argument to .encode(), such as x=alt.X("field:Q", scale=alt.Scale(type="log")) or color="field:N". Each transform entry is a call such as .transform_filter("datum.year == 2007").
Explain in two or three sentences what was wrong with the original chart for the manager's question, and why your version answers it better.
Check yourself: Explain in two or three sentences what was wrong with the original chart for the manager's question, and why your version answers it better.
Which of the course's traps do you think had fooled you before? Where did you see it: at work, in the news, in a report of your own?
Reflect: Which of the course's traps do you think had fooled you before? Where did you see it: at work, in the news, in a report of your own?
Only you and your Maestro read this. It's kept on your account and deleted with it.
What you can do now
When you started, a chart was something the program suggested to you. Now you have an order for thinking it through:
- Choose by the question. Comparing goes with sorted bars; change over time, with a line; how values are spread, with a histogram; part of a whole, with 100% stacked bars or a pie with few slices; the relationship between two things, with a scatter plot.
- Read any chart in four moves: what's measured, axes and baseline, scale, what to compare.
- Recognize the traps: the cut axis on bars, the pie with many slices, the dual axis, the areas that exaggerate and the average that hides the spread.
That's what anyone who reads charts to make decisions needs. If you want to keep going, it depends on what you want to do now.
What's next
If you want an AI assistant to make charts with you, to your taste and checked the way you'd check them, the next step is Data Viz Co-Pilot. You make your first chart with an agent like Claude or ChatGPT, learn a four-question critique, write the style guide your agent works from, and finish by delivering a real chart.
If you want to write your own charts in Python, the code you touched in this course's exercises is Altair. Effective Data Visualization with Altair teaches it from the beginning: twelve lessons that start with what makes a visualization effective and end with a portfolio piece built on your own data.
And if what you make are dashboards with many charts together, the next question is no longer which chart to use, but what goes on the dashboard and for whom. What you learned here still holds for every chart on the dashboard.
Not sure which one is right for you? Tell your Maestro what you do with charts at work and it'll help you decide.