At most one email a week. Unfollow anytime with one click.
Transcript
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Show transcript
Introduction
Hi, welcome back to the data analysis session. We've found insights in Sunset Coffee's data. Now we make them visible. We'll choose the right type of chart, ask Claude to build it as an Artifact, improve it so it can be read in five seconds, and avoid misleading charts.
Step 1: Choosing a chart type
Imagine Laras showing a report to Dimas between serving customers. Dimas only has time for a five-second glance. A good chart has to work in that little time. The key is one sentence: one chart, one message. Before making a chart, decide what message you want to deliver. The chart type is decided by its purpose. To compare several things, use a bar chart. For example, revenue by product. To see change over time, use a line chart, or bars per period. For example, revenue by week. For parts of a whole, a pie chart is fine, as long as there are few slices, four at most. Beyond that, the slices are hard to compare. And if readers need exact numbers, such as price details, a table is often better than any chart. If you're unsure, just ask Claude: I want to show this message to this reader, which chart fits best? It will explain why. Let's try asking Claude. Laras wants to show that repeat customers brought in sixty-one percent of revenue, and asks which chart fits and why. Claude suggests one horizontal bar split in two, because there are only two parts. A pie chart also works. The interesting part is the reason: the message is about share, so a table doesn't fit as well. By asking for the reason, you also learn to choose charts yourself. Over time you won't need to ask anymore.
Step 2: A first chart as an Artifact
Let's start. Laras asks for a bar chart of revenue by product from the October orders file. The chart appears right away as an Artifact in the right panel. The numbers are correct, the same as the analysis in the previous episode. Claude also offers a Code tab, but we don't need to open it. But take a five-second glance. What do you take away? The bars aren't in order, the title only names the content, and the unit isn't clear: is eleven point one six millions or thousands? A first chart like this is normal. The numbers are right, but the message doesn't come through yet. Now let's improve it together with Claude. There are four improvements with the most impact. First, sort the bars from largest. The reader's eye finds the winner right away. Second, a title that carries the message. Not revenue by product, but a sentence that says what it means, such as one-liter iced coffee brings in the most revenue. Third, number labels complete with units, so nobody has to guess millions or thousands. Fourth, one highlight color. The bar you want to talk about gets a standout color, and the rest are gray. These four improvements work for any chart, whether it's made in Claude or in a spreadsheet. Make it a habit to check them every time you make a chart.
Step 3: Improving the chart with Claude
Laras gives the first three improvements in one clear request. She can ask for the highlight color later, when the chart is used in a presentation. Claude updates the Artifact to version two. Now the title delivers the message right away, the bars are sorted from largest, and every bar has a label in millions of rupiah. Claude also adds a small note below: the data source and the period. This note answers a question that comes up often: where is the data from, and which month? Take another five-second glance. The message comes through: the one-liter iced coffee is the revenue champion. Dimas doesn't need to read the numbers one by one. So you don't start from scratch every month, save this chart request template. The data section names the file and the period. The goal section holds the main message in one sentence. This becomes the title. The format section contains all four improvements at once. And the limits keep the chart from misleading: the number axis starts at zero, no three-dimensional effects, and there's always a source note. We'll cover why in the last step.
Step 4: Time charts and charts for meetings
Now a time chart. Laras asks for a chart of weekly revenue throughout October. At a glance, week five looks like it crashed. But Claude gives an important warning: week five has only three days, the twenty-ninth to the thirty-first. This is a very common trap. Laras asks for a fairer method: the average revenue per day in each week. The result flips. The average per day is actually highest in week five, about one point five million per day. The same chart can give the opposite impression, depending on how we calculate. A chart for an office meeting starts from the readers' question. Dewi explains: the board wants to know which stationery categories grew and which fell compared with last quarter. Claude suggests a difference bar chart: one bar per category, different colors for up and down, sorted from the biggest gain, plus one concluding sentence. A chart that directly answers the board's question saves meeting time. Notice that Dewi starts with the board's question, not by asking for a particular chart. Let Claude suggest the shape, and then you judge it. A chart also needs to fit where it will be shown. If it's sent via WhatsApp, use big text and few bars, because it will be read on a phone screen. If it's shown in a meeting, use one chart per slide with the conclusion in the title. Don't cram four charts onto one screen. If it's printed, make sure it's still readable in black and white. Set the highlighted bar apart with a pattern or a label, not just color. Just say where it will be shown, and Claude will adjust.
Step 5: Avoiding misleading charts
Charts can mislead by accident. The first trap: the number axis doesn't start at zero. A small gap, say from eighty-five to eighty-eight, can look like a gigantic jump. Second, three-dimensional effects. They look fancy, but the bar sizes become hard to compare. Third, too many colors. The reader's eye doesn't know where to look. Fourth, a pie chart with many slices. Beyond four parts, the slices are hard to tell apart. Fifth, unequal periods, like the three-day week earlier. Ads often use these five traps to make numbers look more impressive. We don't want to fool our readers, so avoid them all, including when Claude makes the chart. Before sharing a chart, check four things. First, the title delivers the main message. Second, the number labels and units are clear. Third, there's a note on the source and period below the chart. Fourth, it passes the five-second test. Show the chart to someone else for five seconds, then ask what they took away. Laras tries it on Dimas, and Dimas answers right away: the liter-size coffee is the biggest. It passes. If it fails, don't blame the reader. Fix the title or the highlight, then test again. One chart that's understood is worth more than five confusing ones.
Done!
We've chosen chart types by purpose, built and improved charts as Artifacts, read time charts carefully, and avoided five chart traps. Charts show what has already happened. In the next episode we look ahead: building simple scenarios and estimates, such as what happens if prices go up. See you there!
Practice exercises
Try these on your own after watching to lock in what you learned.
Write one main message from your data in one sentence, then make it the chart title
Ask Claude for a chart, then improve it: sort it, add number labels, highlight one bar
Show the chart to a coworker for five seconds and ask what they took away
Ask the class material
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
My notes
Notes are saved with the current video time and are only visible to you.
Questions & answers
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.
Your instructor
Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
53 video classes1,290 episodes
We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.
