Topic 9 — Data visualization and dashboards

Goal: Choose the right chart and build dashboards that make answers obvious.

Lesson 9.1 — Why a chart beats a table

Marcus, Perch's Head of Marketing, drops by Nadia's desk on a Tuesday. "How are sign-ups doing?" She had pulled the numbers that morning: twelve rows, one per month, daily sign-up totals in a neat column. She turns her laptop toward him. He squints, scans, and after a few seconds asks the only question the table left open: "So... up or down?"

That pause is the problem. The answer was sitting right there, but Marcus had to work to find it. Nadia did the analysis; the table made him redo it in his head.

So she drops the same twelve numbers into a line chart. Now the shape does the talking: sign-ups climbed steadily until March, then flattened. Marcus gets it before she says a word.

That is the entire reason the job exists. A table of numbers hides patterns; a good chart reveals them. The eye spots a trend, a spike, or an outlier in a picture far faster than in rows of figures. Visualization is how an analyst makes an insight obvious to the people who didn't do the analysis.

If you found the answer but your audience still has to dig for it, you only finished half the work.

Nadia came to Perch from retail operations, where she'd built weekly sales reports by hand for years, so she knew this in her bones. A wall of inventory figures told a store manager nothing; the same figures as a chart told them which shelf was bleeding money. The tool is new; the instinct is old.

Lesson 9.2 — One chart, one point

A week later Nadia builds her first "real" chart for leadership and goes all in. Three metrics on one graph, two axes, a splash of color for each product line, a 3-D tilt because it looked impressive in the software. She's proud of it.

Priya, her mentor, looks at it for a long moment. "What's the one thing you want them to walk away knowing?"

Nadia opens her mouth and realizes she has three answers. Which means the chart has none.

That's the golden rule, and the one most beginners break: a chart should make one main point clear at a glance. If a viewer has to study it, the chart has failed. A graph carrying three messages delivers zero, because the reader doesn't know which to read.

So clarity beats decoration, every time. The 3-D tilt distorts the bars so you can't compare their real heights. The rainbow of colors makes the eye hunt for meaning that isn't there. Nadia strips all of it out and rebuilds: one product line, plain bars, a title that states the point. Anyone can read it in two seconds.

The test Priya taught her: show it to someone for three seconds, take it away, and ask what they saw. If they name your point, the chart works. If they shrug, it's back to the drawing board, no matter how pretty it looked.

Lesson 9.3 — Matching the chart to the question

Most bad charts aren't dishonest. They're just the wrong type for the question being asked. Nadia keeps a short cheat sheet taped to her monitor, because four shapes cover almost everything she's ever asked:

• Comparing categories — sofa sales vs. desk sales vs. shelving — use a bar chart. It's the reliable, hard-to-misread default; when in doubt, reach for it first.
• Change over time — repeat purchases month by month — use a line chart. Time always goes on the horizontal axis, left to right, so the trend reads the way we read.
• The relationship between two numbers — does a bigger discount actually lift order size? — use a scatter plot. Each dot is one order; the cloud's shape tells you whether the two move together.
• One key figure — yesterday's revenue — just show the big number. No chart adds anything; a single bold figure says it instantly.

Then there's the pie chart, the one everybody reaches for and almost everybody overuses. A pie chart shows parts of a whole, and it only works for two or three slices. The human eye is bad at judging angles, so the moment you have eight wedges in similar sizes, nobody can tell which is bigger. A bar chart, sorted tallest to shortest, answers the same "which is biggest" question in a glance.

Marcus once asked Nadia to "break down where new customers come from" across nine marketing channels. Her instinct said pie; she caught herself and built a sorted bar chart instead. He saw the top three channels immediately. Nine pie slices would have been a guessing game.

When she's unsure, Nadia defaults to a bar or a line, almost always a safe, honest, readable choice.

Lesson 9.4 — Honest axes and labels

Here's where a chart stops being merely unclear and starts being a lie.

Nadia builds a bar chart of sign-ups for two months: April at 1,000, May at 1,050. A real 5% bump. But when she lets the software auto-scale the axis, it starts the bottom at 990. Now the May bar towers over April's, looking three times as tall. The picture screams "huge jump." The data whispers "small one."

That's the single most common way charts deceive, and it happens by accident as often as on purpose. The rule is simple and non-negotiable: start the value axis of a bar chart at zero. A bar's whole job is to let you compare lengths; chop off the bottom and the length lies. (Line charts that track a trend can sometimes zoom in, but bars, never.)

Dana, the VP of Operations, would spot a trick like that in a second, and once she catches you exaggerating, she stops trusting every chart you send. Honest scaling is more than ethics; it's how an analyst keeps the credibility the whole job runs on.

The second rule is quieter but just as important: label everything. Title, both axes, units. "Revenue (USD), by month, 2026" tells the reader what they're looking at. A chart with bare numbers and no labels is an unanswered question, and the reader will either guess wrong or come ask you, which defeats the point of making the chart at all.

An unlabeled chart is an unanswered question. Title it, scale it from zero, name the units — then it can stand on its own.

Lesson 9.5 — Dashboards and BI tools

By her third month, Nadia notices a pattern: every Monday, the same four people ask her the same four questions. Marcus wants sign-ups. Dana wants repeat-purchase rate. Each request means a fresh query and a fresh chart, every single week.

That repetition is what a dashboard fixes. A dashboard is a collection of charts and key numbers gathered on one screen, so people can monitor the metrics that matter without coming to you each week. Build it once, and it refreshes itself.

Dashboards live in a BI tool — short for business intelligence — and the three names Nadia hears most are Tableau, Power BI, and Looker. Each connects directly to the company's data warehouse (the central store from Topic 3), so the numbers stay live. Tableau is loved for polished, flexible visuals; Power BI is the budget-friendly choice for companies already on Microsoft; Looker defines each metric once in code so every dashboard agrees on what "repeat purchase" means. Perch runs Power BI, but the skill carries across all three: pick a metric, choose a chart, point it at the warehouse table Tom maintains.

The hard part of a dashboard is restraint, not the building. Nadia's first version crammed in fifty charts because every chart felt useful. Nobody opened it twice. Priya's fix: a good dashboard is focused — the few metrics that actually drive decisions, the most important one at the top, every panel clearly labeled.

She rebuilt it with five numbers: sign-ups, repeat-purchase rate, average order value, revenue, and refund rate. The five things leadership watches, and nothing else. That version got pinned to Dana's browser. A clean dashboard with the right five numbers becomes indispensable; a cluttered one with fifty gets ignored. The focused one also frees Nadia to do the deeper analysis only she can do.

Worked example — The "are repeat customers growing?" dashboard

Dana sends a one-line email: "I need to know if our repeat-purchase push is working. Can I see it without pinging you every week?" That last clause is the real ask — she wants a dashboard, not a one-off chart.

Nadia starts with the question, the way Priya drilled into her, before touching the software. Is the share of customers who buy a second time going up over time? That's a change-over-time question, so the centerpiece is a line chart: months on the horizontal axis, repeat-purchase rate on the vertical, scaled from zero so a small real gain doesn't get dressed up as a big one. Title: "Repeat-purchase rate, % of customers, by month." Marcus's old "up or down?" can never happen here.

At the top she pins the single number Dana cares about most: this month's repeat-purchase rate, big and bold, with last month beside it for context. One glance, done.

Below the line chart she adds a bar chart comparing repeat rates across the three product lines — sofas, desks, shelving — sorted tallest to shortest, so it's instantly clear which category is pulling repeats and which is lagging. She resists the urge to add a fourth and fifth panel; Dana asked one question, and the dashboard answers exactly that one.

She builds it in Power BI, connected to the warehouse table Tom owns, so it refreshes on its own every morning. She checks his repeat-purchase flag, confirms it's clean, labels every axis, and shares the link.

The next Monday, Dana doesn't email. She opens the dashboard, sees the rate ticking up, and forwards it to the CEO. Nadia turned a weekly interruption into a self-serve answer — the whole craft of this topic in one screen.