Data Analysis · Tutorial 8 of 10
Chart Chooser Reference: which chart for which question
By the FUTEAC Team · 21 November 2026 · 11 min read · Beginner
In short
Pick the chart by the question, not by how it looks. Comparing values takes a bar chart, change over time takes a line chart, a share of a whole takes a pie chart only with five slices or fewer, a spread of values takes a histogram, and a link between two measures takes a scatter plot.
- The question picks the chart. "Which is biggest?" and "How did it change?" are different questions, and they need different charts. A chart that answers the wrong question is decoration.
- A bar chart is the default. When in doubt, use bars. They are the easiest chart to read, they work for comparisons, rankings and shares, and they are the hardest to get wrong.
- A pie chart has a strict limit. Five slices or fewer, and only when the shares add up to a whole. Six or more slices, and the small ones blur into slivers nobody can compare.
- A line chart needs time on the horizontal axis. Categories such as product names belong on bars. A line joining categories implies a trend that does not exist.
- Every chart needs a title that states the finding. "Monthly sales" is a label. "Sales rose 62.5% from January to June" is a finding, and it tells the reader what to notice.
- This reference feeds the Spreadsheets for Data Analysis course (/course/spreadsheets-for-data-analysis) and the Data Visualisation (/skill/data-visualisation) and Statistics Basics (/skill/statistics-basics) skills.
Before you start
How to use this reference
Start with the question you want the chart to answer, find it in the chooser table below, and read the matching section for the worked example. The two example datasets run through the whole reference, so every number can be checked:
Dataset A, monthly sales (thousand PKR): Jan 120, Feb 150, Mar 135, Apr 180, May 210, Jun 195. Total 990, mean 165.0. Sales rose 62.5% from January to June, and the biggest single jump was April, up 33.3% on March.
Dataset B, sales by category (thousand PKR): Shirts 420, Trousers 310, Shoes 260, Accessories 150. Total 1,140. Shares: Shirts 36.8%, Trousers 27.2%, Shoes 22.8%, Accessories 13.2%.
Choosing a chart is like choosing a vehicle for a trip: a motorbike, a car and a truck all move, but only one fits the load. The question is the load. Pick the chart that carries it, not the one that looks newest.
The chooser table
| Your question | Use this chart | Example in this reference | Watch out for |
|---|---|---|---|
| Which is biggest, and by how much? | Bar chart | Categories, Dataset B | Sorting matters: biggest first reads fastest |
| How did it change over time? | Line chart | Monthly sales, Dataset A | Needs real time on the horizontal axis, not categories |
| What share of the whole is each part? | Pie chart, 5 slices or fewer | Categories, Dataset B | More than 5 slices: use a bar chart instead |
| How are values spread out? | Histogram | Delivery times, 15 orders | Too few bins hides the shape, too many bins hides it too |
| Do two measures move together? | Scatter plot | Ad spend vs sales | Correlation is not proof that one causes the other |
| What is the one number that matters? | Single KPI figure, no chart | Total sales 990 | A lone number needs context: last period, or a target |
Comparing values: bar chart
Use it when: the question is "which is biggest?" or "rank these". Do not use it when: the values are parts of a whole and the whole itself matters, that is pie or stacked bar territory.
Dataset B as bars, sorted biggest first: Shirts 420, Trousers 310, Shoes 260, Accessories 150. The bars run horizontally, longest at the top, so the ranking reads top to bottom in one glance. Shirts outsell Accessories by 270, and the gap is visible without reading a single number.
Two rules make a bar chart honest. First, the value axis starts at zero. Second, sort the bars unless the order itself means something, such as months in a year. An unsorted bar chart is a list with decoration; a sorted one is an answer.
In Excel: select the category and value columns, Insert, Bar Chart. In Google Sheets: Insert, Chart, Bar chart. Both do the same thing, and both let you sort the data first.
Change over time: line chart
Use it when: the horizontal axis is time and the question is "how did it change?" Do not use it when: the horizontal axis is categories such as product names. A line joining Shirts to Trousers to Shoes implies a trend across products, and no such trend exists.
Dataset A as a line: Jan 120, Feb 150, Mar 135, Apr 180, May 210, Jun 195. The line climbs overall, dips in March, and falls in June after the May peak of 210. The mean is 165.0, so February, March and January sit below the average line, and April, May and June sit above it. The growth from January to June is 62.5%.
Add the mean as a second, flat line if the reader needs to see which months beat the average. One extra line is fine. Three extra lines, a target, last year, and a forecast, turn the chart into a puzzle.
In Excel: select the months and sales, Insert, Line Chart. Keep the markers on: a line without markers hides the fact that there are only six real data points.
Part of a whole: pie chart, and when not to use one
Use it when: there are five slices or fewer, the shares add up to 100%, and the question is "what share?". Do not use it when: there are six or more parts, the parts do not add up to a whole, or the reader needs exact comparisons.
Dataset B as a pie: Shirts 36.8%, Trousers 27.2%, Shoes 22.8%, Accessories 13.2%. Four slices, and the two biggest are easy to compare. But notice what the pie hides: Trousers at 27.2% and Shoes at 22.8% look close, and the eye cannot tell a 4.4 point gap on a circle. A bar chart shows that gap in a glance. That is the honest limit of a pie: it shows the shape of the whole, not the gaps between parts.
A pie chart is like slicing a cake: four slices are easy to serve and compare, twelve slices are crumbs. If your data has more than five parts, group the small ones into "Other" or switch to a bar chart. A donut chart follows the same rules as a pie, with the same limit.
In Excel: select the categories and values, Insert, Pie Chart. Label each slice with its percentage, not just its name: a slice without a number asks the reader to guess.
Spread of values: histogram
Use it when: the question is "how are the values spread out?" or "what is typical, and what is unusual?" Do not use it when: you have fewer than about ten values, or the question is about one value at a time.
Fifteen delivery times in minutes: 21, 22, 23, 24, 25, 26, 27, 28, 28, 29, 30, 31, 32, 33, 35. The mean is 27.6, the fastest is 21 and the slowest is 35. A histogram with bins of 5 minutes shows most orders landing between 25 and 30 minutes, with a thin tail at each end. That shape is the finding: deliveries are usually 25 to 30 minutes, and 35 is unusual, not normal.
A histogram is not a bar chart. In a bar chart each bar is a category; in a histogram each bar is a range of values, and the bars touch because the ranges are continuous. Bin size is the judgement call: bins of 1 minute would scatter fifteen orders into noise, and one bin of 15 minutes would hide everything. Start with 5 to 10 bins and adjust until the shape is stable.
In Excel: select the values, Insert, Histogram. Excel picks the bins for you, and you can change the bin width from the axis options.
Link between two measures: scatter plot
Use it when: the question is "do these two measures move together?" Do not use it when: one of the measures is time, that is a line chart, or when you only have a handful of points.
Six months of ad spend (thousand PKR) against sales: spend 5, 8, 6, 12, 15, 11 against sales 120, 150, 135, 180, 210, 195. Plot spend on the horizontal axis and sales on the vertical axis, and the points climb from bottom left to top right. The correlation is 0.97, which means the two move together very closely in this data.
Read that carefully. A correlation of 0.97 says the two rise together, not that ads caused the sales. June may have had a holiday, or a new product launch. A scatter plot shows the link; explaining the link needs the business context, and that context is not in the chart. Never write "ads caused sales" under a scatter plot.
In Excel: select the two numeric columns, Insert, Scatter. Do not add a line joining the points: the points are the data, and a joining line implies an order that is not there.
The one number: KPI figure
Use it when: the whole finding is one number. Do not use it when: the number alone misleads, which is most of the time.
Total sales for the six months: 990 thousand PKR. On its own, that number says nothing: 990 against what? Put the context next to it, in the same glance: last period, a target, or the mean. "Total sales 990, up 62.5% on the January run rate" is a finding. "Total sales 990" is a number on a wall.
KPI figures belong on dashboards, one per tile, with the comparison beside them. The dashboards tutorial later in this queue builds on this reference.
The wrong turn: the axis that lies
Here is the mistake that ruins more bar charts than any other. Take Dataset B, Shirts 420, Trousers 310, Shoes 260, Accessories 150, and draw the bars with the value axis starting at 300 instead of zero. Shirts now looks three times as tall as Trousers, and Accessories nearly vanishes. The numbers on the axis are still correct, and the chart is still a lie: the eye compares bar lengths, not axis labels.
Start every bar axis at zero. The one exception is a line chart, where starting above zero is acceptable because the eye follows the slope, not the height. If a bar chart needs a truncated axis to show a difference, the difference is too small for a bar chart: use a table, or a dot plot, and say so.
Chart anatomy checklist
Run this list before any chart goes into a report. It takes one minute and catches the errors that make readers distrust the rest of your work.
- Title states the finding. "Sales rose 62.5% from January to June", not "Monthly sales".
- Axes are labelled with units. Thousand PKR, minutes, percent. A number without a unit is a guess.
- The source is noted under the chart. "Shop sales records, Jan to Jun 2026" tells the reader where the numbers came from and how fresh they are.
- Colours mean something, or nothing. One colour for all bars, or one colour for the bar that matters and grey for the rest. Random colours suggest differences that do not exist.
- No 3D, no shadows, no decoration. A 3D bar distorts the height it is meant to show. Flat bars, thin gridlines, and white space do the work.
Try it yourself
Take Dataset A and Dataset B from the top of this reference and build all six chart types in Excel or Google Sheets: bar, line, pie, histogram, scatter, and one KPI figure. Then deliberately build two wrong ones: a pie chart with the four categories split into eight slices, and a bar chart with the axis starting at 300. Put the right and wrong versions side by side and write one sentence under each pair explaining what the wrong one hides. That pair is the fastest way to train your eye, and it is the same comparison you will make every time you choose a chart at work.
Continue in the free course: Spreadsheets for Data Analysis →
Common mistakes
- Using a pie chart for six or more parts. The small slices blur into slivers, and nobody can compare them. Group the small parts into "Other" or switch to bars.
- Drawing a line across categories. A line implies change over time or order. Categories have no order, so a line across them invents a trend.
- Starting a bar axis above zero. The eye reads bar length, not axis labels. A truncated axis exaggerates every gap, and a careful reader will stop trusting your charts.
- Charting raw counts when the question is about shares. If the question is "what share of sales?", chart the percentages, not the raw totals. The raw totals answer a different question.
- Adding a chart because the page looks empty. A chart that answers no question is decoration, and decoration trains readers to skip your charts. If no question needs a chart, a sentence with the number is better.
Frequently asked questions
How many charts should a one-page report have?
Two or three at most. One chart per question, and a page usually has two or three questions. More than that, and the reader skims instead of reading. The one-page summary lesson in the Spreadsheets for Data Analysis course works through exactly this.
Can I use a pie chart if I have six categories?
Group the smallest ones into "Other" until five slices remain, or use a bar chart. Six slices is where a pie stops being readable, and the bar chart you switch to will show the gaps between slices better anyway.
What is the difference between a bar chart and a histogram?
A bar chart compares categories, and its bars have gaps. A histogram shows how values are spread across ranges, and its bars touch because the ranges are continuous. The data decides which one you need, not the look.
Should the line chart or the bar chart show monthly sales?
Both can, but they answer different questions. A bar chart answers "which month was biggest?" A line chart answers "how did sales move?" If the trend is the point, use the line. If the ranking is the point, use the bars.
Do I need a charting tool, or is Excel enough?
Excel or Google Sheets is enough for every chart in this reference, and for most analyst work. Move to a dedicated tool only when the chart needs to update itself from a database, or when you are building a dashboard that several people will use.
Follow the Data Analyst Roadmap
Turn raw data into decisions with spreadsheets, SQL, Python and dashboards. Free, step by step, at your pace.
View the Roadmap