Tables People Can Read
A table in a report keeps only the columns a reader needs, renames them in plain words, rounds its numbers, shows the top ten instead of every row, and uses df.style.format to show money as $20.00 and shares as 97.2% with the index hidden.
Journeyman · Python · Updated
After this lesson you can
- Make a table readable by keeping the columns a reader needs, renaming them, rounding, and showing the top ten
- Format money and percentages with df.style.format, and hide the index with hide(axis='index')
Key takeaways
- Keep only the columns the reader needs, and rename them in plain words with rename(columns={...}).
- Round numbers to what a reader can use, and show the top ten with sort_values and head(10), or nlargest, instead of every row.
- df.style.format shows money and percentages the way people read them, such as $20.00 and 97.2%, and hide(axis="index") drops the row labels.
- df.style changes how a table looks, not its numbers, and it needs the Jinja2 package, which nbconvert installs.
The tables pandas prints are made for you, the person writing the code: every column, names with underscores, numbers to six decimal places. In a report, a table like that makes the reader do the work. Whenever a table goes in front of someone else, a manager, a client or a class, a few steps turn it into one they can read at a glance.
The example is Pine Valley Community Center's classes, classes.csv, one row for each of its 588 class sessions from November 2025 to October 2026. Denise, the director, asks: which classes fill up, and what do they cost?
A table made for the code, not the reader
A summary with one row per class has the answer in it:
import pandas as pd
classes = pd.read_csv("classes.csv")
by_class = classes.groupby("class_name", as_index=False).agg(
category=("category", "first"),
sessions=("class_id", "count"),
capacity=("capacity", "sum"),
signed_up=("signed_up", "sum"),
fee=("fee", "first"),
)
by_class["fill"] = by_class["signed_up"] / by_class["capacity"]
by_class.head(4)
class_name category sessions capacity signed_up fee fill
0 Acrylic Painting arts 24 288 197 15.0 0.684028
1 Computer Basics technology 6 60 55 0.0 0.916667
2 Junior Chess kids 24 288 183 0.0 0.635417
3 Kids Art Club kids 48 768 684 5.0 0.890625
VS Code draws the table as a grid; this lesson shows it as pandas prints it as text. Denise has to work out that fill is the share of seats taken, that 0.684028 means about 68%, that 15.0 is dollars, and that the classes are in A to Z order, not best first. All 15 classes are in the full table, seven columns wide, and she needs two of the columns.
Fewer columns, names for readers
Keep only what answers the question, then give each column a name a person would say. rename() takes a dictionary, old name to new name:
table = by_class[["class_name", "fee", "fill"]]
table = table.rename(columns={"class_name": "Class", "fee": "Fee", "fill": "Seats filled"})
table.head(3)
Class Fee Seats filled
0 Acrylic Painting 15.0 0.684028
1 Computer Basics 0.0 0.916667
2 Junior Chess 0.0 0.635417
Rename for the report only. In your code, fill is easier to type than Seats filled, so make the renamed table at the end, right before it's shown.
Rounding and the top ten
A reader can't use six decimal places, and 15 rows is already more than most people scan. Sort the best first, keep ten, and round:
top = table.sort_values("Seats filled", ascending=False).head(10)
top["Seats filled"] = top["Seats filled"].round(3)
top
Class Fee Seats filled
11 Summer Science Camp 20.0 0.972
9 Smartphone Basics 0.0 0.942
1 Computer Basics 0.0 0.917
7 Pottery Wheel 25.0 0.914
3 Kids Art Club 5.0 0.891
5 Little Movers 5.0 0.876
10 Spin 10.0 0.858
6 Morning Yoga 8.0 0.814
13 Watercolor Basics 15.0 0.778
8 Senior Strength 0.0 0.721
Three decimal places of a share is a tenth of a percent: 0.972 is 97.2%. nlargest() is a shorter way to get the same ten rows: table.nlargest(10, "Seats filled"). With 588 sessions, or 600 rows of anything, "the top ten" is a table someone reads; all of them is a table someone skips.
The numbers down the left, 11, 9, 1, are the old row labels, the index. They mean nothing to Denise.
Money and percentages with df.style.format
Every DataFrame has a style property that makes a Styler: the same table, drawn for people. Its format() method takes a dictionary of column names and format codes, and hide(axis="index") leaves out the row labels:
top.style.format({"Fee": "${:.2f}", "Seats filled": "{:.1%}"}).hide(axis="index")
In a notebook, the cell shows this table:
| Class | Fee | Seats filled |
|---|---|---|
| Summer Science Camp | $20.00 | 97.2% |
| Smartphone Basics | $0.00 | 94.2% |
| Computer Basics | $0.00 | 91.7% |
| Pottery Wheel | $25.00 | 91.4% |
| Kids Art Club | $5.00 | 89.1% |
| Little Movers | $5.00 | 87.6% |
| Spin | $10.00 | 85.8% |
| Morning Yoga | $8.00 | 81.4% |
| Watercolor Basics | $15.00 | 77.8% |
| Senior Strength | $0.00 | 72.1% |
The format codes are Python's, the ones f-strings use. "${:.2f}" is a dollar sign, then the number with two decimal places (.2f, read "point two f"). "{:.1%}" multiplies by 100 and adds a percent sign, with one decimal place.
The Styler changes only how the table looks: top["Seats filled"] still holds 0.972, so keep computing with top and show the Styler. A Styler is drawn as a web table, so it shows in the notebook and in a web page made from it. It needs the Jinja2 package, which nbconvert installs; without it, .style stops with AttributeError: The '.style' accessor requires jinja2.
Check yourself
2 questions to try on the lesson page.
So what
A table in a report answers the reader's question and nothing else: the columns they need, in words they'd use, the top rows, and numbers shown as money and percentages.