Bar Plot Gallery¶
Bar plots compare values across discrete categories. They're the workhorse chart for retail analytics: sales by store, revenue by channel, market share by brand. The eye reads height (or length) directly, so they work well for ranking, comparison, and part-to-whole questions when the categories are few.
Bar plots excel at:
- Category comparison: rank products, regions, or stores against each other
- Grouped comparison: line up multiple metrics (target vs actual, this year vs last) per category
- Part-to-whole: combined with percentage labels, show what each bar contributes to the total
- Long category names: switch to horizontal bars when category labels won't fit on the x-axis
import matplotlib.pyplot as plt
import pandas as pd
from openretailscience.plots import bar
from openretailscience.plots.styles.graph_utils import set_axis_shorthand
Basic Bar Plot¶
The simplest call: pass a single value_col and the DataFrame's index becomes the category axis. Use this shape when the index already carries the categorical meaning (product names, region codes), so there's no need for a separate x_col.
ax = bar.plot(
product_sales,
value_col="sales",
eyebrow="Category performance",
title="Home & Garden tops category sales by 80K over Electronics",
subtitle="Total sales ($) by product category",
y_label="Sales ($)",
)
set_axis_shorthand(ax.yaxis)
plt.show()
Using x_col¶
When the category labels live in a column rather than the index, point x_col at that column. Common when the data comes back from a SQL query that puts the category in its own column. The DataFrame's row order is preserved, so the bars appear in the order they sit in the data.
ax = bar.plot(
store_performance,
x_col="store",
value_col="revenue",
eyebrow="Store performance",
title="Shoreditch outpaces every other London store on revenue",
subtitle="Total revenue ($) by London store location",
y_label="Revenue ($)",
)
set_axis_shorthand(ax.yaxis)
plt.show()
Multiple Value Columns (Grouped Bars)¶
Pass a list to value_col and each category gets a clustered set of bars, one per series. Use this shape when the upstream data is already in wide format (one column per series), as is common with SQL pivot output. Three or four series stays readable; more than that and the bars get too thin to compare.
ax = bar.plot(
quarterly_channel_sales,
x_col="quarter",
value_col=["Online", "Stores", "Wholesale"],
eyebrow="Channel mix",
title="Stores remain the largest channel but Online closes the gap each quarter",
subtitle="Quarterly sales ($) by channel, full year",
y_label="Sales ($)",
legend_title="Sales Channel",
move_legend_outside=True,
)
set_axis_shorthand(ax.yaxis)
plt.show()
Horizontal Bar Plot¶
Set orientation="horizontal" to lay the bars on their side. Pick this whenever category labels are long enough to crowd or rotate awkwardly on the x-axis. Bars run left-to-right, so the value axis is now the x-axis (use set_axis_shorthand(ax.xaxis) to format it).
ax = bar.plot(
category_revenue,
x_col="category",
value_col="revenue",
eyebrow="Category revenue",
title="Consumer Electronics earns roughly 3x the revenue of Books & Media",
subtitle="Total revenue ($) by product category, long category names",
x_label="Revenue ($)",
orientation="horizontal",
)
set_axis_shorthand(ax.xaxis)
plt.show()
Sorted Bar Plot¶
Pass sort_order="descending" to rank vertical bars from tallest to shortest, so the eye lands on the leader first. Use "ascending" instead when the chart's job is triage rather than celebration. (For horizontal bars the convention flips: "ascending" puts the largest bar at the top of the chart.)
ax = bar.plot(
sales_rep_data,
x_col="sales_rep",
value_col="sales_amount",
eyebrow="Sales team",
title="Top three reps generate over 60% of all sales",
subtitle="Sales ($) by sales representative, ranked highest first",
y_label="Sales Amount ($)",
sort_order="descending",
)
set_axis_shorthand(ax.yaxis)
plt.show()
Bar Plot with Absolute Data Labels¶
Pass data_label_format="absolute" to write each bar's raw value above its top edge. Useful for small charts (under ten bars) where exact numbers matter and the reader shouldn't have to estimate from the y-axis. Values are formatted in shorthand (e.g. 285K, 1.2M) and truncated to num_digits=3 significant digits by default.
ax = bar.plot(
regional_sales,
x_col="region",
value_col="sales",
eyebrow="Regional sales",
title=r"North leads with \$285K, East trails by nearly \$90K",
subtitle="Total sales ($) by region",
y_label="Sales ($)",
data_label_format="absolute",
)
set_axis_shorthand(ax.yaxis)
plt.show()
Percentage Data Labels by Series¶
Pass data_label_format="percentage_by_series" to label each bar as its share of the column's total. Pick this for part-to-whole questions like market share or category mix where the total adds up to a meaningful 100%. The companion "percentage_by_bar_group" mode does the same per category group when you're plotting grouped bars.
ax = bar.plot(
market_share_data,
x_col="brand",
value_col="market_share",
eyebrow="Market share",
title="Cadbury and Mars together hold 63% of the chocolate market",
subtitle="Share of UK chocolate market by brand (%)",
y_label="Market Share (%)",
data_label_format="percentage_by_series",
)
plt.show()
Bar Plot with Hatch Patterns¶
Set use_hatch=True to overlay a hatch pattern on each series. Useful when the chart will be printed in black and white, projected on a low-contrast screen, or needs to remain readable with colour-blind-friendly palettes. Each series gets a different hatch style, so they stay distinguishable even if the colours blur together.
ax = bar.plot(
monthly_performance,
x_col="month",
value_col=["Target Sales", "Actual Sales"],
eyebrow="Target tracking",
title="Sales beat target in March and May, missed it in the other four months",
subtitle="Monthly sales ($), target versus actual",
y_label="Sales ($)",
legend_title="Sales Type",
use_hatch=True,
move_legend_outside=True,
)
set_axis_shorthand(ax.yaxis)
plt.show()
Plotting a Pandas Series¶
A Series works directly: the function uses the index for category labels and the series name for the legend. Useful when the upstream computation (e.g. a groupby().mean()) already returns a Series and you don't want to wrap it just to plot.
ax = bar.plot(
customer_satisfaction,
value_col="Satisfaction Score",
eyebrow="Voice of customer",
title="Support and Quality lead satisfaction; Price scores lowest",
subtitle="Customer satisfaction ratings by category, 1 to 5 scale",
y_label="Rating (1-5)",
)
plt.show()
Advanced Styling Options¶
Two knobs worth knowing: source_text adds a small attribution line at the bottom of the chart for export-ready images, and move_legend_outside=True lifts the legend to the right edge so it never sits on top of the data. Both are no-cost additions on any chart that's headed for a slide or a public dashboard.
ax = bar.plot(
department_data,
x_col="department",
value_col=["Q1", "Q2", "Q3"],
eyebrow="Team performance",
title="Operations holds the top performance score in every quarter",
subtitle="Quarterly performance scores by department",
y_label="Performance Score",
legend_title="Quarter",
move_legend_outside=True,
source_text="Source: Internal Performance Dashboard",
)
plt.show()
