storeNum Year Week p1sales p2sales p1price p2price p1prom p2prom country
1 101 1 1 127 106 2.29 2.29 0 0 US
2 101 1 2 137 105 2.49 2.49 0 0 US
3 101 1 3 156 97 2.99 2.99 1 0 US
4 101 1 4 117 106 2.99 3.19 0 0 US
5 101 1 5 138 100 2.49 2.59 0 1 US
6 101 1 6 115 127 2.79 2.49 0 0 USData Visualization
MKT 566 · Fall 2026
Same numbers, four different stories
All four datasets share the same means, variances, correlation (r = 0.82), and regression line (y = 3 + 0.5x). Always plot your data first.
An (almost) perfect example
Source: The Groupon Effect on Yelp Ratings: A Root Cause Analysis (Byers, Mitzenmacher, and Zervas 2012), Figure 1a.
Bar chart
Bar chart: order your bars
Alphabetical order is by default; sorted order requires fct_reorder() in R.
Pareto chart
Sorted bars + cumulative share: identifies the “vital few” driving most of the total.
The 80/20 rule in marketing
Revenue is almost always concentrated in a small share of customers. This motivates segmentation, loyalty programs, and CLV analysis (all later in this course).
Treemap
Space-filling overview: area ∝ count. Good for hierarchies (category → brand → SKU).
Pie chart
Use sparingly: humans compare lengths far better than angles. With more than 3–4 slices, a sorted bar chart is almost always clearer.
Waterfall chart
Each bar starts where the previous ended: shows incremental contributions to a total (common in revenue-bridge and budget decks).
Line chart
Area chart
Stacked areas show composition over time, but only the bottom series sits on a flat baseline, so read the others with care.
Bar + line combo
Two scales on one chart: level (bars, left axis) vs. running total (line, right axis).
Histogram
Density plot
A smoothed histogram: highlights modes, hides bin-choice artifacts.
Box plot
Median, quartiles, whiskers, outliers: a five-number summary per group.
Violin plot
Distributions in the wild: online ratings
Review ratings are typically J-shaped: delighted and furious customers review, the indifferent middle stays silent. The mean rating hides this: plot the distribution. Much more on reviews and reputation later in the course.
Scatter plot
P1 and P2 sales are negatively associated. What marketing story could explain that?
Bubble chart
A third variable mapped to point size: do high-sales weeks coincide with low prices?
Ad spend and diminishing returns
The response curve is concave: doubling spend does not double sales. A scatter plus a smoother is often the first diagnostic for budget-allocation questions.
Choropleth map
Heatmap
Two categorical axes + a color-encoded value: spot which stores skew high or low.
The marketing classic: cohort retention
Rows are acquisition cohorts, columns are months since signup: the standard view for subscription businesses. Are newer cohorts retaining better or worse?
Grouping: everything in one panel
Faceting: one panel per group
5. Annotate and highlight key insights
- Direct labels with
geom_text()orggrepelfor calling out peaks, thresholds, or outliers - Annotations (
annotate(),geom_vline()/geom_hline()) to mark events: product launches, policy changes, seasonal holidays
How to mislead with a chart
Same data, different y-axis. Truncated axes exaggerate differences, a favorite trick of dashboards and ad claims. When you see a bar chart, check where the axis starts.
Step 0: the default plot
Step 1: label the axes
Step 2: larger fonts, more breathing room
ggplot(data = store.df) +
geom_point(
mapping = aes(x = p1price, y = p1sales)
) +
labs(
x = "\nPrice of P1 ($)",
y = "Weekly Sales of P1 (units)\n"
) +
theme(
axis.title.x = element_text(size = 14),
axis.title.y = element_text(size = 14),
axis.text.x = element_text(size = 12),
axis.text.y = element_text(size = 12)
)
Step 3: a cleaner theme
ggplot(data = store.df) +
geom_point(
mapping = aes(x = p1price, y = p1sales)
) +
labs(
x = "\nPrice of P1 ($)",
y = "Weekly Sales of P1 (units)\n"
) +
theme(
axis.title.x = element_text(size = 14),
axis.title.y = element_text(size = 14),
axis.text.x = element_text(size = 12),
axis.text.y = element_text(size = 12)
) +
theme_few()
Step 4: fix the overplotting
ggplot(data = store.df) +
geom_jitter(
mapping = aes(x = p1price, y = p1sales),
width = 0.02, alpha = 0.4
) +
labs(
x = "\nPrice of P1 ($)",
y = "Weekly Sales of P1 (units)\n"
) +
theme(
axis.title.x = element_text(size = 14),
axis.title.y = element_text(size = 14),
axis.text.x = element_text(size = 12),
axis.text.y = element_text(size = 12)
) +
theme_few()
geom_jitter() adds a little horizontal noise; alpha makes overlaps visible.
Step 5: add the trend
ggplot(data = store.df) +
geom_jitter(
mapping = aes(x = p1price, y = p1sales),
width = 0.02, alpha = 0.4
) +
labs(
x = "\nPrice of P1 ($)",
y = "Weekly Sales of P1 (units)\n"
) +
theme(
axis.title.x = element_text(size = 14),
axis.title.y = element_text(size = 14),
axis.text.x = element_text(size = 12),
axis.text.y = element_text(size = 12)
) +
geom_smooth(
mapping = aes(x = p1price, y = p1sales),
method = "lm", color = "blue"
) +
theme_few()
Step 6: add a title
ggplot(data = store.df) +
geom_jitter(
mapping = aes(x = p1price, y = p1sales),
width = 0.02, alpha = 0.4
) +
labs(
x = "\nPrice of P1 ($)",
y = "Weekly Sales of P1 (units)\n",
title = "Weekly P1 Sales Decline with Price"
) +
theme(
axis.title.x = element_text(size = 14),
axis.title.y = element_text(size = 14),
axis.text.x = element_text(size = 12),
axis.text.y = element_text(size = 12)
) +
geom_smooth(
mapping = aes(x = p1price, y = p1sales),
method = "lm", color = "blue"
) +
theme_few()
A downward-sloping demand curve. How much sales fall when price rises is the price elasticity; we will estimate it with regression in week 4.