Data Visualization

What is Data Visualization for Data Analysts?

Data Visualization is how Data Analysts turn raw numbers into clear, trustworthy visuals that drive decisions. You will choose the right chart for the question, design readable visuals (labels, scales, colors), avoid misleading patterns, and assemble dashboard-ready views for recurring monitoring.

Who this is for

• Aspiring and junior Data Analysts who present findings to stakeholders.
• Analysts shifting from ad-hoc tables to reproducible charts and dashboards.
• Anyone who needs to communicate data clearly and ethically.

Prerequisites

• Basic descriptive statistics (mean, median, percent, rates).
• Comfort with spreadsheets or Python/pandas for simple aggregations.
• Understanding of the business question you want to answer.

Learning path

1. Foundations: chart selection, clear labeling, axes/scales, color & accessibility.
2. Core charts: bar (categories), line (trends), histogram/density (distributions), scatter (relationships), stacked/100% stacked (composition), small multiples (compare subgroups).
3. Interactive basics: tooltips, filters, and highlights for exploration.
4. Quality & ethics: avoid misleading visuals; document assumptions.
5. Dashboard-ready: consistent styles, clear titles, minimal clutter.

Milestones

• Pick the right chart for a question without second-guessing.
• Label clearly: units, time frames, and definitions are obvious.
• Use linear vs log scales intentionally; set sensible axis ranges.
• Apply color responsibly and accessibly; emphasize what matters.
• Compose small multiples for clean subgroup comparisons.
• Ship a tidy, consistent dashboard that answers one core question.

Worked examples (with code and reasoning)

# Python (pandas + seaborn)
import pandas as pd, seaborn as sns
import matplotlib.pyplot as plt

sales = pd.DataFrame({
    'category':['A','B','C','D','E'],
    'revenue':[120000, 95000, 43000, 41000, 15000]
})
# Sort for readability
sales = sales.sort_values('revenue', ascending=False)

sns.set(style='whitegrid')
ax = sns.barplot(data=sales, x='revenue', y='category', color='#5B8FF9')
ax.set_title('Revenue by Category — Q4 (USD)')
ax.set_xlabel('Revenue (USD)')
ax.set_ylabel('Category')
for i, v in enumerate(sales['revenue']):
    ax.text(v, i, f'  ${v/1000:.1f}k', va='center')
plt.show()

import pandas as pd, seaborn as sns
import matplotlib.pyplot as plt

weeks = pd.date_range('2024-01-01', periods=20, freq='W')
df = pd.DataFrame({'week': weeks, 'signups':[90,110,105,150,130,160,170,165,180,175,190,210,205,220,235,230,245,260,255,270]})
df['ma_3'] = df['signups'].rolling(3).mean()

sns.lineplot(data=df, x='week', y='signups', label='Weekly')
sns.lineplot(data=df, x='week', y='ma_3', label='3-week MA')
plt.title('Weekly Signups — with 3-week Moving Average')
plt.ylabel('Signups (count)'); plt.xlabel('Week')
plt.show()

import numpy as np, pandas as pd, seaborn as sns
import matplotlib.pyplot as plt

np.random.seed(0)
mins = np.clip(np.random.normal(42, 8, 400), 15, 80)

sns.histplot(mins, bins=20, kde=True, color='#5AD8A6')
plt.title('Delivery Time Distribution (minutes)')
plt.xlabel('Minutes'); plt.ylabel('Orders')
plt.show()

import numpy as np, pandas as pd, seaborn as sns
import matplotlib.pyplot as plt

np.random.seed(1)
sessions = np.random.randint(1, 20, 100)
rev_per_user = sessions * 3.5 + np.random.normal(0,7,100)

sns.regplot(x=sessions, y=rev_per_user, scatter_kws={'alpha':0.6}, line_kws={'color':'#F6BD16'})
plt.title('Sessions vs Revenue per User')
plt.xlabel('Sessions (30 days)'); plt.ylabel('Revenue per user (USD)')
plt.show()

import pandas as pd
import matplotlib.pyplot as plt

months = ['Jan','Feb','Mar','Apr']
data = pd.DataFrame({
    'month': months,
    'Paid':[120,150,160,170],
    'Organic':[80,90,95,100],
    'Referral':[30,40,45,50]
})

# Stacked (absolute)
fig, ax = plt.subplots(1,2, figsize=(10,4))
ax[0].stackplot(months, data['Paid'], data['Organic'], data['Referral'], labels=['Paid','Organic','Referral'])
ax[0].legend(loc='upper left'); ax[0].set_title('Stacked Signups (Absolute)')

# 100% Stacked (relative)
row_sums = data[['Paid','Organic','Referral']].sum(axis=1)
perc = data[['Paid','Organic','Referral']].div(row_sums, axis=0)
ax[1].stackplot(months, perc['Paid'], perc['Organic'], perc['Referral'], labels=['Paid','Organic','Referral'])
ax[1].set_title('Stacked Signups (Percent of Total)')
plt.tight_layout(); plt.show()

import numpy as np, pandas as pd, seaborn as sns
import matplotlib.pyplot as plt

vals = np.concatenate([np.random.lognormal(mean=1.5, sigma=1.0, size=300), [5000,10000]])
sns.histplot(vals, bins=30)
plt.xscale('log')
plt.title('Skewed Metric on Log Scale')
plt.xlabel('Value (log scale)'); plt.ylabel('Count')
plt.show()

Drills — quick practice

• [ ] Replace a pie chart with a ranked bar chart and add data labels.
• [ ] Redesign one chart with fewer gridlines and clearer units.
• [ ] Create a small multiple of line charts: same y-axis range for fair comparison.
• [ ] Build a histogram with 3 different bin counts; pick and justify the best.
• [ ] Recolor a chart to be colorblind-safe and add an emphasis color for the key series.
• [ ] Add a subtitle that states time frame and sample size (e.g., n=1,204).

Common mistakes and debugging tips

• Using the wrong chart: Bars for categories; lines for continuous time; scatter for pairs. If it feels forced, revisit the question.
• Ambiguous labels: Always include units, date range, and definitions. Put them in title/subtitle/axis labels.
• Truncated axes on bars: Bars should start at zero. If not possible, switch to a dot/line chart and disclose ranges.
• Over-coloring: Use one emphasis color; keep the rest neutral-gray or a muted palette.
• Inconsistent scales across panels: For small multiples, keep axes consistent unless explicitly stated.
• Unreliable interactivity: Tooltip values rounded inconsistently confuse users. Standardize formats.
• Hiding uncertainty: For small samples, annotate cautions and consider error bars or ranges.

Mini project — KPI story dashboard

Build a 1-page dashboard that answers: “Are we acquiring and retaining users efficiently this quarter?”

1. Top tile: Current quarter KPI (e.g., signups) vs target; small delta indicator.
2. Trend: Weekly line chart with 4-week moving average; annotate launches.
3. Acquisition mix: 100% stacked bar by channel per month.
4. Quality: Histogram of time-to-first-action; mark median.
5. Efficiency: Scatter of sessions vs revenue per user (color by segment).
6. Small multiples: Line charts by region with consistent axes.
7. Interactivity: Filter by region/segment; tooltip shows exact values.

Subskills

• Chart Selection Principles: Map question to chart: compare, trend, distribution, relationship, composition.
• Clear Labeling and Annotations: Titles with time frame, units, n; concise callouts for events.
• Working with Axes and Scales: Zero-baselines for bars, sensible ranges, log scale when justified.
• Color and Accessibility Basics: Colorblind-safe palettes, use color sparingly to emphasize.
• Comparing Categories Bar Charts: Sorted, aligned bars with data labels if short.
• Trends Line Charts: Continuous time, smoothing (MA), annotate events.
• Distributions Histograms Density Plots: Bin selection, KDE overlays, median markers.
• Relationships Scatter Plots: Trend lines, grouping, discuss outliers and correlation.
• Composition Stacked Charts: Absolute vs 100% stacked; when to avoid too many segments.
• Small Multiples: Same scales; faceted charts for clean subgroup comparison.
• Interactive Charts Basics: Tooltips, filters, highlights that add clarity.
• Avoiding Misleading Visuals: No cherry-picking; disclose scale, range, and filters.
• Dashboard Ready Charts: Consistent style, tight titles, minimal clutter, aligned to the KPI.

Next steps

• Pick one weekly report and replace every table with the most suitable chart.
• Create a style guide: fonts, colors, number formats, and axis rules.
• Automate your data prep so charts refresh reliably.

Practical projects

• Product funnel snapshot: Bar + 100% stacked to compare drop-off by channel.
• Marketing mix over time: Line small multiples by region; annotate campaigns.
• Quality dashboard: Distribution of response times; scatter of load vs errors; weekly trends.