9 Hidden Plotly Tricks Every Data Scientist Needs to Know
Why Your Current Visualizations Aren’t Telling the Full Story
When it comes to data visualization, Plotly is one of my go-to tools. It’s intuitive, interactive, and offers endless possibilities. But there are several hidden techniques in Plotly that many people don’t fully explore — these can take your visualizations from good to extraordinary. I will cover 9 “hidden” or advanced techniques or “hidden” techniques that people don’t talk about enough in Plotly. For accessibility, I will use the famous UCI Machine Learning Repository datasets. I will definitely explain to you why these are “hidden”, the value of using these charts, and the insights. Along the way, I’ll explain why these tricks are “hidden,” what value they bring, and how they unlock insights that simpler visuals just can’t deliver.
1. Custom Pairwise Correlation Matrix with Colors
Why It’s Hidden:
Why did I include the correlation matrix as the first one? Every data analyst or scientist uses correlation to visualize the relationships in the data. Most data professionals visualize correlations using default heatmaps but skip the step of tailoring them to highlight meaningful relationships. Adding annotations and custom color scales brings a layer of clarity often overlooked.
What it Does:
I computed the correlation matrix using the Wine Quality dataset and created a color-coded heatmap with annotations. This highlights which features are most closely related — like alcohol and quality.
How it Adds Value:
A correlation matrix reveals feature relationships at a glance. With color coding and annotations, stakeholders can immediately see which variables matter most, like alcohol’s strong correlation with wine quality.
import pandas as pd
import plotly.express as px
from ucimlrepo import fetch_ucirepo# Fetch Wine Quality dataset
wine_data = fetch_ucirepo(name='Wine Quality')
df = pd.DataFrame(wine_data.data.features, columns=wine_data.metadata.features)# Compute correlation matrix
corr = df.corr()# Plot correlation heatmap
fig = px.imshow(corr, text_auto=True, color_continuous_scale='Viridis', title='Wine Quality Feature Correlation Matrix')
fig.show()
2. Dynamic Data Highlights
Why It’s Hidden:
We need to use conditional logic to highlight data across multiple columns more. Most users color points by a single variable but miss opportunities to highlight nuanced patterns.
What it Does:
We return to the Wine Quality dataset used in the first example. Here, I highlighted wines with high alcohol and low pH. These are what make a wine a premium wine, so I chose to highlight them. This technique of highlighting Dynamic Data makes it easy to spot them.
How it Adds Value:
Like Excel’s conditional formatting, this dynamic data highlight guides attention to critical data points. By visually separating “normal” wines from outliers, this visualization simplifies analysis for winemakers or quality controllers.
import pandas as pd
import plotly.express as px
from ucimlrepo import fetch_ucirepo# Fetch Wine Quality dataset
wine_data = fetch_ucirepo(name='Wine Quality')
df = pd.DataFrame(wine_data.data.features, columns=wine_data.metadata.features)
df['quality'] = wine_data.data.targets# Highlight outliers
df['highlight'] = df['quality'].apply(lambda x: 'High Quality' if x > 7 else 'Normal')# Scatter plot with highlights
fig = px.scatter(df, x='alcohol', y='pH', color='highlight', title='Wine Quality: Alcohol vs. pH with Highlights')
fig.show()
3. Density Contours for Class Distribution
Why It’s Hidden:
A technique to make your scatter plots more appealing: Density contours! They do, however, require precise parameter tuning. This is to avoid overwhelming the consumer of the data with excessive lines. Many Data professionals skip them altogether in favor of simpler scatter plots. Why make our lives complex when we can do the bare minimum right? (I’m being sarcastic, of course, Haha!)
What it Does:
To show density contours in use, I’ve used it back in the Iris dataset to show where species overlap or cluster.
How it Adds Value:
Contours add a layer of depth by visually quantifying density. For classification tasks, they’re a quick way to evaluate separability between classes.
# Fetch Iris dataset
iris_data = fetch_ucirepo(name='Iris')
df = pd.DataFrame(iris_data.data.features, columns=iris_data.metadata.features)
df['species'] = iris_data.data.targets# Density contour plot
fig = px.density_contour(df, x='sepal length', y='sepal width', color='species', title='Density Contour of Sepal Dimensions by Species')
fig.update_traces(contours_coloring='fill', contours_showlabels=True)
fig.show()
4. Faceted Histogram by Class
Why It’s Hidden:
Faceted histograms are often skipped because they require categorical variables and careful interpretation of the data. If you want to look at subgroups, these are an amazing tool to have in your arsenal!
What it Does:
I chose the Car Evaluation dataset in the UCI ML repository for this. Here, I used facets to compare buying price distribution by class.
How it Adds Value:
Facets enable car buyers to see exactly how a class of the car can contribute or correlate to its buying.
import pandas as pd
import plotly.express as px
from ucimlrepo import fetch_ucirepo
# Fetch Car Evaluation dataset
car_data = fetch_ucirepo(name='Car Evaluation')
df = pd.DataFrame(car_data.data.features, columns=car_data.metadata.features)
df['class'] = car_data.data.targets
# Faceted histogram
fig = px.histogram(df, x='buying', color='class', facet_col='class', title='Car Evaluation: Buying Price Distribution by Class')
fig.show()
5. Adding Threshold Lines
Why It’s Hidden:
Want to make sure your data has decision boundaries? Use Threshold lines! Threshold lines are rarely used because they require extra customization. However, they’re a simple, effective way to emphasize those decision boundaries.
What it Does:
I added a line for the Blood Donation dataset to show the critical threshold of months since the last donation.
How it Adds Value:
This draws immediate attention to actionable insights, such as identifying donors due for their next visit.
success, making it easy to target interventions for failing students.
import pandas as pd
import plotly.express as px
from ucimlrepo import fetch_ucirepo
# Fetch Blood Transfusion dataset
blood_data = fetch_ucirepo(id=176)
df = pd.DataFrame(blood_data.data.features, columns=blood_data.metadata.features)
df['donated'] = blood_data.data.targets
print(df.columns)
# Scatter plot with threshold line
fig = px.scatter(df, x='Recency', y='Frequency', color='donated', title='Blood Donation: Recency vs. Frequency')
fig.add_shape(type='line', x0=6, x1=6, y0=0, y1=df['Frequency'].max(), line=dict(color='Red', dash='dash'))
fig.show()
6. Custom Annotations
Why It’s Hidden:
Annotations are your data labels. They require manual placement, you need to add them manually. That means extra work for you, the data scientist. You will prefer skipping this because it means more work for you. What we need to understand is that for our end users who are consuming, this could save so much more time should you label the data or annotate it. You earn more recognition in your company as a result, so yeah, that is a super win for you, if there ever was one! So, use these correctly because they transform visuals into storytelling tools.
What it Does:
For this trick, I chose the Car Evaluation dataset again in the UCI ML repository. Here, the annotations help the cheapest car options stand out. This gives a lot of clarity to the consumer with their life’s big decision of buying the right car.
How it Adds Value:
Annotations can provide context to your data and help narrate a story, thereby helping end users interpret key insights without needing a data scientist to narrate to your audience explicitly.
# Fetch Car Evaluation dataset
car_data = fetch_ucirepo(name='Car Evaluation')
df = pd.DataFrame(car_data.data.features, columns=car_data.metadata.features)
df['class'] = car_data.data.targets# Scatter plot with annotations
fig = px.scatter(df, x='buying', y='maint', color='class', title='Car Evaluation with Annotations')
fig.add_annotation(x='low', y='low', text='Cheapest Option', showarrow=True, arrowhead=2)
fig.show()
7. 3D Scatter with Class Labels
Why It’s Hidden:
This technique may be complex for the average data consumer. 3D scatter plots can be confusing, so they are kept away. If you execute them well, they help reveal some never-before-seen patterns in data that you may have missed while looking at a 2D view.
What it Does:
I plotted sepal and petal dimensions in the Iris dataset in 3D, showing clear class separability.
How it Adds Value:
Your data will not have just one feature and one target variable. You will not use the area you live in to determine your house’s price. It will depend on the number of bedrooms, bathrooms, sq ft, and neighborhood/city, among many other factors. Because the real world is not 2D.
3D, so we need to ensure we don’t miss any relationships that a 2D visual will miss. We need to look at how different independent features relate to each other in addition to how they relate to the target variable, which in the case above was housing price.
# Fetch Iris dataset
iris_data = fetch_ucirepo(name='Iris')
df = pd.DataFrame(iris_data.data.features, columns=iris_data.metadata.features)
df['species'] = iris_data.data.targets# 3D scatter plot
fig = px.scatter_3d(df, x='sepal length', y='sepal width', z='petal length', color='species', title='3D Scatter Plot of Iris Dataset')
fig.show()
8. Animated Visualizations
Why It’s Hidden:
When I first came across PowerPoint, I remember I was in middle school or maybe high school, and someone was showing me some really cool animations in the slides they were showing. However, my poor, naive self didn’t know what it was, but it looked and sounded really cool. I was very fascinated by it, and I believe if you are a millennial like me or older, you can also appreciate this.
Animations in Plotly are more complex than fun PowerPoint animations. But they are more fun. Human beings are, in general, more visual. I believe in today’s day and age, we are more familiar with technology than we were 20 years ago or even 10 years ago. I think we can understand animations in Plotly because they help us visualize temporal trends — the direction in which data can change over time.
What it Does:
Here, I chose the Energy Efficiency dataset. I created an animation showing how compactness and surface area influence heating and cooling over time.
How it Adds Value:
The animations to your plotly visuals help reveal very dynamic patterns in your data. This dataset I chose helps identify key design variables for energy-efficient buildings.
import pandas as pd
import plotly.express as px
from ucimlrepo import fetch_ucirepo# Fetch Energy Efficiency dataset
energy_data = fetch_ucirepo(name='Energy Efficiency')
df = pd.DataFrame(energy_data.data.features, columns=energy_data.metadata.features)# Rename columns to meaningful names based on metadata
df.rename(columns={
'X1': 'Relative Compactness',
'X2': 'Surface Area',
'X3': 'Wall Area',
'X4': 'Roof Area',
'X5': 'Overall Height',
'X6': 'Orientation',
'X7': 'Glazing Area',
'X8': 'Glazing Area Distribution',
'y1': 'Heating Load',
'y2': 'Cooling Load'
}, inplace=True)# Add a time component (e.g., sequential order of data points)
df['time_step'] = range(len(df))# Animated scatter plot using available features
fig = px.scatter(
df,
x='Relative Compactness',
y='Surface Area',
animation_frame='time_step',
size='Wall Area',
color='Orientation',
title='Energy Efficiency Animation Over Time'
)# Customize plot for better insights
fig.update_layout(
xaxis_title='Relative Compactness',
yaxis_title='Surface Area',
legend_title='Orientation'
)
fig.show()9. Custom Tooltips with Extra Data
Why It’s Hidden:
Want to make your charts or visuals deeper? Try customizing the tooltips. It is very often skipped, but don’t worry, it’s a non-intrusive way to add depth to visuals.
What it Does:
I included age, education level, and income class in the Adult Income dataset in the tooltip.
How it Adds Value:
Customing your tooltips can make data exploration more interactive for your end users. It provides detailed information about your visuals and, as an added bonus, doesn’t clutter your chart if done right.
# Fetch Adult Income dataset
income_data = fetch_ucirepo(name='Adult')
df = pd.DataFrame(income_data.data.features, columns=income_data.metadata.features)
df['income'] = income_data.data.targets# Custom tooltips
fig = px.scatter(df, x='age', y='education-num', color='income', title='Custom Tooltip Example')
fig.update_traces(hovertemplate='Age: %{x}<br>Education: %{y}<br>Income: %{marker.color}')
fig.show()Conclusion
In these 9 techniques, I discussed how you can unlock hidden potential in Plotly. I hope you are using this already, and if not, you are now equipped to use them Today! Why should you wait to turn your basic charts into powerful storytelling tools? If you are like me, you likely noticed how these techniques were not just producing visually appealing charts but also functional, interactive, and designed to draw actionable insights. They give you an opportunity to go deeper into your data. Beyond the surface-level understanding of the data. It powers you to dive deep into your data, Truly!
If you’re ready to step up your visualization game, try these — and let me know where you end up using them in the comments. It is always fascinating to see how creative we can get with our problem-solving, especially regarding our data.