Notebook – Metric Dashboard

Notebook – Metric Dashboard#

Where Your Metrics Learn to Dance 💃 (and Impress the CFO)

“Spreadsheets make people yawn. Dashboards make them clap.” 👏

Welcome to the Metric Dashboard Lab — your mission: Combine your model performance metrics, visuals, and business storytelling into one interactive dashboard that would make even your manager say,

“Wait, you built this in Python?” 🤯

🎬 Business Hook: The Boardroom Moment#

You’ve trained a sales forecast model. You have metrics. You have visuals. But in your meeting, someone asks:

“Can you show how this changes by product line and region?”

You gulp. Then smile. Because you built this lab. 😎

🧩 Your Objectives#

By the end of this lab, you’ll be able to: ✅ Build a dashboard-like notebook to compare models ✅ Display interactive visuals using Plotly ✅ Present key metrics with a business narrative ✅ Let users filter and explore model results

⚙️ Step 1. Load the Sample Data#

import pandas as pd

# Example performance data
data = {
    'Model': ['Linear Regression', 'Random Forest', 'XGBoost', 'Prophet'],
    'MAE': [250, 180, 160, 200],
    'RMSE': [300, 220, 190, 240],
    'R2': [0.82, 0.91, 0.93, 0.88],
    'Business Unit': ['Retail', 'Retail', 'Wholesale', 'Wholesale']
}
df = pd.DataFrame(data)

df

🧠 Pro Tip: Replace this with your own metrics! Try loading data from your previous chapters — regression, classification, etc.

🧮 Step 2. Compute Derived KPIs#

You can turn your metrics into business-friendly indicators.

df['Improvement_vs_Baseline'] = ((df['R2'] - df['R2'].min()) / df['R2'].min()) * 100
df['Error_Reduction_%'] = ((df['RMSE'].max() - df['RMSE']) / df['RMSE'].max()) * 100

df[['Model', 'Improvement_vs_Baseline', 'Error_Reduction_%']]

💬 “Because no one gets excited about RMSE = 190, but they do get excited about ‘error reduced by 35%.’” 📉

📈 Step 3. Visualize the Metrics (Plotly Style)#

import plotly.express as px

fig = px.bar(df, x='Model', y='R2', color='Business Unit', text='R2',
             title='📊 Model Performance by R²',
             labels={'R2': 'R² Score'})
fig.update_traces(texttemplate='%{text:.2f}', textposition='outside')
fig.update_layout(yaxis_range=[0, 1])
fig.show()

🎨 Plotly Tip: The fewer colors, the more “executive-friendly” your chart looks.

💼 Step 4. Create KPI Cards#

This is the “wow” part of your dashboard — mini metric boxes that show results at a glance.

from IPython.display import display, Markdown

for _, row in df.iterrows():
    display(Markdown(f"""
    ### 🚀 {row['Model']}
    - **R²:** {row['R2']:.2f}
    - **RMSE:** {row['RMSE']}
    - **MAE:** {row['MAE']}
    - **Error Reduction:** {row['Error_Reduction_%']:.1f}%
    """))

💬 “Congratulations — you just built mini KPI cards like a pro dashboard designer.”

🧠 Step 5. Add Interactivity (Optional but Awesome)#

Use widgets to filter your dashboard — because executives love clicking buttons.

import ipywidgets as widgets
from IPython.display import clear_output

model_dropdown = widgets.Dropdown(options=df['Model'].unique(), description='Choose Model:')
output = widgets.Output()

def show_model_metrics(change):
    with output:
        clear_output()
        model = change['new']
        row = df[df['Model'] == model].iloc[0]
        display(Markdown(f"""
        ## 📊 {model} Metrics
        - **R²:** {row['R2']:.2f}
        - **RMSE:** {row['RMSE']}
        - **MAE:** {row['MAE']}
        - **Error Reduction:** {row['Error_Reduction_%']:.1f}%
        """))

model_dropdown.observe(show_model_metrics, names='value')
display(model_dropdown, output)

🧩 “Interactivity turns a static notebook into a Netflix for data nerds.”

🧪 Step 6. Create the Business Summary#

Use Markdown, visuals, and plain language to summarize:

“The Random Forest model achieved the best tradeoff between accuracy and interpretability, improving forecast precision by 25% over the baseline. This translates into an estimated $420K reduction in overstock costs.” 💰

💬 “A good dashboard answers questions. A great one creates budget justification.” 😏

🧩 Step 7. Export or Share#

✅ Download Notebook (via toolbar) ✅ Open in Colab for team collaboration ✅ Deploy with Streamlit for interactive business apps

Remember: A dashboard is not the end — it’s the conversation starter.

🧠 Practice Tasks#

🎯 Level 1 – Warm-Up:

  • Load your own dataset from a previous exercise (e.g. regression or classification).

  • Compute RMSE, MAE, R², and display them in a Plotly bar chart.

🎯 Level 2 – Business Twist:

  • Add calculated “error reduction %” or “accuracy gain %”.

  • Annotate results with business insights.

🎯 Level 3 – Dashboard Hero:

  • Add filters for business units or time periods.

  • Include a “summary text” explaining the best model choice.

  • Style it with emojis, colors, or icons (CFOs love polish).

🧭 Recap#

Step

Skill

Output

1

Load performance data

Pandas DataFrame

2

Compute KPIs

Derived metrics

3

Visualize

Plotly charts

4

Summarize

KPI cards

5

Interact

Filters & widgets

6

Communicate

Business insights

7

Share

Dashboard ready

💬 Final Thought#

“Your dashboard isn’t just about showing numbers — it’s about telling the story of your model’s impact on business.” 📈

🔜 Next Up#

👉 In the next chapter — Core ML Models — we’ll go from measuring models to building them. Get ready to meet Regression, Classification, and their weird cousin, Clustering — and find out why every ML algorithm has trust issues with your data. 😜

# Your code here
Contents