app.py

"""
Lead.AI Customer Predictor — Live Demo
Predicts customer churn and purchase likelihood for small businesses.
Visit https://www.lead-ai.us for a custom deployment.
"""

import gradio as gr
import numpy as np
from sklearn.ensemble import GradientBoostingClassifier
import warnings
warnings.filterwarnings("ignore")

# ── Train demo models ─────────────────────────────────────────────────────────
np.random.seed(99)
n = 3000

tenure      = np.random.randint(1, 60, n)
purchases   = np.random.randint(0, 50, n)
avg_spend   = np.random.uniform(10, 800, n)
support_tix = np.random.randint(0, 15, n)
days_since  = np.random.randint(1, 180, n)
email_opens = np.random.uniform(0, 1, n)

X = np.column_stack([tenure, purchases, avg_spend, support_tix, days_since, email_opens])

# Churn: high support + long since last purchase + low email engagement
churn_score = (support_tix * 0.3 + days_since * 0.01 - purchases * 0.05
               - email_opens * 0.5 - tenure * 0.005)
y_churn     = (churn_score > np.percentile(churn_score, 65)).astype(int)

# Purchase: recent + engaged + history
buy_score   = (purchases * 0.4 + email_opens * 0.3 - days_since * 0.008 + avg_spend * 0.001)
y_buy       = (buy_score > np.percentile(buy_score, 50)).astype(int)

churn_model = GradientBoostingClassifier(n_estimators=100, random_state=42)
churn_model.fit(X, y_churn)

buy_model = GradientBoostingClassifier(n_estimators=100, random_state=42)
buy_model.fit(X, y_buy)

SEGMENTS = {
    (False, True):  ("🌟 High-Value Active",    "This customer is engaged and ready to buy. Prioritize for upsell offers."),
    (False, False): ("✅ Stable Retained",       "Low churn risk but not currently primed to buy. Nurture with content."),
    (True,  True):  ("⚡ At-Risk, Still Buying", "Buying but showing churn signals. Act now with a retention offer."),
    (True,  False): ("🚨 High Churn Risk",       "This customer is disengaging. Send a personal win-back message today."),
}

def predict_customer(tenure, purchases, avg_spend, support_tickets,
                     days_since_purchase, email_open_rate):
    feats = np.array([[tenure, purchases, avg_spend, support_tickets,
                       days_since_purchase, email_open_rate]])

    churn_prob = churn_model.predict_proba(feats)[0][1]
    buy_prob   = buy_model.predict_proba(feats)[0][1]

    churn_pct  = round(churn_prob * 100, 1)
    buy_pct    = round(buy_prob   * 100, 1)

    is_churn = churn_pct >= 50
    is_buy   = buy_pct   >= 50

    segment, action = SEGMENTS[(is_churn, is_buy)]

    churn_bar = f"{'█' * int(churn_pct // 5)}{'░' * (20 - int(churn_pct // 5))} {churn_pct}%"
    buy_bar   = f"{'█' * int(buy_pct   // 5)}{'░' * (20 - int(buy_pct   // 5))} {buy_pct}%"

    report  = f"## {segment}\n\n"
    report += f"**Recommended Action:** {action}\n\n"
    report += "---\n\n"
    report += "### Prediction Scores\n\n"
    report += f"**Churn Risk:**    `{churn_bar}`\n\n"
    report += f"**Purchase Likelihood:** `{buy_bar}`\n\n"
    report += "---\n\n"
    report += "### Key Signals\n\n"

    signals = []
    if days_since_purchase > 60:
        signals.append(f"⚠️  Last purchase was **{days_since_purchase} days ago** — engagement is dropping")
    if support_tickets >= 5:
        signals.append(f"⚠️  **{support_tickets} support tickets** — customer may be frustrated")
    if email_open_rate < 0.2:
        signals.append("⚠️  **Low email engagement** — re-engagement campaign recommended")
    if purchases >= 10:
        signals.append(f"✓  **{purchases} purchases** — loyal customer history")
    if email_open_rate >= 0.5:
        signals.append("✓  **High email engagement** — customer is paying attention")
    if tenure >= 12:
        signals.append(f"✓  **{tenure} months** customer — long-term relationship")

    if signals:
        report += "\n".join(signals) + "\n\n"

    report += "---\n\n"
    report += "> ⚠️ Demo model trained on synthetic data. Your production system will learn "
    report += "from your actual customer history for accurate predictions.\n\n"
    report += "**[→ Get a Custom Customer Predictor for Your Business](https://www.lead-ai.us)**"

    return report


with gr.Blocks(
    title="Lead.AI Customer Predictor",
    theme=gr.themes.Soft(primary_hue="blue"),
    css=".footer { text-align:center; margin-top:20px; color:#666; }"
) as demo:

    gr.Markdown("""
# 🎯 Lead.AI Customer Predictor
### Know Which Customers Are About to Leave — Before They Do

Enter customer data below. The AI will predict churn risk, purchase likelihood,
and tell you exactly what action to take.

> 💼 This is a live proof-of-concept. [Request a custom system →](https://www.lead-ai.us)
""")

    with gr.Row():
        with gr.Column():
            gr.Markdown("### Customer Profile")
            tenure_in    = gr.Slider(1, 60, value=12, step=1,
                                     label="Customer Tenure (months)")
            purchases_in = gr.Slider(0, 50, value=8, step=1,
                                     label="Total Purchases")
            avg_spend_in = gr.Slider(10, 800, value=150, step=5,
                                     label="Average Order Value ($)")
            support_in   = gr.Slider(0, 15, value=1, step=1,
                                     label="Support Tickets (last 90 days)")
            days_in      = gr.Slider(1, 180, value=30, step=1,
                                     label="Days Since Last Purchase")
            email_in     = gr.Slider(0.0, 1.0, value=0.4, step=0.05,
                                     label="Email Open Rate (0 = never, 1 = always)")
            btn          = gr.Button("🔍 Analyze This Customer", variant="primary", size="lg")

        with gr.Column():
            gr.Markdown("### AI Prediction")
            report_md = gr.Markdown()

    btn.click(
        fn=predict_customer,
        inputs=[tenure_in, purchases_in, avg_spend_in, support_in, days_in, email_in],
        outputs=report_md,
    )

    gr.Examples(
        examples=[
            [2,  1,  45,  8, 90, 0.05],
            [36, 28, 320, 0,  5, 0.75],
            [8,  12, 180, 4, 45, 0.30],
            [18,  3,  90, 0, 120, 0.10],
        ],
        inputs=[tenure_in, purchases_in, avg_spend_in, support_in, days_in, email_in],
        label="Try These Customer Profiles",
    )

    gr.Markdown("""
---
<div class="footer">
🌐 <a href="https://www.lead-ai.us">www.lead-ai.us</a> &nbsp;|&nbsp;
💻 <a href="https://github.com/Lead-AI-US/lead-ai-customer-predictor">GitHub</a> &nbsp;|&nbsp;
🤗 <a href="https://huggingface.co/lead-ai-labs">Hugging Face</a>
<br><br>
<strong>Need this running on your real customer data?</strong>
<a href="https://www.lead-ai.us">Request a Custom Lead.AI Setup →</a>
</div>
""")

if __name__ == "__main__":
    demo.launch()