generate_sample_data.py

"""
Generates synthetic order data for RFM analysis.
Produces data/sample_orders.csv with ~1,000 customers and ~8,000 orders.
"""
import pandas as pd
import numpy as np
from pathlib import Path

np.random.seed(42)

N_CUSTOMERS = 1_000
N_ORDERS    = 8_000
DATE_START  = pd.Timestamp("2022-01-01")
DATE_END    = pd.Timestamp("2024-12-31")

# Customer segments (simulate real-world skew)
SEGMENT_WEIGHTS = {
    "champion":   0.10,
    "loyal":      0.20,
    "at_risk":    0.25,
    "hibernating":0.25,
    "lost":       0.20,
}

def make_customers(n: int) -> pd.DataFrame:
    segments = np.random.choice(
        list(SEGMENT_WEIGHTS.keys()),
        size=n,
        p=list(SEGMENT_WEIGHTS.values())
    )
    return pd.DataFrame({
        "customer_id": [f"C{str(i).zfill(5)}" for i in range(1, n + 1)],
        "segment":     segments,
    })

def make_orders(customers: pd.DataFrame, n: int) -> pd.DataFrame:
    # Champions and loyals get more orders, recency bias toward recent
    freq_map = {
        "champion":    (12, 30),
        "loyal":       (6,  15),
        "at_risk":     (2,   6),
        "hibernating": (1,   3),
        "lost":        (1,   2),
    }
    monetary_map = {
        "champion":    (200, 800),
        "loyal":       (80,  300),
        "at_risk":     (40,  150),
        "hibernating": (20,   80),
        "lost":        (10,   50),
    }
    recency_bias = {
        "champion":    0.85,   # mostly recent
        "loyal":       0.70,
        "at_risk":     0.40,
        "hibernating": 0.20,
        "lost":        0.05,   # mostly old
    }

    rows = []
    total_days = (DATE_END - DATE_START).days

    for _, cust in customers.iterrows():
        seg = cust["segment"]
        lo, hi = freq_map[seg]
        n_orders = np.random.randint(lo, hi + 1)
        bias = recency_bias[seg]

        for _ in range(n_orders):
            # Weighted random date (bias toward recent or old)
            u = np.random.beta(bias * 3 + 0.5, (1 - bias) * 3 + 0.5)
            order_date = DATE_START + pd.Timedelta(days=int(u * total_days))

            mlo, mhi = monetary_map[seg]
            amount = round(np.random.uniform(mlo, mhi), 2)

            rows.append({
                "customer_id": cust["customer_id"],
                "order_date":  order_date.date(),
                "grand_total": amount,
            })

    return pd.DataFrame(rows)

if __name__ == "__main__":
    print("Generating customers...")
    customers = make_customers(N_CUSTOMERS)

    print("Generating orders...")
    orders = make_orders(customers, N_ORDERS)
    orders = orders.sort_values("order_date").reset_index(drop=True)

    out = Path(__file__).parent / "data" / "sample_orders.csv"
    out.parent.mkdir(parents=True, exist_ok=True)
    orders.to_csv(out, index=False)

    print(f"\nSaved {len(orders):,} orders for {orders['customer_id'].nunique():,} customers")
    print(f"Date range: {orders['order_date'].min()} → {orders['order_date'].max()}")
    print(f"Revenue:    ${orders['grand_total'].sum():,.2f}")
    print(f"Output:     {out}")