Data Science Applications in Retail Real-World Examples

import pandas as pd

sales = pd.read_csv("store_sales.csv", parse_dates=["date"])
weekly = sales.set_index("date")["units"].resample("W").sum()

train, test = weekly[:-8], weekly[-8:]
# Seasonal naive: value from the same week one year (52 weeks) earlier
baseline = weekly.shift(52)[-8:]

mae = (test - baseline).abs().mean()
print("Seasonal naive MAE:", round(mae, 1))

from statsmodels.tsa.holtwinters import ExponentialSmoothing

# Needs at least two full years of weekly history
model = ExponentialSmoothing(
    train, trend="add", seasonal="add", seasonal_periods=52
).fit()
forecast = model.forecast(8)

mae = (test - forecast).abs().mean()
print("Holt-Winters MAE:", round(mae, 1))

from itertools import combinations
from collections import Counter
import pandas as pd

tx = pd.read_csv("transactions.csv")  # columns: invoice_id, product
baskets = tx.groupby("invoice_id")["product"].apply(lambda s: sorted(set(s)))
n = len(baskets)

item_count = Counter(p for b in baskets for p in b)
pair_count = Counter(pair for b in baskets for pair in combinations(b, 2))

rows = []
for (a, b), c in pair_count.items():
    support = c / n
    lift = support / ((item_count[a] / n) * (item_count[b] / n))
    rows.append((a, b, c, round(support, 4), round(lift, 2)))

rules = pd.DataFrame(rows, columns=["item_a", "item_b", "baskets", "support", "lift"])
print(rules[rules["baskets"] >= 20].sort_values("lift", ascending=False).head(10))

import pandas as pd

orders = pd.read_csv("orders.csv", parse_dates=["order_date"])
snapshot = orders["order_date"].max() + pd.Timedelta(days=1)

rfm = orders.groupby("customer_id").agg(
    recency=("order_date", lambda d: (snapshot - d.max()).days),
    frequency=("order_id", "nunique"),
    monetary=("amount", "sum"),
)
print(rfm.describe())

from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans

X = StandardScaler().fit_transform(rfm[["recency", "frequency", "monetary"]])
km = KMeans(n_clusters=4, n_init=10, random_state=42)
rfm["segment"] = km.fit_predict(X)

# Profile each segment so marketing can name it
print(rfm.groupby("segment").mean().round(1))

Project: Weekly demand forecast for 10 products
Data: Public retail sales dataset (2+ years, weekly)
Baseline: Seasonal naive (same week last year)
Model: Holt-Winters or gradient boosting with
       promo and holiday features
Metric: MAE per product, compared with baseline
Output: Chart of forecast vs actual + ordering note

Project: Products bought together
Data: Public online retail transactions
Method: Pair counts, support, lift
Filter: Minimum 20 baskets per pair
Output: Top 10 pairs with a bundle idea for each
Limitation: Correlation, not proof of cause

Project: Customer segments for a loyalty campaign
Data: Orders with customer IDs and dates
Features: Recency, Frequency, Monetary
Method: StandardScaler + KMeans (try k = 3 to 6)
Output: Segment profiles, names and one offer each
Check: Are segments stable on a different month?