65. How to Forecast Sales with Time Series in Python (pandas and sklearn)

65.1 Why Forecast Next Months from Past Months

By Ravindra Bagale, trainer · LinkedIn · Last updated: 31 October 2026

Why. Priya Deshmukh (Pune ops) asks: “How many rupees of stock should Mauli Kirana order for Diwali month?” Looking at only last week is weak. Looking at the same month last year is better. A clear forecast turns a guess into a plan.

What.

• Forecast — a careful guess of future values from past values.
• Time series — data points in time order (day, week, month).
• Horizon — how far ahead you predict (next 1 month, next 6 months).

How.

Worked example A — Retail in Pune (₹).

Dataset note: fictional Mauli Kirana monthly sales. 36 rows. Columns: month, sales_inr. Path used in labs: mauli_monthly_sales.csv.

import pandas as pd

df = pd.read_csv("mauli_monthly_sales.csv", parse_dates=["month"])
print(df.head(3))
print(df.tail(3))
print("rows:", len(df), "columns:", list(df.columns))

Output you should see.

       month  sales_inr
0 2023-11-01     440000
1 2023-12-01     357700
2 2024-01-01     363000
        month  sales_inr
33 2026-08-01     694700
34 2026-09-01     799900
35 2026-10-01     878700
rows: 36 columns: ['month', 'sales_inr']

What it means. Oldest month first. Newest month last. That order matters for every method in this chapter.

Worked example B — Healthcare in Manchester (£).

Dataset note: fictional City Light Clinic monthly appointments. 6 rows. Columns: month, appointments. Fee context £45 per visit.

import pandas as pd

clinic = pd.DataFrame({
    "month": pd.date_range("2025-05-01", periods=6, freq="MS"),
    "appointments": [210, 198, 205, 240, 255, 248],
})
print(clinic)
print("Naive guess for next month:", int(clinic["appointments"].iloc[-1]))

Output you should see.

       month  appointments
0 2025-05-01           210
1 2025-06-01           198
2 2025-07-01           205
3 2025-08-01           240
4 2025-09-01           255
5 2025-10-01           248
Naive guess for next month: 248

What it means. A first forecast can be “same as last month” — we improve it later.

Worked example C — Comparison card.