Guides

Why Python for Data Science — and Your First Steps Order

By Ravindra Bagale, Data Science & Data Analytics trainer · LinkedIn · Published: 1 October 2026 · Last updated: 3 October 2026

Read this guide in: English · मराठी · हिंदी

Python is the main language for this Data Science course because it reads clearly, has pandas for tables, and runs well in Jupyter. First steps order: tiny Python → Jupyter → pandas → clean → EDA → charts. No sklearn / ML on this path yet.

Python basics → Jupyter → pandas → clean → EDA → charts

Language: English · मराठी · हिंदी

What do I need before this guide?

• Willingness to type small code (even 10 lines helps).
• Optional: Python already installed (next guide covers install).
• Course: Variables and types.

Before and after (look at the tables first)

Before: random tabs and sklearn-first confusion. After: ordered path with a pandas sum check of 190.

Why Python for data work?

What it is good at

1. Reading CSV and Excel into memory.
2. Cleaning columns with clear code you can re-run.
3. Grouping, joining and charting in one language.
4. Sharing a notebook so a teammate sees your steps.

Why you care

1. Job posts for analysts and data roles list Python often.
2. Repeatable scripts beat one-off mouse clicks when the file updates weekly.
3. The same skills transfer to Infosys-style client spreadsheets and shop exports.

What Python is not (here)

1. Not a requirement to know advanced OOP on day one.
2. Not a free pass to skip Excel forever — Excel still helps.
3. Not equal to “I train deep learning models”.

Why this order (and not ML first)

1. Models amplify whatever is in the table — including dirt.
2. Managers ask for clear tables and charts long before they ask for a model.
3. This course’s map stops at storytelling visuals on purpose.

Tiny story:

1. You get shop_sales.csv.
2. Wrong path: copy a Kaggle model, fail on column names, blame yourself.
3. Right path: read_csv → head → filter Online → sum 190 → bar chart.

Step 1 — Tiny Python (classroom minimum)

Vertical list:

1. Numbers and strings in variables.
2. A list of sales: [120, 70].
3. A dict for one row: {"city": "Mumbai", "sales": 120}.
4. An if to flag high sales.
5. A for loop to walk rows.
6. A small function that returns a total.

sales = [120, 70]
total = sum(sales)
print(total)  # 190

What you see: 190 in the cell output. What it means: you can check pandas later against the same number.

Course depth: Lists and dictionaries, if / for / functions.

Step 2 — Jupyter as your desk

1. A notebook is a page of cells.
2. Code cells run Python.
3. Markdown cells hold notes for your future self.
4. Tables from pandas print under the cell — easy teaching.

Why you care: data work is half explanation. Jupyter keeps code and notes together.

Step 3 — pandas as the table engine

1. import pandas as pd
2. df = pd.read_csv("shop_sales.csv")
3. df.head()
4. df["sales"].sum()

Suppose Online rows are Mumbai 120 and Pune 70:

online = df[df["channel"] == "Online"]
print(online["sales"].sum())  # 190

Real-world suppose examples (training style)

1. Shop sales CSV — city, channel, sales.
2. Instagram-like engagement — post_id, city, likes, comments.
3. Netflix-style watches — title, watches, missing cells to clean later.
4. Infosys client project spreadsheet — client, hours, week — groupby later.

Use small files first. Big data bravado can wait.

What you will not do yet

1. sklearn.fit anything.
2. Tune neural nets.
3. Pretend accuracy scores replace cleaning.

Packages that matter first

1. pandas — tables.
2. matplotlib / seaborn — charts.
3. openpyxl — Excel read/write helpers.
4. NumPy — under the hood arrays (course covers essentials).

Not first: sklearn, tensorflow, random AutoML demos.