What a Data Analyst Does

By the end of this lesson you will know what a data analyst actually does from morning to evening, which tasks fill most of the week, which job titles mean similar work, and which skills each task needs. You will also run three small checks that analysts really run every day.

Job adverts for analysts often list ten tools and say little about the work itself. So let us follow one analyst, Meera, who works for an online store in Pune, through an ordinary Tuesday.

A day in the life

Notice the mix. Some work repeats every week, some arrives without warning, and some is a slow, deep project. Very little of it is "build a fancy model". Most of it is careful counting, checking and explaining.

The four kinds of work

1. Monitoring. Every morning someone must check that the numbers look normal. Think of a railway station master walking the platform before the first train. He is not fixing anything yet. He is looking for anything unusual.

Here is a typical morning check on our practice shop: how many orders are in each status?

SELECT status,
       COUNT(*) AS orders
FROM orders
GROUP BY status
ORDER BY orders DESC, status;

status    | orders
----------+-------
delivered | 10
pending   | 2
shipped   | 2
cancelled | 1
(4 rows)

What you should see: four statuses, with "delivered" far ahead. If one morning "cancelled" suddenly jumped to the top, Meera would investigate before anyone asks.

2. Ad-hoc requests. "Ad hoc" means "for this one purpose". These are quick questions from managers: "Can you pull the list of...?" They must be answered fast and correctly. A good analyst repeats the question back first, to be sure what is wanted.

SELECT o.order_id,
       c.first_name,
       c.city,
       o.order_date,
       o.status
FROM orders AS o
JOIN customers AS c ON c.customer_id = o.customer_id
WHERE o.status IN ('pending', 'shipped')
ORDER BY o.order_date;

order_id | first_name | city      | order_date | status
---------+------------+-----------+------------+--------
108      | Vikram     | Rohtak    | 2025-03-15 | shipped
111      | Arjun      | Hyderabad | 2025-04-10 | pending
113      | Rohan      | Mumbai    | 2025-05-02 | shipped
115      | Aarav      | Rohtak    | 2025-05-20 | pending
(4 rows)

What you should see: four orders that have not reached the customer yet, oldest first. Sorting oldest first matters: the manager will chase the oldest orders first.

3. Recurring reports. Weekly sales, monthly targets, quarterly reviews. These should be automated as much as possible, with a saved query, a pandas script or a Power BI report that refreshes. If you copy and paste the same numbers every Monday, you have found something to automate.

4. Analysis projects. These are the deep questions: "Why did repeat purchases fall in Lucknow?" They take days, need several data sources, and end with a written recommendation. They are the most interesting work, and they are what you show in interviews.

Data quality is part of the job

Before Meera trusts any file, she runs a quick health check: how many rows, which dates, which blanks. This takes one minute and saves hours of embarrassment later.

import pandas as pd

sales = pd.read_csv("sales.csv")

print("Rows and columns:", sales.shape)
print("Dates from", sales["order_date"].min(), "to", sales["order_date"].max())
print("Duplicate rows:", sales.duplicated().sum())
print("Blank cells per column:")
print(sales.isna().sum()[sales.isna().sum() > 0])

Rows and columns: (1217, 10)
Dates from 2025-01-01 to 2025-12-31
Duplicate rows: 8
Blank cells per column:
quantity         6
payment_mode    11
dtype: int64

What you should see: about 1,200 orders covering the whole of 2025, a few exact duplicate rows, and blanks in two columns. None of this is unusual. Real data always has some mess. The analyst's job is to find it before it reaches a report. You will clean this file properly in A Data Cleaning Checklist.

Job titles you will see

Companies use many names for similar work. Read the job description, not just the title.

Many people start as MIS executives or junior analysts and grow into product or senior analyst roles. The core skills stay the same.

Skills each task needs

Common mistakes

• Answering a different question. The manager asked about pending orders; you sent all orders. Repeat the question back before you start.
• Skipping the health check. Reporting duplicate orders as real sales is one of the fastest ways to lose a manager's trust.
• Doing manual work forever. If you run the same steps every week, save the query or write a script.
• Hiding in the data. Analysts must talk to people. Ask the sales team what a strange number means; they often know.

Key points

• An analyst's week mixes monitoring, ad-hoc requests, recurring reports and deeper analysis projects.
• Morning checks look for anything unusual before anyone else notices.
• Always run a quick health check (rows, dates, duplicates, blanks) before trusting data.
• Titles vary (MIS, BI, product, business analyst); read the duties, not just the title.
• Communication and documentation matter as much as SQL and Python.

SELECT o.order_id,
       c.first_name,
       c.city,
       o.order_date
FROM orders AS o
JOIN customers AS c ON c.customer_id = o.customer_id
WHERE o.order_date BETWEEN '2025-03-01' AND '2025-03-31'
ORDER BY o.order_date;