Python for Business: What Should You Automate First?

Choose a task with stable rules

Good starting projects include checking a CSV export for missing identifiers, combining files with a consistent layout, or comparing two lists for unmatched records. Start with a report or exception list that a person reviews. Learn how to handle unusual inputs before letting a script change live records.

Python is not automatically the best tool. A one-time small analysis may be quicker in Excel. A recurring transformation may fit Power Query. A supported business-system feature may provide better ownership and controls than a custom script. Choose based on the recurring problem and who will maintain the solution.

Compare possible first projects

Write the specification in plain English

For the example below: read invoice records, require InvoiceID and Amount headers, count data rows, count blank IDs after removing surrounding spaces, and list nonblank IDs that appear more than once. Keep identifiers as text. Do not delete rows, add amounts, or decide whether a repeated ID is wrong.

Define the grain before applying the rule. An invoice-line export can legitimately repeat an invoice number across lines. In a multi-company file, the true key may be CompanyID plus InvoiceID. A duplicate test is only meaningful relative to its stated key and grain.

A runnable beginner example

This original example uses Python 3 and only its standard library. With Python installed, save the code in a plain-text file named invoice_check.py and run it from that folder using python invoice_check.py or the Python command configured on your system. It reads fictional data embedded in memory, prints a summary, and does not write a file or connect to an account.

import csv
from collections import Counter
from io import StringIO

sample = """InvoiceID,Amount
001,120.00
002,80.00
001,120.00
,50.00
"""

reader = csv.DictReader(StringIO(sample, newline=""))
required = {"InvoiceID", "Amount"}
if not required.issubset(reader.fieldnames or []):
    raise ValueError("Missing required columns")

rows = list(reader)
if any(None in row or any(value is None for value in row.values())
       for row in rows):
    raise ValueError("Row does not match the header")

ids = [row["InvoiceID"].strip() for row in rows]
counts = Counter(value for value in ids if value)
duplicates = sorted(key for key, count in counts.items() if count > 1)
blank_count = sum(not value for value in ids)

print("Rows:", len(rows))
print("Blank invoice IDs:", blank_count)
print("Repeated invoice IDs:", ", ".join(duplicates) or "None")

Expected result:

Rows: 4
Blank invoice IDs: 1
Repeated invoice IDs: 001

Understand what the code does

DictReader turns each CSV row into fields addressed by header name. The header check stops the script when required columns are absent. The row-shape check catches extra fields or missing trailing fields in parsed rows. The list of IDs retains strings, so “001” does not silently become the number 1. strip() removes surrounding spaces according to our stated rule.

Counter counts each nonblank ID. The sorted duplicate list includes IDs with counts greater than one. A blank row field is counted separately. CSV files can contain quoted commas and multiline fields, so use a CSV parser rather than splitting every line on a comma.

The script deliberately does not validate Amount. A successful run therefore says nothing about whether amounts are numeric, complete, or correct. Extend the specification and checks before using the result for financial analysis.

Test changes against known answers

1. Change the second “001” to “003.” The repeated-ID result should become None.
2. Fill the blank ID with “004.” The blank count should become zero.
3. Change the InvoiceID header to another name. The script should stop with “Missing required columns.”
4. Add an extra field to a data row. The row-shape check should reject it.

These tests check both normal behavior and failure behavior. When moving to a real CSV, read a copy with an explicit encoding and newline="", retain the original export, and confirm headers, row counts, and identifier rules. Unexpected schema changes should be visible rather than silently accepted.

Make the time-saving case honestly

Suppose a weekly review takes 30 minutes manually. A script takes 5 minutes to run and 10 minutes to review. That releases 15 minutes per week. Over 40 runs, gross time released is 600 minutes, or 10 hours. If development and maintenance require 6 hours, the illustrative net benefit is 4 hours. Time released becomes cash savings only if it actually changes spending.

When is a script ready for routine use?

Keep a brief runbook with the input location, schema, command, expected output, checks, owner, and recovery steps. Separate outputs from source files. For recurring work, record when each run occurred and which input it used. If it later writes records, design duplicate-run protection, review, and recovery before deployment. A script someone can maintain is more valuable than one only its author understands.

Frequently asked questions

Sources and further learning

Official documentation checked November 2, 2026. Examples and learning sequences on this page are original teaching illustrations. Product interfaces and plan availability can change.

• Python documentation: CSV reading and writing
• Python Packaging User Guide: environments and packages