Why Every Automation Tester Should Learn Pandas in 2026

import pandas as pd

# 200 synthetic test users with a mix of roles and states
n = 200
users = pd.DataFrame({
    "username": [f"qa_user_{i:03d}" for i in range(n)],
    "email":    [f"qa_user_{i:03d}@example.com" for i in range(n)],
    "role":     (["admin", "editor", "viewer", "viewer"] * (n // 4)),
    "active":   [i % 10 != 0 for i in range(n)],   # every 10th user inactive
})

print(users["role"].value_counts())
print(users["active"].value_counts())

users.to_csv("test_users.csv", index=False)

import pandas as pd

# Boundary and invalid values for a signup form
edge_cases = pd.DataFrame([
    {"case": "min length name",  "name": "Al",         "age": 18,  "expect": "accept"},
    {"case": "too short name",   "name": "A",          "age": 25,  "expect": "reject"},
    {"case": "max age",          "name": "Meera",      "age": 120, "expect": "accept"},
    {"case": "under age",        "name": "Ravi",       "age": 17,  "expect": "reject"},
    {"case": "unicode name",     "name": "Zoë",        "age": 30,  "expect": "accept"},
    {"case": "spaces only",      "name": "   ",        "age": 30,  "expect": "reject"},
])

edge_cases.to_csv("signup_cases.csv", index=False)
print(edge_cases.groupby("expect").size())

case,cart_total,coupon,expected_total
no coupon,1000,,1000
ten percent,1000,SAVE10,900
flat hundred,1000,FLAT100,900
flat on small cart,80,FLAT100,0
invalid coupon,1000,BADCODE,1000
zero cart,0,SAVE10,0

import pandas as pd
import pytest


def apply_discount(total, coupon):
    # Stand-in for the real code or API call under test
    if coupon == "SAVE10":
        return round(total * 0.9, 2)
    if coupon == "FLAT100":
        return max(total - 100, 0)
    return total


cases = pd.read_csv("discount_cases.csv", dtype={"coupon": str})
cases["coupon"] = cases["coupon"].fillna("")

# Guard the data file itself
assert not cases["case"].duplicated().any(), "duplicate case names"
assert cases["expected_total"].notna().all(), "missing expected values"


@pytest.mark.parametrize("row", cases.to_dict("records"),
                         ids=cases["case"].tolist())
def test_discount(row):
    actual = apply_discount(row["cart_total"], row["coupon"])
    assert actual == row["expected_total"]

# Run the suite with XML output each time, e.g.:
#   python -m pytest --junitxml=results/run_01.xml
import glob
import xml.etree.ElementTree as ET
import pandas as pd

rows = []
for path in sorted(glob.glob("results/run_*.xml")):
    for tc in ET.parse(path).getroot().iter("testcase"):
        failed = tc.find("failure") is not None or tc.find("error") is not None
        skipped = tc.find("skipped") is not None
        rows.append({
            "run": path,
            "test": f"{tc.get('classname')}::{tc.get('name')}",
            "time": float(tc.get("time", 0)),
            "status": "skipped" if skipped else ("failed" if failed else "passed"),
        })

results = pd.DataFrame(rows)
summary = (results[results["status"] != "skipped"]
           .groupby("test")["status"].nunique())
flaky = summary[summary > 1].index.tolist()
print("Flaky tests:", flaky)

slowest = results.groupby("test")["time"].mean().nlargest(5)
print(slowest)

import pandas as pd

expected = pd.read_csv("expected_orders.csv")   # e.g. saved from a DB query
actual   = pd.read_csv("downloaded_report.csv")  # file the app exported

compare = expected.merge(actual, on="order_id", how="outer",
                         suffixes=("_exp", "_act"), indicator=True)

missing_in_report = compare[compare["_merge"] == "left_only"]
extra_in_report   = compare[compare["_merge"] == "right_only"]

both = compare[compare["_merge"] == "both"]
wrong_amount = both[(both["amount_exp"] - both["amount_act"]).abs() > 0.01]

print(len(missing_in_report), "orders missing from report")
print(len(extra_in_report), "unexpected orders in report")
print(len(wrong_amount), "orders with wrong amount")
print(wrong_amount[["order_id", "amount_exp", "amount_act"]].head(10))

TEST DATA FILE CHECKLIST (before committing a CSV)

File: ______________________   Used by: test_____________.py

[ ] Synthetic data only, no real personal information
[ ] Each row has a unique, readable case name
[ ] No blank expected values (isna().sum() == 0)
[ ] Includes boundaries: min, max, just outside
[ ] Includes invalid formats and empty values
[ ] IDs and codes loaded as text (dtype=str)
[ ] Reviewed by a second person in the pull request

WEEKLY FLAKY TEST REPORT

Period        : ________ to ________   Runs analysed: ___

Test                                  Pass  Fail  Likely cause
------------------------------------  ----  ----  -------------------
tests/test_checkout.py::test_coupon    __    __   timing / wait
tests/test_search.py::test_filters     __    __   shared test data

Slowest tests (mean seconds)
  1. ______________________  __._ s
  2. ______________________  __._ s

Actions: quarantine ___, fix waits in ___, owner: ________

DATA DEFECT REPORT

Title      : Monthly sales export shows wrong amount for discounted orders
Environment: staging, build ______
Check used : pandas merge of DB query vs downloaded CSV on order_id
Scope      : ___ of ___ orders affected
Examples   :
  order_id  expected  actual
  ________  ________  ________
  ________  ________  ________
Pattern    : all affected orders used coupon FLAT100
Expected   : export amount equals amount stored after discount
Attached   : comparison script and mismatch CSV