Python Mock Interview — Real Questions & Model Answers

Excel: pivot tables, charts, quick one-time analysis, sharing with non-technical stakeholders. Best for < 100K rows, interactive exploration.

Python: automation, 1M+ rows, complex transformations, reproducibility, APIs, scheduling, ML. Best when the same task runs repeatedly or on large data.

I use Excel for quick stakeholder reports and Python when building a repeatable pipeline or working with data too large for Excel.

Interviewers want to know you understand both tools and can choose the right one. Showing pragmatism (using the right tool, not always Python) demonstrates professional maturity.

df["purchase_date"] = pd.to_datetime(df["purchase_date"], format="%d-%m-%Y")
df["year"]  = df["purchase_date"].dt.year
df["month"] = df["purchase_date"].dt.month

# Verify:
print(df[["purchase_date","year","month"]].head())

Always specify format= for non-standard date strings — it is 5x faster than auto-parsing on 500K rows. dt.year and dt.month are vectorised, no loop needed. The print verify shows you test your own code.

# Step 1: Check percentage:
null_pct = df["revenue"].isnull().sum() / len(df) * 100  # 2.4%

# Step 2: Understand why — random or systematic?
df[df["revenue"].isnull()].describe()  # any pattern?

# Step 3: Choose strategy:
# If 2.4% and random → fill with median
df["revenue"] = df["revenue"].fillna(df["revenue"].median())

# If nulls cluster in a segment → fill by group mean:
df["revenue"] = df.groupby("region")["revenue"].transform(
    lambda x: x.fillna(x.median())
)

2.4% null rate is low — imputation is appropriate. Always understand WHY values are missing before deciding how to handle them. Group-level imputation is more accurate than global imputation.

# Step 1: Nulls:
null_pct = df.isnull().sum() / len(df) * 100
high_null = null_pct[null_pct > 50].index  # >50% null — likely useless

# Step 2: Variance — constant columns:
zero_var = df.select_dtypes("number").columns[df.select_dtypes("number").std() == 0]

# Step 3: Cardinality of categoricals:
for col in df.select_dtypes("object").columns:
    print(col, df[col].nunique())
# 1 unique → useless, 500K unique → likely ID (not analytical)

# Step 4: Correlation with target:
df.corr()["revenue"].sort_values(ascending=False)

Column triage before analysis is a professional skill. Columns with >50% nulls, zero variance, or 1 unique value are almost always dropped. High cardinality strings are IDs, not features. This four-step process shows systematic thinking.

top5_by_region = (
    df.groupby(["region", "product"])["revenue"]
    .sum()
    .reset_index()
    .sort_values(["region", "revenue"], ascending=[True, False])
    .groupby("region")
    .head(5)
)

print(top5_by_region)

The pattern: sum by region+product → sort by revenue descending within each region → take top 5 per region using groupby().head(5). This is one of the most common interview questions — memorise this exact pattern.