I would first understand why the values are missing and determine how important the column is to the analysis. If the column is critical, I would consider an appropriate imputation technique such as the mean, median, or mode depending on the data. If the missing values are excessive and the column provides little value, I might remove it after evaluating its business impact.

I would remove rows when only a small percentage of observations are missing and deleting them is unlikely to affect the overall analysis. If removing rows would introduce bias or significantly reduce the dataset, I would prefer an appropriate imputation method instead.

The median is less affected by extreme values or outliers. When numerical data is skewed, using the median usually provides a better representation of the typical value than the mean.

Missing values can reduce model accuracy, introduce bias, and produce unreliable reports. If not handled correctly, they may lead to incorrect business conclusions and poor decision-making.

Common functions include:

isnull()
isna()
sum()
dropna()
fillna()

These functions help identify, remove, or replace missing values during data cleaning.

I would first check whether complete duplicate rows exist using duplicated records. Then I would check key columns such as customer ID, email, phone number, order ID, or transaction ID because duplicates are not always identical across every column.

No, duplicate rows should not always be removed immediately. Some duplicates may be valid, such as repeat purchases or multiple transactions from the same customer. I would first understand the business context before deciding whether the duplicate is an error or a valid record.

Duplicate customer records can lead to incorrect customer counts, inaccurate sales reports, wrong churn calculations, duplicate marketing emails, and poor customer experience. They can also affect dashboards and business decisions.

I would group the data by customer ID and check whether fields like name, email, phone number, address, or date of birth have multiple different values. This helps identify cases where the same ID is linked to conflicting customer details.

df.duplicated()
df.drop_duplicates()
df[df.duplicated()]
df.duplicated(subset=["customer_id"])
df.drop_duplicates(subset=["customer_id"])

An outlier is a data point that is significantly different from the rest of the dataset. Outliers can distort statistical measures such as the mean, affect visualizations, reduce machine learning model accuracy, and lead to incorrect business conclusions if not investigated properly.

Common methods include:

Boxplots
Interquartile Range (IQR)
Z-score
Scatter plots
Histograms

The choice of method depends on the data distribution and business problem. For skewed data, the IQR method is often preferred.

No. Outliers should only be removed after understanding their cause. Some outliers represent data entry errors, while others reflect genuine business events such as holiday sales, premium customers, or large corporate purchases. Removing valid outliers may eliminate valuable business insights.

I would first verify whether the spike is accurate by checking the data source and looking for duplicate or incorrect records. If the data is valid, I would investigate possible business reasons such as promotional campaigns, seasonal events, product launches, holidays, or bulk customer purchases before deciding whether it should be treated as an outlier.

It is important to be transparent about limitations in the data.

I would explain:

Missing or incomplete data.
Small sample sizes.
Assumptions made during analysis.
Potential sources of error.
Confidence in the conclusions.

Being honest about uncertainty builds trust and helps stakeholders make informed decisions.

✅ Key Takeaway: Confidence comes from transparency, not from pretending the data is perfect.

Outliers can significantly increase or decrease the mean, making it less representative of the typical data. They may also influence regression models, clustering algorithms, and distance-based machine learning techniques, leading to less accurate predictions. In some cases, using robust statistics such as the median or transforming the data may be more appropriate than removing outliers.

Consistent date formatting ensures that dates are interpreted correctly during filtering, sorting, merging, and time-based analysis. Inconsistent formats such as 01/05/2025 and 2025-05-01 can lead to incorrect reports, inaccurate calculations, and errors in dashboards or machine learning models.

I would first identify all variations of the same value, then replace them with a single standardized format using business rules or a mapping table. Standardizing categorical values improves grouping, reporting accuracy, and prevents the same category from being counted multiple times.

When numbers are stored as text, mathematical operations such as sums, averages, and comparisons may produce incorrect results or fail completely. It can also prevent proper sorting, filtering, and visualization, leading to inaccurate analysis and business reports.

Standardized column names make datasets easier to read, maintain, and process. Using consistent naming conventions avoids errors in code, simplifies joins and automation, and improves collaboration among team members working on the same project.

Before creating reports, I would verify that:

Date columns are stored as Date/Datetime
Numerical fields are stored as Integer or Float
Categorical variables are stored as String or Category
Boolean fields contain only True/False values
Identifier columns (such as Customer ID or Order ID) have the correct data type

Verifying data types helps ensure calculations, visualizations, and business metrics are accurate.

After cleaning the data, I verify that the number of records is correct, missing values have been handled appropriately, duplicates are removed where necessary, data types are correct, and business metrics such as totals or averages remain accurate. I also compare the cleaned data with the original source to ensure no important information was lost.

Before sharing the dataset, I would check:

Missing values
Duplicate records
Correct data types
Invalid or inconsistent values
Data ranges (e.g., age cannot be negative)
Total record count
Business metrics such as revenue or customer counts
Consistency with source data

These checks help ensure the data is accurate and reliable for decision-making.

I compare the data against expected business rules and logical constraints. For example, negative sales amounts, ages greater than 120 years, future birth dates, or impossible percentages above 100% may indicate invalid values. I also use summary statistics, filters, and validation rules to detect anomalies.

Data cleaning focuses on correcting issues such as missing values, duplicates, formatting inconsistencies, and incorrect data types. Data validation is the process of confirming that the cleaned data is accurate, complete, and meets business rules before it is used for reporting or analysis. In simple terms, cleaning fixes the data, while validation confirms the fixes were successful.

I would compare key statistics before and after cleaning, verify record counts, review summary reports, validate important business metrics, and perform spot checks on sample records. Whenever possible, I would also document each cleaning step so the process is transparent, reproducible, and easy to audit.

I would first identify all date formats used across the datasets and convert them into a single standard format, such as YYYY-MM-DD. After standardizing the dates, I would verify that no records were incorrectly converted and confirm that all branches can be combined without affecting time-based analysis or reporting.

I would compare other customer information such as email address, purchase history, account creation date, and recent activity. If the records represent the same customer, I would retain the most complete and up-to-date information or merge the records according to business rules rather than deleting data immediately.

Before making any changes, I would determine whether the negative values are valid business transactions, such as refunds or returns, or simply data entry errors. I would review transaction types, source systems, and business rules before deciding whether to correct, remove, or retain the records.

No. I would first investigate the cause of each outlier. Some outliers may represent genuine business events, such as large enterprise orders, seasonal promotions, or bulk purchases. Removing valid observations could lead to inaccurate analysis and poor business decisions, so I would explain the impact before making any changes

I would compare the source data used by both reports and check for differences in date filters, duplicate records, missing values, joins, calculations, and aggregation methods. I would also verify that both reports use the same business definitions for metrics such as revenue, profit, or customer count before identifying the root cause of the discrepancy.