Replace blank cells with a constant, average, or interpolated value
Replace blank CSV cells with a constant, column mean, median, previous value, or 'Unknown'.
🛡️ Your rows never leave your browser.
Why Onfyt: Applies different imputation strategies per column in one pass—mean for numeric, mode for categorical, constant for text fields—without multiple tool runs.
Interactive Sample Data
| id | name | revenue | region |
|---|---|---|---|
| 1 | Grace Kerubo | 4500 | Nairobi |
| 2 | Peter Njoroge | Mombasa | |
| 3 | Mercy Akinyi | 3200 |
How this tool works
Blank cells in a training dataset are not neutral — they introduce bias. A model trained on rows that silently omit values for lower-income respondents learns the missingness pattern, not the underlying distribution. Three imputation strategies cover the most common scenarios: column mean for numeric attributes without heavy outliers, column median for skewed distributions such as income or duration, and a fixed sentinel value for categorical fields where the absence of a value is itself meaningful information. DataFix applies these strategies per-column, letting you choose different strategies for different columns in the same file. Handling missing data is a fundamental data preparation step. Depending on analytical requirements, blanks must be populated with specific default constants (such as "0", "Unknown", or "N/A") or statistical measures like the column mean or median. DataFix provides flexible missing-value imputation rules applied consistently across any selected column. Imputing default fallback values (such as zero, 'Unknown', or custom sentinels) into blank cells eliminates null-handling exceptions in downstream data processing applications. Imputing default values into blank cells prevents null-pointer exceptions in microservices and ensures complete datasets for analytics. Applying explicit fallback constants standardizes records across reporting pipelines. Multiple sentinel tokens (including 'NaN', 'null', 'N/A', and whitespace) are recognized and replaced simultaneously in a single processing pass. Replacing blank fields with explicit placeholder tokens prevents numeric aggregators from skipping missing observations.
Things that usually go wrong
- Mean/median fill requires all non-null values to be numeric — mixed columns produce errors.
- Forward fill at the top of a column leaves the first cell empty if it was originally null.
Edge cases we tested
🧪 Column where 90% of values are blank
Mean/median computed from the non-blank 10%; warning about imputation reliability.
🧪 Fill with mean on a column with outliers
Mean is skewed; user suggested to use median instead.
Frequently Asked Questions
What is forward fill?
Each null cell takes the value of the most recent non-null cell above it.
What is backward fill?
Each null cell takes the value of the nearest non-null cell below it.
Related tools
Clean
Split Full Name into First and Last
Separate full name columns into dedicated first_name and last_name columns in CSV and Excel spreadsheets.
Open Tool →
Clean
Combine First and Last Name
Merge separate first and last name columns into a single clean full_name column, trimming extra spaces.
Open Tool →
Clean
Remove Duplicate Rows
Eliminate duplicate rows across your entire dataset or based on specific unique key columns.
Open Tool →