Sum
S = Σ xᵢ
Only numerical values participate in this addition, without imputing zeros for excluded cells.
CSV Statistics and Data Profile
Profile CSV data locally: column types, missing and unique values, minimum, maximum, mean, median, mode, and standard deviation.
- Clipboard
Data not sent
Data profile
Was this tool useful?
How does this tool work?
Before adding a column or joining an export, inspect its missing values, diversity and actual spellings. CSV profiling examines each column and provides a descriptive diagnosis rather than automatic repair. It separates populated cells from absences and computes indicators using values that can be interpreted as numbers. An identifier containing only digits can therefore be detected as numerical without being a meaningful quantity to add. Business interpretation remains essential when choosing the next transformation, and a plausible average alone does not establish that every source record was valid or included.
The reader detects commas, semicolons, tabs or vertical bars and respects quoted fields, including multiline cells. Headers start enabled; without them, generic column names are created and the first record contributes data. Each column is profiled independently. A missing field, empty cell or whitespace-only value counts as missing. Its rate uses all data records as the denominator, while the populated-value count excludes those absences. Header records are not included in the data population when their option is enabled.
Distinct values count different populated spellings, not merely values appearing once. Case, spaces in non-empty cells and numerical notation can therefore affect cardinality. Calculations, however, recognise decimal points or commas and valid thousands spaces. A non-numerical annotation is never forced to zero. Empty, number, boolean, date, text and mixed types are inferred from the cells. They are practical inspection hints rather than comprehensive validation of dates or conformity with your application's expected schema. A numerical-looking reference may still be an identifier.
Minimum, maximum, sum, mean, median and standard deviation use convertible numbers only. Median sorts those values and selects the centre, averaging the two central values for an even count. Standard deviation uses the population convention, dividing by n rather than n − 1. Numerical mode is reported only when a value repeats; if several modes tie, the smallest is displayed. A single numerical value has zero population standard deviation. Display uses up to four decimal places in the page's locale, while computation uses JavaScript floating-point precision.
How to use this tool
Supply the export with correct settings
Paste or import CSV and confirm whether the first record is a header. The upload control limits files to 5 MiB. Data and the resulting profile are processed in memory within your browser.
Read counts before interpreting averages
Check record and column totals, followed by populated, missing and distinct counts. An unexpected type or substantial missingness needs explanation before a numerical indicator can safely support your interpretation of the file.
Interpret and export the diagnosis
Compare mean and median, inspect extremes and read the mode when available. Copy or export the text report to document the review, then perform justified corrections separately using the source information.
Calculations used by the profile
The xᵢ are the n recognised numerical cells after excluding missing values and invalid text. N denotes every data record and m the missing cells in the column. Sums, means and standard deviations retain the analysed values' unit; the missing-value rate is expressed as a percentage.
Arithmetic mean
x̄ = S / n
The denominator counts numerical values rather than every record in the source file.
Population standard deviation
σ = √(Σ(xᵢ − x̄)² / n)
This convention describes the supplied series without applying the n − 1 sample correction.
Missing-value rate
p = 100 × m / N
The rate uses all data records, including those contributing no numerical value.
Use cases
Locate an incomplete column
Measure missingness in an identifier or measurement field before deciding whether to repair the source, exclude specific records or deliberately retain the absences as part of the dataset.
Investigate mixed data types
Find numerical values mixed with annotations or unsupported spellings. The mean then describes only the numerical subset and may not represent every entity or measurement in the business population.
Compare profiles before and after cleanup
Review counts, cardinality and descriptive figures after a transformation to check that it addressed the intended problem without unexpectedly removing records or changing meaningful values in the source.
Examples
Measure an absence without treating it as zero
Three values are populated and one is missing, giving 25% missingness. Their sum is 300; mean and median are 100. No number repeats, so there is no numerical mode. Population standard deviation is approximately 16.3299.
Input
amount
100
120
""
80Result
amount · type number · 3 values · 1 missing (25%) · 3 unique · min 80 · max 120 · sum 300 · mean 100 · median 100 · standard deviation 16.3299 · mode -Compare an amount with a status field
Amount contains four numbers and one absence, with mean 22.5 and mode 20. Status contains two distinct values across five populated cells. Cardinality two does not mean that only two records have a value in that column.
Input
amount;status
10;ok
20;ok
20;NA
40;ok
;NAResult
amount · type number · 4 values · 1 missing (20%) · 3 unique · min 10 · max 40 · sum 90 · mean 22.5 · median 20 · standard deviation 10.8972 · mode 20
status · type text · 5 values · 0 missing (0%) · 2 uniqueTips and best practices
- High cardinality is normal for a unique identifier but suspicious in a field expected to contain only a few categories. Interpret distinct counts according to the field's role rather than as a universal quality score.
- Spellings 1 and 1.0 can be distinct for cardinality while representing the same numerical value in calculations. This difference between text comparison and numerical interpretation does not mean that the counters contradict one another.
- A plausible mean does not prove every cell contributed to it. Check mixed types and missingness, then inspect rejected spellings before making decisions based solely on a summary statistic that may describe a smaller subset.
Sources and references
- NIST/SEMATECH, e-Handbook of Statistical Methods: Measures of Location (opens in a new tab)Page in English
- Weisstein E.W., Standard Deviation (opens in a new tab), Wolfram MathWorldPage in English
Frequently asked questions
Are empty cells included as zeros?
No. They increase the missing count and rate but contribute neither to the numerical sum nor to the mean's denominator. A genuine zero remains a populated value and participates in every applicable calculation rather than being treated as an absence.
Why does standard deviation differ from my spreadsheet?
This profile uses population standard deviation, dividing by n. A sample-standard-deviation function divides by n − 1 and gives a different result. Also verify that both applications selected exactly the same numerical cells rather than treating annotations differently.
Does unique values mean values appearing only once?
Here the label means the number of distinct populated values. A value repeated a hundred times contributes one unit to that cardinality. Missing cells are excluded, while non-empty spelling variants may be counted separately.
Does a date or number type guarantee correct data?
No. Inference helps inspection but knows neither your schema nor the meaning of its fields. A numerical reference may be an identifier, and a date-like spelling does not replace dedicated business-rule and calendar validation.
Related tools
Keep going with these tools
Tools similar or complementary to this one.
Similar tools
Remove duplicates, empty rows and unnecessary spaces.
Data & spreadsheetsUse this tool
Create a new numeric column from a formula that references other table columns.
Data & spreadsheetsNewUse this tool
Cross a row dimension with a column dimension, then calculate count, sum, average, minimum, or maximum.
Data & spreadsheetsFeaturedUse this tool
Read next
GuideBest practicesBeginner
Data and tables: understanding CSV, Excel, JSON and dataset structure
Learn how tabular data is structured, how CSV, Excel, JSON and TSV differ, and how to choose the right format before cleaning, converting or automating a workflow.
29 August 20263 min
GuideBest practicesBeginner
How to clean and prepare a CSV or Excel file before using it
Learn a practical order for removing empty rows, trimming values, standardizing headers, checking identifiers and validating a CSV or Excel dataset.
29 August 20263 min
GuideBest practicesBeginner
Compare two CSV or Excel files and identify the differences
Learn how to compare two versions of a table using a stable key and distinguish added, removed and modified records.
29 August 20262 min
GuideBest practicesBeginner
Split a CSV or Excel table into smaller files
Discover when to split a large table by row count or by a column value, and how to preserve a useful structure in every output file.
29 August 20261 min