CSV, Excel and Data Analyzer

1. 1Drop a file onto the box, tap it to choose one, or paste a copied file. CSV, TSV, TXT, XLSX and XLS files up to 20 MB are accepted. To paste text instead, use the text box below it.
2. 2For a workbook with several sheets, choose the sheet from the Sheet list that appears.
3. 3Read the Overview: rows, columns, missing values, the delimiter and any columns that held stray values.
4. 4Open Summary for per-column statistics, then Profile for types, percentage missing, distributions and the stray values.
5. 5Open Correlation to see how the numeric columns move together, with blue for positive and red for negative.
6. 6Open Preview to read the rows, 50 per page. Type in the filter box to keep matching rows, and press a column header to sort.
7. 7Open Chart, pick a chart type and a column (two for a scatter plot), or press Open Chart Maker for more chart types.
8. 8Press the download button in Summary to save the column statistics as a CSV file.

• Open .csv, .tsv, .txt, .xlsx and .xls files up to 20 MB, or paste a file or text.
• Parsing runs in a Web Worker for files and pastes, never on every keystroke. The paste box waits half a second after you stop typing.
• Automatic delimiter detection among comma, semicolon, tab and pipe, with quoted fields and a leading byte-order mark handled.
• Excel support through SheetJS, loaded only when a workbook is opened, with a sheet picker.
• Columns are tracked by position, so two columns with the same header stay separate. The second one is shown as "name (2)".
• Majority-type inference: a column is numeric when 95 percent or more of its non-missing values are numbers, and a date column on the same rule. Stray values are listed.
• Missing values recognised as empty cells, NA, N/A, NaN, null, none, nil, a single or double dash and #N/A, in any letter case.
• Per-column count, missing, unique, mean, median, quartiles, minimum, maximum and sample standard deviation.
• Minimum and maximum are found with a loop, so columns of hundreds of thousands of values are safe.
• Pearson correlation matrix of the numeric columns (the first 30 when there are more), using only rows where both values are present.
• Paged preview, 50 rows per page, with a filter and sortable headers, inside a scroll area that works on a phone.
• Chart builder: bar chart of counts, histogram and scatter plot, plus a link to Chart Maker.
• Download of the per-column summary as a CSV file.

numeric  count(numbers) ÷ count(non-missing) ≥ 0.95
std      s = √( Σ(x − x̄)² / (n − 1) )
quantile position = (n − 1) × q, linear interpolation
r(A, B)  = Σ(a − ā)(b − b̄) / √( Σ(a − ā)² × Σ(b − b̄)² )
bins     k = ⌈log₂ n⌉ + 1

• The built-in sample: name, age, score, city for five people, one city blank → gives age: mean 32, median 30, std 6.6708, min 25, max 42. score: mean 79.6, median 79, std 12.341. city: 4 values, 2 unique, 1 missing.

  The correlation between age and score in this tiny sample is 0.42.

• A column of the numbers 1 to 19 and one cell that says abc → gives Type: number. 19 numbers are used, 1 stray value is left out and listed: abc. Mean 10.

  19 of 20 is exactly 95 percent. With two stray values the column would read as text.

• A header row of score, score, city → gives Columns: score, score (2), city.

  Columns are addressed by position, so the two score columns keep their own values and statistics.

• The sample sorted by score, high to low, in Preview → gives Carol 95, Alice 88, Eve 79, Bob 72, Dan 64.

• Download summary CSV for the sample → gives Header: column,type,count,missing,excluded,unique,mean,median,std,min,q1,q3,max.

  Numbers in the download are rounded to six decimals. The excluded column counts stray values.

• Files are read by your browser and parsed by JavaScript in a worker on this page. The contents are not sent to Ulkaa or to any other server. The Excel reader is part of this site and loads from it, not from a third party, and only when you open a workbook.
• Nothing is stored. Reloading the page clears everything. The summary CSV is created on your device when you press the download button.

• If a column you expected to be numeric shows as text, open Profile and read its stray values. A currency symbol, a decimal comma or a word is usually the cause.
• Check the missing count before looking at averages. A mean computed from only 60 percent of the rows can mislead.
• A correlation matrix only covers numeric columns. Encode a yes or no column as 1 and 0 first if you want it included.
• Quote any field that contains your delimiter, such as "1,200" in a comma file, so the row keeps its column count.
• For a very large file, open it instead of pasting it. The paste box is meant for small pastes.
• Delimiter detection looks at the first five lines. Make sure they use the same separator as the rest.