Why comparing two spreadsheets is harder than comparing two files
Diffing text is a solved problem: line by line, in order, and the answer is unambiguous. Spreadsheets defeat that immediately, because the row order carries no meaning. Export the same data twice from the same system and the rows may arrive in a different sequence, at which point a line-based diff reports that every row changed. It is technically correct and completely useless.
So a spreadsheet comparison has to start by deciding what makes two rows the same row. That is the key column: an id, an SKU, an email address, an invoice number, whatever uniquely identifies a record in your data. Rows are matched on the key, then compared field by field. Choosing the key is the one decision the tool cannot make for you, and choosing it wrongly is the reason most comparisons produce nonsense. If the key is not unique, two different records collide and the differences reported between them are meaningless.
The second trap is numeric equality. Spreadsheets store numbers as IEEE 754 doubles, and values that are equal in arithmetic are frequently not equal in bits. 0.1 plus 0.2 is not 0.3 in any language that uses doubles, including the one your spreadsheet application is written in. Compare two exported price lists with exact equality and you will find "differences" of one ten-billionth of a rupee scattered through them. A comparison that is useful for money needs a tolerance, and the tolerance needs to be visible so you know what it is treating as identical.
Formatting is the third. A cell holding the number 1000 and a cell holding the text "1,000" look the same to a person and are not the same value. Dates are worse: a date is a number with a display format, and the same underlying day can appear as 08/20/2026, 20-08-2026 or 46,258 depending on locale and format. Comparing what is displayed rather than what is stored produces false differences; comparing what is stored without normalising produces the opposite.
All of this runs in your browser, which for this tool is usually the point. The files that people most need to reconcile, bank statements against ledgers, supplier price lists, payroll exports, are exactly the files that should not be uploaded to a stranger to answer a question as small as "what changed".