Anonymise a spreadsheet (and put the real data back)
You have the client list, the payment schedule or the attendance register and you would like it analysed, but you cannot hand over names and contacts. Here you decide column by column what leaves the house: names become stable placeholders, so the counts stay right, and at the end the real values go back into the answer.
Put the real values back
Why names can be masked here
This tool's sibling, Anonymise a text, states that it does NOT recognise proper names, and that is true: inside a sentence «Rossi» is indistinguishable from any other word. In a table the question does not arise: the column header already says that column is names. Here names are not guessed, they are read. That is why you decide by column and not by word: if a «notes» column has a name in the middle of a sentence, that column has to go, or be passed to the other tool.
The same client written four ways
This is the fault you cannot see: «Mario Rossi», «MARIO ROSSI» and «Mario Rossi » with a trailing space would become three different clients, the table would look perfect and the analysis would count three people instead of one. Spaces are always stripped. Case is not, by default: «Rossi S.r.l.» and «Rossi Srl» must not be merged without asking. The preview does tell you how many values differ only by spaces or case, so the call is yours.
The placeholders and the return of the values
The placeholder is shaped like the other tool's, [NAME_7], because something written like «Client 7» gets rewritten by the model: it declines it, translates it, sums it up. The numbering follows the order in which the values appear in the file, so the same file always gives the same placeholders. The mapping lives only in this tab: it is not saved and not sent, so closing the page loses it, but redoing the same file brings it back identical.
When there is nothing to put back
If the answer you get talks about averages, totals or trends by month, it names no row: there is nothing to put back, and that is fine. The return of the values matters when the answer cites the rows one by one, for instance «the client with the most unpaid invoices is [NAME_12]». That is why «0 put back» is not a fault, and it is written in the result too.
Pseudonyms, not anonymity
It has to be said plainly: this is pseudonymisation. If the date of birth, the postcode and the amount stay in the table, the person is still recognisable even without the name, and two tables masked with the same mapping can be cross-referenced. When you leave columns like that in the clear the tool writes it down for you, instead of letting you find out later. Do you have an .xlsx file? Open it first with Excel (XLSX) to CSV, or copy the cells from Excel and paste them here.