Survey Comment Redactor
Survey comments made safe to pass on: the names and numbers taken out, and the comments that give their writer away pointed out for you to read.
Reads .xlsx and .csv inside your browser. Nothing is uploaded, which matters more here than anywhere: these are the comments people were promised nobody would trace.
The survey export
Drop a spreadsheet here
One row per response, with the comment columns and anything else the survey exported. Only the columns you choose go into the shared file.
What a pattern can catch, and what it cannot
Email addresses, phone numbers, links and ID numbers have shapes, so they always come out. Names are harder, and are found four ways: from a roster if you give one, after words such as “my manager” or “Dr”, from a list of common first names, and from capitalized words that the file never writes in lower case, which is what a name or a place looks like to something that does not know your people. Those last are guesses, listed first so you can keep the harmless ones.
“As the only woman on the night shift” names nobody and identifies somebody. No pattern can remove that safely, and a tool that rewrote it would change what the person said. So comments that describe their writer, whether as the only one of something, by years of service, by age, by pay or by health, are flagged and left exactly as written. The number of them sits in the headline beside the number redacted, because it is the one that still needs you.
Words that look like names, and are not
Will, Grace, Hope and Mark are names and ordinary words. They count as names only when written as one, mid-sentence and capitalized, so “Will this change?” survives and “I asked Will” does not. A word the file itself writes in lower case anywhere is an ordinary word in this file, so “Leadership” at the start of one comment is not mistaken for a person.
The columns beside a comment give it away too
A comment shown with a department of two is as good as signed. Any column you share has its groups under the minimum combined, and when the combined group is still too small the next smallest joins it, the same protection the Survey Cut Analyzer gives scores. Rows are shuffled so the file’s order cannot be matched to when people responded. A column that names or numbers the respondent is refused outright.
Two files, and only one to share
The redacted file holds the comments and the columns you chose. The review sheet lists what came out of each comment, word for word, so you can check the work; it is for you, and is named so nobody forwards it by mistake. When two people appear in one comment they become [name 1] and [name 2], so the comment still reads, and the numbering starts again in every comment so nobody can be followed from one to the next.
What this does not do
It does not summarize, score or theme comments, and it does not decide what is fair to share. It removes what can be found, points at what only a person can judge, and keeps a record of both.