Learn → Comparing two files

Comparing two files

"What changed since last month?" is the question a review usually turns on, and it is the one a spreadsheet is worst at answering. Prism matches rows by a key column and reports what was added, removed, and changed — while ignoring the reformatting that makes a naive comparison useless.

How matching works

You choose a key column — invoice number, employee ID, account code — and Prism pairs rows between the two files on that key. The result is four buckets:

• Added — keys present in the new file only
• Removed — keys present in the old file only
• Changed — keys in both, where at least one field differs
• Unchanged — the rest, which is usually most of the file

Why a plain text comparison fails

This is the part that decides whether a diff tool is usable on real exports.

Two exports of the same report, a month apart, routinely differ in ways that carry no meaning. An amount comes out as 1,200.00 in one and 1200 in the other. A date is a real date cell in one file and its text form in the other. Nothing changed — but a character-by-character comparison reports thousands of differences, and the dozen changes you actually needed to see are buried somewhere inside them.

Prism compares values, not their formatting:

• "1,200.00", "1200" and "$1,200" are the same number, so they do not count as a change.
• Two representations of the same calendar day are equal, even when one carries a time component the other lost on export.
• Leading and trailing whitespace does not, by itself, constitute a change.

In our test fixtures, thirty formatting-only rewrites are correctly reported as unchanged while the real deltas — seven added, five removed, twelve changed — come through exactly.

Duplicate keys are reported, not swallowed

If the key column contains duplicates — precisely the defect the quality engine checks for — then matching on it is ambiguous. A naive implementation keeps the last row for each key and silently compares the wrong pair of records, producing a confident, wrong answer.

Prism excludes duplicate and blank keys from matching and tells you about them. A comparison that quietly compared the wrong rows would be worse than no comparison at all, because you would believe it.

Controls

• Ignore columns — exclude fields that always change and never matter, such as an export timestamp or a row number, so they stop drowning the result.
• Swap direction — flip which file is "before", without reloading either.
• Export the diff to CSV — the changes as a file, for a working paper or to send to whoever produced the data.

Two ways to run it

If you are trying the app on a sample dataset, the in-app Compare view is demonstrated against a built-in "corrected" version of that sample, so you can see the feature working before buying anything.

Typical uses

• Period over period. This month's trial balance against last month's — what moved, what appeared, what disappeared.
• Before and after a client's corrections. You sent a list of issues, they sent back a revised file. Confirm they changed what they said they changed, and nothing else.
• Two systems that should agree. An export from one system against an export from another, matched on a shared reference.
• Checking your own edits. Compare a reviewed copy against the original before it goes anywhere.