How to compare two versions of a lease in Word, and where it fails

A counterparty sends back a clean lease draft with a note saying they made only minor changes. You have their prior version and this one. The question is what actually moved, and the honest answer is that you cannot know without comparing the files.

Word will do it. It does it imperfectly, in ways worth understanding before you rely on the output.

Running the comparison

In Word, go to the Review tab, choose Compare, then Compare again from the dropdown. Set the original document to your last version and the revised document to the one you just received. Under More, confirm that comparison is set to Word level rather than character level, and that Comparison settings include Formatting only if you actually want formatting noise. For a lease, turn formatting off. You are looking for words.

Word produces a third document showing insertions and deletions as revision marks. Save it. That comparison document, not the clean draft, is what you read.

Two settings people miss. First, "Show changes in" should be set to a new document, not to the original, so you do not contaminate your own copy. Second, if either file has unaccepted tracked changes already in it, Word will accept them silently before comparing, which means the comparison reflects a document neither party has actually agreed to. Check both files for existing markup before you start.

Where the output misleads

A generated comparison is an algorithm's guess about what corresponds to what. It is not a record of what anyone did. That distinction produces four recurring problems.

Moved text reads as a deletion plus an insertion. When a clause is relocated from Article 12 to Article 19, the comparison shows the whole clause struck in one place and inserted in another. Nothing changed, but you now have two large marked blocks to read. Worse, when a clause is moved and edited, the substantive edit is buried inside a block that looks like a pure relocation, and it is very easy to wave through.

Renumbering swamps everything. Insert one new subsection early in the document and every subsequent number changes. The comparison shows hundreds of marks, and the four that matter are somewhere inside them. This is the single biggest reason comparison output gets skimmed rather than read.

Formatting noise. Style changes, list restarts, and font substitutions can generate marks throughout a document that is substantively identical. Turning formatting comparison off helps and does not fully solve it.

Tables and exhibits compare badly. Rent schedules in tables and site plan exhibits frequently produce either garbage marks or no marks at all. If the rent table looks unchanged in a comparison, verify the numbers by eye anyway.

The four checks the comparison will not do for you

Whatever tool produced the marks, these four failures are not the kind a diff can catch.

Exhibits compared separately. If the guaranty, work letter, or rules and regulations are separate files, they need their own comparison. A revised good guy guaranty produces no marks in the lease body at all, and the guaranty exhibit is where a landlord's counsel can widen personal exposure without touching a page anyone is reading closely.

Defined terms checked against every use. A comparison shows you that the definition of Operating Expenses changed. It will not tell you that the change interacts with the gross-up provision in Article 5 and the expense cap in Article 6, both of which now mean something different while showing no marks.

Cross references verified. A deleted subsection leaves behind references to it. A surviving clause that says "subject to Section 8.4(c)" where 8.4(c) no longer exists is unenforceable in the direction you were relying on, and no comparison flags it.

Your own open requests reconciled. The comparison shows what they changed. It does not show what you asked for that they quietly ignored. That check has to be run against your own list, and it is the check people skip.

Why tracked changes beat a generated comparison

When a counterparty edits with track changes on, the revision marks are stored in the file as a record of what was actually done, attributed to an author with a timestamp. There is no guessing, because nothing was inferred. A moved clause is a move. A renumbering is a renumbering. The signal to noise ratio is dramatically better, and you can trust that the marks correspond to human decisions.

This is worth asking for explicitly, early, and in writing. "Please return the draft with track changes on" is a normal request in a lease negotiation and almost everyone honors it. The rounds where someone sends a clean document are the rounds worth the extra scrutiny, which is a useful signal on its own.

Doing it without the manual pass

The reason redline review takes an evening is that the comparison tells you what changed and nothing else. Someone still has to decide which of the 86 changes matter, look up what the prior language said, and remember how the same clause resolved two rounds ago.

CRE Redline reads the tracked changes directly out of the .docx revision marks rather than generating a diff, so the change list cannot contain an edit that is not in the file. Each change comes back with the deleted and inserted text shown verbatim, a plain English reading of what it does and who it favors, and a materiality rating, so the abated rent clawback sorts above the cross reference fix. Contested clauses carry their history forward, which is what turns round four from an archaeology exercise into a list.

The first round of every deal runs free, which is the honest way to find out whether it works on your own lease. For the broader method, see how to read a commercial lease redline.