Comparing contract versions: what to look for beyond the obvious edits

Most contracts open with a block of definitions, and every capitalised term after that points back to it. Rewriting one definition silently rewrites every clause that uses the term. If “Services” is narrowed from “the services described in Annex 1 and any related work” to “the services described in Annex 1”, nothing else in the document looks different — but the scope has shrunk in twenty places at once.

Watch for the opposite move too: a party renamed from “Contractor” to “Supplier” throughout. It looks like a cosmetic edit and produces a large, noisy diff — which is exactly the kind of change that hides one real edit in the middle of forty harmless ones. Read those paragraphs at word level rather than skimming the count.

Numbers are where money lives, and they are easy to skim past because they are short. Compare the fee, the instalment split, the interest rate, the notice period, the term and the renewal mechanics side by side rather than from memory. A payment schedule that moves from “50% on signature” to “40% on signature, 40% on delivery, 20% on acceptance” shifts real cash flow even though the total is unchanged.

Then read the unit next to the number. “Thirty (30) days” and “fourteen (14) business days” are not the same thing, and neither are “days” and “banking days” when a payment has to clear. The same applies to percentages with a period attached: 0.05% per day is roughly 18% a year, and 0.08% per day is nearly 30%.

Insert one new clause in the middle of a contract and every clause after it is renumbered. A paragraph-level comparison will show that shift honestly — a long tail of “rewritten” paragraphs whose only change is the leading number — and that tail is worth two minutes of attention, because cross-references usually do not move with it.

Search the new version for every reference of the form “clause 12.3”, “Section 8” or “the Annex referred to in paragraph 4” and check that each still points where it used to. A liability cap that refers to the wrong clause is not a typo; it is an unenforceable cap. Numbering shifts are also the reason to keep the old version open: you need both numbers to make sense of a reference.

An added sentence draws the eye. A deleted one leaves no trace in the new file at all — which is why a comparison that only reads the new version is useless. The sentences most often removed are the boring protective ones: a survival clause, a carve-out from a limitation, an obligation to notify, a right to audit, a cap on an indemnity.

Deletions inside a paragraph matter as much as whole clauses. “The Client shall pay for work completed up to the termination date” and “work completed and accepted up to the termination date” differ by one word, and that word moves the risk of unaccepted work from one party to the other. Word-level marks inside rewritten paragraphs are what make this visible.

Half of the substance of a commercial contract usually sits in the annexes: the specification, the price list, the service levels, the data-processing terms. They are also the parts most often sent separately, updated without a note, or attached in a different format. Compare them as documents in their own right, and confirm that the annex numbering in the body still matches the files you were sent.

Then find the order-of-precedence clause — the one that says which document wins if the body and an annex disagree. A quiet reversal there (“in case of conflict, the Annex prevails”) can hand control of the scope to whoever edits the annexes.

Read the comparison twice. The first pass is by type: look at every deletion, then every addition, then the rewritten paragraphs — this is where an independent tool with a filter beats reading the document top to bottom, because deletions cluster instead of hiding. The second pass is in document order, so you see how the changes interact: a new clause 9 plus a reference to “clause 9” elsewhere is one story, not two.

Keep the output. Export the change report, attach it to the version you approved, and note the questions you asked about each item. When the same counterparty sends version four, that file is the fastest way to remember what was already agreed — and it is the document you will want if anyone later claims a change was never flagged.