A useful comparison starts with equivalent passages and ends with a question worth investigating.
A difference report can tell you that words moved. It cannot, by itself, tell you that the business changed. The research task is to establish what the new language supports, what the surrounding document qualifies, and what additional evidence would change your interpretation.
The SEC’s guide to reading a 10-K identifies distinct sections for the business, risk factors, management’s discussion, and financial statements. Those sections serve different purposes. A sentence in a risk discussion should not be treated as if it were a reported financial result.
Start with a comparison you can defend
Record the company, form, reporting period, filing date, and source link for both documents. Keep an amendment separate from the original. Match equivalent sections before comparing individual paragraphs: moved headings, reordered notes, and changes in document extraction can otherwise look like substantive additions or deletions.
For a first review, choose one research question. “What changed in the service model?” creates a more useful reading boundary than “Summarize everything new.” Read the relevant section in both periods before asking an automated comparison to identify candidate changes.
Keep the passage wider than the highlight
A short highlight helps a reviewer find the change. The surrounding paragraph helps the reviewer understand it. Preserve both. If the sentence includes a qualification, a time horizon, or a reference to another section, keep that material attached to the observation.
A change in the described revenue mix does not, by itself, establish that revenue recurs, that margins improved, or that demand accelerated. Treat each as a separate research question. The comparison should retain enough context to distinguish a change in wording from a change in reported business activity.
A helpful comparison record contains:
- The prior and current wording, with a stable source reference for each.
- A short description of the literal change.
- The possible research implication, explicitly labeled as an interpretation.
- An alternative explanation and the next source to inspect.
Give deletions their own review
Removed language can be easy to overread. A deleted sentence might reflect a resolved issue, consolidation elsewhere in the filing, or a change in how the company organizes disclosures. Search the current document for the same concept before concluding that management has stopped discussing it.
When a table changes, check units, period length, column definitions, and footnotes. A clean textual difference can conceal an inconsistent denominator. Keep the original table available; do not let a generated prose summary become the only record of the numbers.
End with a decision about further work
Every candidate change does not need a place in the final brief. Ask whether it alters a question your team is already tracking, introduces a new uncertainty, or deserves monitoring next period. Record why you retained it. That small decision makes the next review more coherent.
Our preferred handoff is a concise change note with a visible evidence trail, rather than an unranked inventory of edits. The note should make it possible for another researcher to reproduce the comparison without trusting the summary’s phrasing.
Gatani’s filing intelligence approach organizes this workflow around source-linked comparisons and researcher review. The Gatani Lens research workspace connects reported Microsoft results to source summaries and unresolved research questions.
This editorial collection was published on . Archive dates organize the series.