The Chronology Is the Deliverable

The chronology is the deliverable

Ask any litigator what they build first on a document-heavy matter and the answer is a chronology. Ask what tool they use and the answer is usually a spreadsheet, assembled by hand, over weeks, and stale within days of being finished.

This is the single most valuable artefact in a document case and the one least served by the software sold into it. Document review tools are built to decide whether a document is responsive. A chronology asks a different question: what happened, when, according to whom, and how do we know?

What a working chronology contains

Not a list of documents in date order. That is an index. A chronology has, for each entry:

• The date, and whether it is the date of the event or the date of the document recording it. These differ, and conflating them produces errors that survive to trial.
• What happened, in one neutral sentence.
• Who says so — the source, cited precisely enough to open.
• Confidence. Is this established, asserted by one side, or inferred?
• Whether it is disputed, and by whom.

The confidence column is what separates a chronology that helps from one that misleads. A timeline mixing established facts with one party’s assertions, presented in the same typeface, will eventually be relied on as though it were all the first kind.

Where the errors come from

Document date versus event date. An email of 14 March describing a meeting “last Tuesday” is evidence about 8 March. Automated extraction routinely files it under the fourteenth.

Time zones and formats. Systems, servers and people record time differently. On a case where sequence matters — who knew what before what — this is not pedantry.

Duplicates presenting as corroboration. The same email in four custodians’ mailboxes is one fact, not four. A chronology built without deduplication overstates its own support.

Attachments separated from their parents. A document detached from the message that sent it loses its date, its author and its purpose.

Inference recorded as fact. The most consequential error. Someone writes “X knew about the defect” when the source says X was copied on an email whose attachment mentioned it. By the time it reaches a brief, the inference has hardened.

Where automation genuinely helps

Extraction of dates and entities at scale, as a first pass to be corrected rather than an output to be trusted.

Sequencing across sources — correspondence, transcripts and filings on one timeline is mechanical work that a machine does quickly.

Finding every mention of an event across the corpus, so the chronology entry cites the best source rather than the first one found.

Flagging conflicts. Two documents dating the same event differently is exactly what you want surfaced — that conflict is often the case.

Keeping it current. A chronology that updates as documents arrive stays useful; one rebuilt by hand each time does not, which is why most are abandoned mid-matter.

Where it does not help

Deciding what belongs. A complete chronology of everything is unusable; the judgement about which events matter is the work.

Characterising an event neutrally. “Defendant refused” and “defendant did not respond by the deadline” describe the same emails and are not the same entry.

Distinguishing assertion from established fact — an extraction model has no view about that, and will present both identically unless a person marks them.

The rule that makes it defensible

Every entry cites a source you can open in one action. Document, page, and — where the source is testimony — line.

This matters for three reasons. Someone will challenge an entry, and the citation ends the argument. The chronology outlives the person who built it, and an entry whose source cannot be found is an entry nobody trusts. And when the chronology becomes a brief, every fact in it needs a record citation anyway — so building it that way from the start is not extra work; it is the same work, done once.

Linking documents and testimony

The chronology usually starts from documents and ends up needing testimony. A witness explains a document, contradicts it, or supplies the date it lacks.

Keeping both in one addressable system removes a category of manual cross-referencing that is otherwise done from memory: following an exhibit from its production through every witness who was asked about it, with each answer citable. For the testimony half of that method, see deposition analysis is a testimony problem.

Where Lawnova PDF fits

Lawnova PDF indexes documents and transcripts together with page and line precision, so a chronology entry can cite either and the citation opens. Search returns surrounding context rather than an isolated hit, which is what distinguishes a usable entry from a misleading one, and it will run against a local airgapped model where the material cannot leave your environment.

The extraction is a first pass. The judgement about what happened, and how confident you are, stays with the person whose name goes on the brief.