← Back to Marketing Analytics Consultants

Case study · Large-scale document analysis

Ten years of records, read together for the first time

Over ten thousand documents across a decade and three states. A succession of attorneys. Well into six figures in fees before any of this began. The problem this page is about is not what the matter concerned — it is that nobody had ever read the whole record at once.

1. The structural problem, not the legal one

A family estate and care matter running more than a decade across three states. Over ten thousand documents — court filings, medical records, financial statements, correspondence and notes spanning eight years of events. A succession of attorneys across the period. Well into six figures in fees before any of this began.

The problem this page is about is not what the matter concerned. It is that in ten years, nobody had ever read the whole record at once.

10,263documents analyzed
8 yrsof events covered
12 yrsof litigation before the review
3states involved

2. Why long matters get read badly

Figure 1

Sequential reading, and what it cannot see

How a long matter is normally read Counsel 1 Counsel 2 Counsel 3 Counsel 8 Each inherits a summary of the last one’s work, not the record itself. What full-corpus review does instead All of it, on one timeline, at the same time A contradiction between a 2014 note and a 2019 record is invisible unless both are in view at once.

This is not a criticism of anyone's counsel. Successive attorneys were engaged on this matter precisely because the volume was beyond what any one of them could hold. And nobody bills a client for re-reading a decade of file they did not create.

The specific thing that gets lost

Contradictions across record types. A medical note and a financial statement four years apart, from different sources, held in different folders, that cannot both be true. Nobody is looking for that, because looking for it means holding the whole record in view simultaneously — which no reader can do and no billing structure would pay for.

3. What the review actually did

StepWhat it produces
Full-corpus ingestionEvery file structured and readable at once, in batches, with nothing summarized away
Timeline reconstructionEvery dated event from every record type on a single chronology
Entity and relationship mappingWho appears where, in what capacity, and when their role changed
Cross-document contradiction detectionStatements from different sources that cannot both be accurate
Citation-ready extractionEach observation tied back to the specific document and page it came from
Thematic indexingEvery document tagged against the questions actually at issue
What it produced, stated carefully

Connections across the record that ten years of sequential review had not surfaced — specifically, relationships between medical, financial and procedural events that are only visible when all three are laid on one timeline. Those are leads. Every one of them then had to be verified against the source document, and some did not survive that check.

4. What this would have cost to buy

Figure 2

What conventional discovery would have required

To reach the same picture across 10,263 documents by conventional means Specialist attorneys 15 to 20 Elapsed time 5 to 8 years Cost $2.5 to $4 million Twelve years of litigation and a succession of attorneys had already run against this record without reaching that picture — the volume was beyond what any one of them could hold. The estimate carries its own caveat: even then, the key material might never have surfaced.

These figures are the estimate prepared for the matter itself and set out in writing to opposing counsel. They are an estimate rather than an invoice — but they are the estimate the party relying on the analysis was willing to put in front of the other side.

Review is consistently reported as sixty to eighty percent of total spend in document-heavy matters. That is not because reading is expensive. It is because a corpus this size cannot be covered by one reader at all — which is why the conventional answer is a team of fifteen to twenty specialists and half a decade.

The part money cannot buy

Even at four million dollars and eight years, a team of twenty reads in parallel rather than together. A contradiction between a note in one year and a financial record five years later is caught only if a single reader holds both — and on a corpus this size no reader holds more than a fraction. That is why the estimate ends where it does. Full-corpus analysis is not merely cheaper. It does a different thing.

5. The limits, stated plainly

What this isWhat it is not
Organization, chronology and cross-referencing at a scale a person cannot holdLegal advice. We are not attorneys and do not practice law
Leads, each tied to a specific pageConclusions. Nothing is established until a person verifies it at the source
A record a professional can work from efficientlyA substitute for counsel, or for a court’s findings
Contradictions surfaced for a human to assessA determination that anyone did anything wrong
The rule the work runs on

Every observation carries a citation to the document and page it came from, and anything that cannot be cited does not appear in the output. That single discipline is what separates a useful analysis from an expensive way to generate plausible-sounding claims.

Details of the underlying matter, the parties and the jurisdictions are not published here and will not be. We work privately.

Marketing Analytics Consultants

Have a record too large to read?

Litigation files, claim histories, audit trails, correspondence archives, board records. We organize, chronologize and cross-reference at a scale a person cannot hold — and everything comes back with a citation.

Prefer email? info@marketinganalyticsconsultants.com