TL;DR: When you need the same facts from many contracts, extraction beats reading. A Review Matrix asks up to 25 questions across a set of documents and returns a grid where every cell is typed (date, currency, yes/no, summary and six more), confidence-scored, and cited to the exact clause. This guide works three real portfolio jobs - a rent roll from leases, a change-of-control sweep before a deal, and renewal tracking across vendor contracts - and shows how to design questions that come back as clean data.
There is a category of legal work that is not really reading - it is data entry with a law degree. Building a rent roll from forty leases, checking which of two hundred customer contracts have change-of-control clauses, finding every auto-renewing vendor agreement before its notice window: in each case the questions are known in advance and identical for every file, and the output is a spreadsheet. That shape of work belongs in a matrix. Here are three workflows that show how, with the question design that makes the difference between clean data and a wall of text.
When bulk extraction beats reading
The test is repetition. If you are reading one complex agreement to understand it, read it - a matrix will not replace that. But when the same five-to-twenty-five questions apply to every document in a set, sequential reading has three costs a matrix does not: it is slow, it drifts (the fortieth lease is not read like the first), and its output is whatever the reader typed into the spreadsheet, with no link back to the page. A Review Matrix asks every question of every document identically, types every answer, marks its confidence, and cites the clause it came from.
The economics compound with portfolio size. The matrix costs roughly the same effort to set up for ten documents as for a hundred - the marginal document is nearly free, which is exactly the opposite of manual review, where the marginal document costs the same as the first and arrives when the reader is most tired.
Design questions that extract clean data
Extraction quality is mostly decided before the run, in how the questions are written. Four rules cover most of it.
- One fact per question. What is the rent and how does it escalate? is two columns pretending to be one. Split them.
- Ask for what is on the page. What is the annual rent? has an answer in the document; Is the rent above market? does not. Matrices extract; they do not opine.
- Type every answer. Judicio offers ten answer types - text, date, currency, amount, monetary amount, percentage, yes/no, list, summary, and tag. A date column that sorts and a currency column that totals are the difference between data and prose.
- Reserve summaries for genuinely narrative facts. Assignment conditions deserve a summary cell; a commencement date does not.
You do not have to write questions from scratch: Judicio drafts them from the documents you selected or from a plain-English brief, typing a column header like Governing Law prompts a tuned question, and matrix question sets are among the 100-per-tool expert templates. However the questions originate, read them once with the four rules in mind before running.
Workflow 1: A rent roll from a stack of leases
The job: a lender or buyer wants a rent roll, and the source of truth is forty signed leases in the File Library. The matrix:
| Column | Question | Answer type |
|---|---|---|
| Tenant | Who is the named tenant? | Text |
| Premises | What premises are demised? | Text |
| Term start | When does the term commence? | Date |
| Term end | When does the current term expire? | Date |
| Annual rent | What is the annual rent payable? | Currency |
| Escalation | What is the annual escalation rate? | Percentage |
| Break option | Does either party have a break option? | Yes/No |
| Security deposit | What security deposit is held? | Currency |
Run it and the output is the rent roll - sortable by expiry, totalled on rent, every figure cited to the lease page it came from. The cells marked not addressed are findings too: a lease with no escalation clause or no deposit provision is exactly what the lender asks about next.
Workflow 2: A change-of-control sweep
The job: a transaction is coming and you need to know which of the company's contracts can be triggered by it - fast, and provably. The matrix runs across the customer and supplier agreements: Does the agreement contain a change-of-control provision? (yes/no), What does it trigger - consent, notice, or termination? (tag), Whose consent is required? (text), What notice period applies? (text), and Summarise the provision (summary).
The grid that comes back is the consent-and-notice plan for the deal: the yes/no column scopes the problem, the tag column separates the fatal from the administrative, and the summary column - each cell cited to the clause - feeds straight into the disclosure schedule. What used to be the reason diligence sampled the contract population becomes a run measured in hours, covering all of it. This sweep is the same first pass an M&A diligence review runs from the buy side.
Workflow 3: Renewal and expiry tracking
The job: an in-house team inherits two hundred vendor contracts and a history of unwanted auto-renewals. The matrix: When does the current term expire? (date), Does the agreement renew automatically? (yes/no), What notice is required to prevent renewal? (text), What is the annual fee? (currency), Can the agreement be terminated for convenience? (yes/no).
Exported to Excel and sorted by expiry date, the grid becomes the renewals calendar the team never had: which contracts renew silently, what the non-renewal notice window is for each, and what the spend at stake is. Because the answers are typed, the sort and the totals need no cleanup - and because each cell is cited, any figure the CFO questions traces to its clause in one click. Re-run the same saved question set as new vendors arrive, and the tracker stays current instead of rotting.
Review flagged cells, then export
A matrix is a first pass with a built-in triage. Every cell carries a confidence signal - clear, ambiguous, low confidence, or not addressed - and the uncertain cells collect in a review queue so you verify where the model hesitated instead of everywhere. Open the citation behind a flagged cell, read the quoted passage, and either accept, correct, or regenerate that single cell without re-running the matrix. The Insights view rolls the whole run up - files, questions, answered cells, cells needing review - and its confidence heatmap shows where the portfolio is ambiguous rather than merely where the model was.
Then export: Excel or CSV for the model, the tracker, or the client's deal team; Word or PDF with citations for the report. Anything you will rely on commercially deserves its citation opened first - the grid does the cross-referencing, and you supply the judgment about what the answers mean. Outputs are not legal advice.
How Judicio helps: extraction with receipts
Judicio's Review Matrix is built for exactly these workflows: up to 25 questions per run across multiple documents, ten answer types, per-cell confidence and citations, a review queue for flagged cells, single-cell regeneration, chat with the grid, an Insights roll-up, and exports to Excel, CSV, Word, and PDF with citations included. Files upload once into the File Library - with automatic OCR for scanned contracts - and the same set feeds Document Review when a flagged agreement needs a full checklist read. Question sets live in the shared template library, so a portfolio job you designed once becomes a workflow anyone on the team can run.
See the due diligence and in-house counsel solution pages for where these extractions sit in the wider workflow.
Getting started with Judicio
Pick the portfolio question your team keeps answering by hand - the rent roll, the renewal list, the change-of-control check - and build it as a matrix this week. Run it on a batch you know well, verify the flagged cells against their citations, and compare the grid to the spreadsheet you built last time.
The 7-day free trial includes 500 credits, no credit card required; Professional access is $200 per month for 5,000 credits. Explore the features or contact us to design your first matrix together.