Cross-Document Matrix
Selecting several documents starts one review per document, grouped into a batch. Each review answers "what is wrong with this document?". The matrix answers the question none of them can: across the whole batch, which documents fail which check, and which were never asked.
Every file against every check
Checks run across the top with their severity, files down the side, and each crossing cell says what actually happened to that pairing. Matrix is the grid itself. Insights rolls the same batch up into the portfolio view.
Opening it
Go to the jobs view and open the Completed tab. Any row that belongs to a batch carries a grid icon in its Actions column, labelled View batch matrix. It opens the page at a URL of the form /dashboard/document-review/batch/<batch id>/matrix.
Rows from reviews that pre-date batch grouping have no such action, because they belong to no batch.
The page fills the width: the Files panel is hidden while it is open and comes back when you leave.
The page header
| Control | What it does |
|---|---|
| Back to reviews | Returns to the Document Review jobs view |
| Cross-document findings matrix | The page title. Under it, the file count, the check count and the total findings in this batch |
| Share controls | Share the batch. A batch review is one shared work item, so the grant covers every document in it |
Two tabs sit under the header: Matrix and Insights.
The Matrix tab
A grid with documents as rows and checks as columns. The top-left cell reads Files / Checks. Both the header row and the file column stay put as you scroll, so a wide batch and a tall one are both readable.
Each column header names the check and, under it, its severity as MUST, SHOULD or NICE. Each row header names the file, and adds its processing status when the review did not complete.
Reading a cell
A cell is one document against one check. What it says depends on what actually happened.
| Cell | Meaning |
|---|---|
| HIGH, MEDIUM or LOW with a finding count | The check produced findings on this document. The word is the highest risk among them |
| Passed | The check was evaluated and satisfied |
| N/A | The check does not apply to this document, so it was not assessed |
| Not evaluated | The check was configured, but the analysis returned no verdict. Review this pairing by hand |
| Not configured | This document was never reviewed against this check. It has not passed; it was not asked |
| No record | The check was configured and nothing was persisted. Treat the document as unreviewed for it |
| A dash | Neither a finding nor an outcome is known for this pairing |
The distinction between Passed, Not evaluated and Not configured is the point of the grid. A blank cell would read as "fine" for all three, and only one of them means it.
Hovering any cell names the document, the check and the outcome in words. Clicking a cell that has findings opens that document's review with the finding pre-selected, and its Back returns you here rather than to the jobs list. A cell with nothing behind it is not a link.
Filtering to a set of documents
Arriving from the Insights tab, the grid is narrowed to the documents that bucket named. A line above the grid reads how many of the batch's documents are showing and why, with Clear filter beside it to return to all of them.
When it is empty
A batch with no files or no checks shows No findings yet for this batch, and says the files and checks will appear once a review completes.
The Insights tab
The same batch, rolled up. Where the matrix shows how each document scored, Insights shows how the set scored.
The stat row
Six cards across the top.
| Card | What it counts |
|---|---|
| Documents | Files in the batch |
| Checks | Distinct checks across the batch |
| Findings | Total findings raised |
| Docs missing a must-have | Documents that failed at least one must-have check |
| Checks not evaluated | Configured checks that returned no verdict. Not evidence a clause is present |
| Unsupported findings | Findings whose cited text does not fully carry the claim, per the independent check |
The last three turn colour only when they are above zero.
The two banners
Under the stats, Judicio raises the gaps that a green grid would otherwise hide.
- Never configured pairs. A count of document and check pairings that were never configured in this batch, with the reminder that not being reviewed against a check is not the same as passing it.
- Missing records. A count of configured checks that produced no record at all, with the instruction to treat those documents as unreviewed for that check.
Each banner is shown only when its count is above zero.
Must-have gaps
The portfolio question: which of your must-have checks are failing, and where. Every gap is a card naming the check, how many of the documents reviewed against it were flagged, and how many could not be verified. Up to five affected file names are listed, with a count of any remainder.
| Control | What it does |
|---|---|
| Show flagged documents | Switches to the Matrix tab filtered to the documents this check flagged |
| Show unverified documents | Switches to the Matrix tab filtered to the documents where support could not be confirmed |
When nothing is missing the section says so directly: every must-have check was satisfied on every document that was reviewed against it.
Coverage by check
One card per check, with its severity badge. Each states how many of the batch's documents the check was configured on, plus the counts of unresolved and unsupported findings where there are any.
Under that sits a row of outcome pills, one per bucket the check produced, each with its count. Hovering a pill explains what it counts.
| Outcome | What the pill counts |
|---|---|
| Failing | Documents with at least one finding against this check |
| Passed | Documents where the check was evaluated and satisfied |
| Not applicable | Documents the check does not apply to, so it was not assessed |
| Not evaluated | Documents where the check was configured but the analysis returned no verdict |
| Not configured | Documents that were never reviewed against this check |
| No record | Documents where the check was configured and nothing was persisted |
Clicking a pill filters the Matrix tab to exactly those documents. A pill with a count of zero is not clickable.
Three of these six outcomes say nothing at all about the document: not applicable, not evaluated and not configured. They are deliberately not green, and they are deliberately not blank. If a check matters to you, read its Not configured and Not evaluated counts before you read its Passed count.
If the rollup does not load
The two tabs fetch independently. If the rollup fails, the grid stays fully usable and falls back to showing findings only, without the outcome chips. The Insights tab then offers the error and a retry rather than an empty page.