No new tool, no new account
Authentication is Microsoft Entra ID and the content sits in the organization's own SharePoint. There is no separate platform for a reviewer to be onboarded into.
Reviewers
There are more reviewers than authors. Design for them
A content operation is usually counted by its authors, and the authors are the smaller group. For every subject matter expert drafting content there are several people who read it, comment on it and send it back. If review is the step that does not fit the tooling, it becomes the step that sets the pace for everything else.
CommentedOn the component, not the document
No new tool Word, or the SharePoint library
The structural problem
Review is where a document model stops scaling
When the unit of work is a finished document, every reviewer receives the whole thing, whether or not their part of it changed. They read past the sections that are not theirs, comment in whatever tool the file arrived in, and the responses come back to be reconciled by hand. Nothing about that gets easier as the portfolio grows.
When the unit of work is the component, review narrows to what actually changed. The comment attaches to the component rather than to a page number in one rendering of it, and the same approved component carries that history into every document it appears in.
What changes
The reviewer keeps their own tools
DitaExchange runs inside the Microsoft 365 the organization already has. A reviewer signs in with the same Entra ID account, opens content in Microsoft Word or in the SharePoint library, and is not asked to learn an XML editor to take part.
Comments attach to the component
Feedback lands against the unit of content it concerns, so it stays with that component through every later assembly rather than being tied to one rendered output.
Review only what moved
Component-level version history makes the changed unit identifiable, so a review cycle does not require re-reading the parts that did not change.
Frequently asked questions
Where does a review request actually turn up?
In the places work already arrives. Dx5 is available as a Microsoft Teams app, so review and approval tasks come through Teams notifications, and tasks also route through Outlook, plus Planner where the organization uses it. There is no separate queue somebody has to remember to open, which is usually the difference between taking part and being chased.
My comment is about how the assembled document reads, not about one paragraph. Where does that go?
Both readings exist. Publishing produces the assembled output in Word, HTML or XML, so a reviewer who has to read the whole thing can read the whole thing. The distinction that matters is where a comment has to land to change something: a remark about one statement belongs on the component that carries it, because that is the unit somebody will edit and approve.
Can the content change while I am still reading it?
Snapshots and approval states exist for that. A snapshot fixes a version of the content at a point in time, so a review or an inspection reads against something that is not moving. Version comparison then shows what changed between revisions, which is how the next cycle starts from the change rather than from the first page again.
Does anything catch the mechanical errors before they reach me?
That is what DxChecker is for. Rules cover approved and banned terms, heading and outline structure, date formats, citation syntax, required metadata, and internal and external links. DxChecker for Word runs against a document in Word 365, and DxChecker Server runs a ruleset across a folder or a SharePoint library on a schedule. Reviewer attention is better spent on judgment than on formatting.
The same text appears in several documents I review. Do I read it once or every time?
Once, when it is one component rather than several copies. A safety warning used in twenty manuals is one approved component referenced twenty times, and where-used reporting lists every document that carries it, so the review you give it applies to all of them. Copies are what create the repeated reading. The component model is what removes the copies.
Start with the review cycle that slows everything else down
Pick the content set where review is the bottleneck rather than the drafting. That is usually where a component model pays for itself first.
