The same question comes up on this forum every few weeks: someone imports a long Word document into Confluence Cloud, gets one endless page, and asks how to turn it into a proper page tree, one page per chapter, sections underneath. Several of those threads are closed now, so here is the whole answer in one place: what Cloud does, the free way around it, and what that way really costs on a big document.
What the Cloud importer actually does
The built-in Word import in Cloud is better than its reputation. It reads .docx, keeps headings, tables, lists and images, and it takes batches of up to 30 files at a time. Two behaviours catch people on the first run.
- One file becomes one page. Not one page per chapter. The whole document, however long, arrives as a single page with the headings inside it.
- Batches land in a holding pen. Import several files at once and Confluence creates a restricted placeholder page called "Imported pages" with the date and time, and puts everything under it. Moving the pages into your real tree is a separate job, by hand.
Splitting by heading levels exists in Confluence Data Center: the import screen there offers "one page" or "multiple pages" built from heading levels. Cloud has no such option, and as of today there is no setting that turns it on.
The free route: split the document in Word first
Word has carried a feature called master documents for decades. It cuts a long document into separate files along heading boundaries, which is exactly the shape Confluence wants.
- View, then Outline. The Outlining tab appears in the ribbon.
- Set Show Level to the level you want to split on, usually Level 1 or Level 2.
- Select the headings that should each become a page.
- Show Document, then Create. Grey boxes appear around each selected heading: those are subdocuments.
- Save into an empty folder. Word writes one .docx per subdocument, named after its heading.
- Import the folder into Confluence, 30 files per batch.
After each batch, open the "Imported pages" placeholder in the page tree, drag its children under the right parent, and delete the empty placeholder. Do it batch by batch: if you import everything first and sort later, a 100+ page document leaves you with several flat piles that look alike.
Where this route costs you
For a handful of documents it is the right answer. On a long document with a deep outline, three things wear on you.
- The hierarchy is gone. Word gives you a flat set of files. Chapters with sections under them arrive as one flat layer, and the parent and child structure is rebuilt by dragging pages one at a time. With two or three heading levels you repeat the split for each level.
- Cross-references break. "See section 4.2" pointed at a bookmark inside one file. Once section 4.2 is a different page, the phrase still reads fine and links nowhere. Nothing warns you.
- Shapes do not survive. Shapes drawn in Word come through as placeholders. A process document full of small flow diagrams drawn with Word shapes is where this hurts most.
For cross-references, the reliable fix is relinking by hand once the pages exist: each "See section 4.2" becomes a link to the page, or to a heading inside it (in Cloud, the page link plus #Heading-text opens the page at that heading). Do the relinking last, after page titles are final, or a rename makes you touch the links twice.
How long it takes on a real migration
A typical case: 150 documents on a shared drive, 20 of them long enough to deserve splitting.
- Split 20 long documents in Word, about 10 minutes each: around 3 hours 20 minutes.
- Import 5 batches of 30 files: around 25 minutes.
- Move roughly 300 resulting pages into the tree, 20 seconds each: around 1 hour 40 minutes.
- Fix broken cross-references and shape placeholders: 2 hours, optimistically.
Call it a day and a half. For a one-time move that is entirely reasonable. The real risk is different: nobody has that day and a half spare, so the migration gets postponed and the documents stay on the shared drive another year.
Which route to pick
A dozen short documents: import them as they are, one page each. The built-in importer does that well. A few long ones: use the Word outline route, it is cheap and you control every cut. A whole department moving off a shared drive: do the arithmetic above first, then decide whether an app is worth it.
Disclosure: I build Confluence apps, and one of them, Word Import for Confluence, does the split during import. It reads the .docx in the browser and builds the parent and child page tree from the heading level you pick, with tables, lists and images. It does not rebuild Word cross-references, so the relinking above still applies. Marketplace listing.
