Search is the feature buyers use most and evaluate least. On a demo it always works, because the demo room is full of clean, digitally created PDFs. A real diligence room is different: decades of signed agreements, scanned leases, faxed consents and spreadsheets with forty tabs.
This guide explains what happens between upload and search result, where documents drop out, and how to find out whether a given room handles the messy files you actually have.
How does full-text search work in a data room?
Every file goes through the same short pipeline. The platform stores the file, pulls out whatever text it can, adds that text to an index, and links each indexed word back to the document, the folder and the permission groups allowed to open it. When someone searches, the platform looks up the index and filters the results by what that person is entitled to see.
From upload to search result: where documents fall out of the index
The file lands in its folder. Nothing is searchable yet.
Seconds
Native PDFs, Word and Excel give up their text layer directly.
Usually automatic
Scans and photos have no text layer. Without OCR they stay invisible to search.
The weak link
Words are stored against the file, the folder and its permissions.
Minutes on large batches
Results are filtered by what the searcher is allowed to open.
Must respect groups
What goes missing when stage 3 is skipped
Scanned contracts
Signed copies scanned to image PDF
Phone photos
Board minutes or permits photographed on site
Embedded content
Text inside images, charts or email attachments
The pipeline is invisible when it works. It becomes visible when a buyer’s lawyer types “change of control”, gets twelve hits, and later finds the thirteenth contract by hand, scanned and unindexed, in a folder nobody thought to open.
Why do scanned PDFs go missing from search?
A PDF can be two very different things. A digitally created PDF, exported from Word or a contract system, contains a text layer: the letters exist as characters. A scanned PDF is a picture of a page. To a search index it is a blank sheet.
OCR converts that picture into characters. Platforms handle it in one of three ways:
| Approach | What happens to a scanned file | What the buyer experiences |
|---|---|---|
| OCR on upload, automatic | Text is recognized as the file is processed | Scans appear in full-text results like any other file |
| OCR on request | An admin triggers recognition per file or folder | Works if someone remembers to run it |
| No OCR | The scan is indexed by file name only | Scans are found only by browsing or exact titles |
Ask which of these applies, and ask whether it covers files already uploaded before a setting was switched on. Some platforms only process new uploads.
OCR quality also varies. Clean office scans recognize well. Skewed pages, handwriting, stamps across text, low-resolution phone photos and tables with thin lines produce errors that make search unreliable even when OCR ran.
Which file types does a data room actually search inside?
PDFs are the easy case. The harder cases are the files a finance or legal team produces in volume:
| File type | Text usually searchable? | Watch for |
|---|---|---|
| PDF with text layer | Yes | Password-protected PDFs may be skipped |
| Scanned PDF or image | Only with OCR | Quality of the scan |
| Word documents | Yes | Tracked changes and comments may or may not be indexed |
| Excel workbooks | Often, cell text | Hidden sheets, formulas and very large files |
| PowerPoint | Usually slide text | Text inside images and charts |
| Email files (.msg, .eml) | Varies | Attachments inside the email are a separate question |
| Zip archives | Rarely | Usually need extracting before upload |
Spreadsheets deserve a separate check, since a model may be converted for viewing in ways that change what is indexed. Our guide to spreadsheets in a data room covers viewing and download rules for Excel models.
Does search respect permissions?
It must, and it is the most important thing to verify. The index is a separate structure from the folder tree. If it is built or filtered incorrectly, a search can reveal a file name, a folder name or a snippet of text from a document the searcher has no right to open.
Even a file name can leak a deal. “Project Falcon, Buyer B revised offer.pdf” tells a rival bidder more than the seller intended. That is why the go-live test in our four-step launch guide includes searching from a dummy guest account for a term you know sits in a hidden folder.
Check snippets as well as titles
Some platforms show a line of matching text under each result. Confirm that snippets are generated only from documents the guest can open, and that a file hidden from a group does not appear as a greyed-out result.
What search features matter beyond a keyword box?
Once the basics hold, these features separate a usable search from a frustrating one:
- Filters by folder, file type, upload date and uploader. Useful when a buyer asks “what was added this week?”
- Search within a document in the viewer, with hits highlighted, so a reader can jump to the clause.
- Index number search. Typing “4.2.3” should take you to that folder or file.
- Exact phrase and exclusion operators, for legal teams hunting specific wording.
- Saved or recent searches, so a reviewer does not retype the same terms every morning.
Some modern rooms now add AI-assisted search that answers questions in plain language. That is useful, but it sits on the same index: if OCR never ran, an AI assistant cannot read the scan either. Our reviews note which providers list AI tools among their features.
How do you test search during a trial?
You need a trial room, a few real or realistic files, and a second account set up as a guest. Use documents you know well so you can tell when a result is missing.
- 1
Upload a mixed batch
Include a native PDF, a scanned signed contract, a phone photo of a page, a Word file with tracked changes, an Excel workbook with a hidden sheet and an email with an attachment. Wait for processing to finish.
- 2
Query 1: a word only in the scan
Search for an unusual word that appears only in the scanned contract. If it returns nothing, OCR is off, on request only, or failed.
- 3
Query 2: a figure inside the spreadsheet
Search for a label or number from the visible sheet, then one from the hidden sheet. Note which return results and whether hidden content is exposed.
- 4
Query 3: a phrase from the email attachment
This shows whether attachments inside email files are indexed, or only the message body.
- 5
Query 4: a hidden-folder term as a guest
Log in as the guest and search for a word that exists only in a folder the guest cannot open. Expect zero results, no file names and no snippets.
- 6
Query 5: an index number
Search for a folder number such as 2.1. A good room jumps to it; a weaker one treats it as text.
Write down the results for each platform you trial. Five queries are not a benchmark, but they reveal the gaps that matter on a real deal far better than a polished demo does.
How can sellers make documents easier to find?
Even with good OCR, preparation helps:
- Scan at 300 dpi or better, straight and in black and white or greyscale, before uploading.
- Run OCR before upload if the platform does not, using the desktop PDF software most teams already own.
- Name files descriptively: date, counterparty, document type. “2024-03 Acme MSA signed.pdf” beats “scan0042.pdf”.
- Split bundles. One 400-page PDF of “all customer contracts” is hard to search and harder to cite in Q&A.
- Number the index so questions can reference locations precisely.
These steps cost an afternoon and save every reviewer hours.
Does search quality affect our scores?
Search is part of how we judge deal workflow and everyday usability, two of the six criteria in our scoring model. We weigh what providers document about indexing, OCR and permission-aware results, and what a buyer can verify in a trial. Our review method explains the weights.
Frequently asked questions
What is OCR in a data room?
OCR (optical character recognition) converts scanned pages and images into searchable text. Without it, scanned documents in a data room can only be found by file name or by browsing folders.
Can you search inside Excel files in a data room?
Often, yes: many platforms index cell text in workbooks. Check whether hidden sheets are indexed, since that can expose content a seller meant to keep out of view, and whether very large files are processed fully.
Why does my data room search miss documents I know are there?
The most common reasons are scanned files without OCR, password-protected PDFs, files still processing after upload, or search operators that do not match how the platform parses phrases.
Can guests find files they are not allowed to see through search?
They should not. Search results must be filtered by permission group, including file names and text snippets. Test this from a dummy guest account before inviting real users.
Do AI search tools fix missing OCR?
No. AI assistants in a data room read the same extracted text as keyword search. If a scan was never converted to text, an AI tool has nothing to work with.