Layer 1: the text
A photo or scan of a page is just a picture. Optical character recognition (OCR) turns the picture into text a computer can search. Without it, the word "Hendricks" in a 1958 letter is invisible to search.
OCR is reliable on clean typed and printed pages. It's much less reliable on handwriting, faint carbon copies, and damaged paper — more on that below.
Layer 2: the details
Full-text search finds words. But many questions are about kinds of records and when they're from:
- "All board minutes from the 1970s."
- "Every letter that mentions the Hendricks family."
- "Deeds and property records, any year."
Answering those needs structured details on each page: dates, names of people and organizations, and the document type. Archivists call this metadata. In small organizations it's usually the part nobody has time to type in — which is why Boxchive extracts dates and names and classifies each page by document type automatically.
Layer 3: the filing
Consistent filenames and folders mean results make sense when you see them, and the archive
still makes sense outside any one piece of software. A result called
1962-03-14_board-minutes_p02.jpg tells you what it is before you open it;
IMG_4021.jpg doesn't. Boxchive files pages under your organization's existing
naming convention — see file naming conventions
if you don't have one yet.
What searching looks like in practice
| The question | What makes it answerable |
|---|---|
| "Did the board ever vote on selling the north lot?" | OCR text of the minutes + document type "minutes" |
| "What do we have from 1962?" | Extracted dates on each page |
| "Every mention of a donor's family name" | OCR text + extracted names |
| "The original deed of gift for this object" | Document type + names + consistent filing |
Where search quietly fails: handwriting
The most dangerous search result is an empty one that's wrong. If a handwritten register was never transcribed, a search for a name in it returns nothing — and it looks exactly like the name isn't there.
Boxchive flags handwritten, faded and damaged pages and has them human-reviewed, and the archive records which pages were transcribed confidently and which weren't. That way you know what a search did and didn't cover. More in handwriting OCR for old records.
A checklist for any "searchable archive" claim
Whether you're evaluating a vendor, a software product, or your own scanned files, ask:
- Can I search across the whole collection at once, not file by file?
- Is there real OCR text behind every page?
- What happens to handwritten pages — and will I know which ones weren't read?
- Can I filter by date and document type?
- Are files named in a way I'd understand without the software?