What this article is about
Type "Amazon" into search and TaxItEasy finds the scanned receipt where "Amazon" only appears inside the document — even though every document is stored encrypted. This article explains how that works in plain English, what exactly is searchable, how queries behave, and what's different on mobile.
Searching without storing readable text
Full-text search and encryption usually fight each other: a classic search index is a readable copy of your documents' text, which would undo the point of encrypting them. TaxItEasy takes a different route:
- Instead of storing words, the index stores one-way cryptographic fingerprints (keyed hashes) of the words in each document.
- The fingerprints are keyed with a secret key unique to your company — the same word produces different fingerprints for different companies.
- When you search, your search terms are fingerprinted the same way and matched against the index.
The result: TaxItEasy can find "Amazon" inside a scanned receipt without keeping the receipt's text readable in the search index — and the index is useless to anyone without your company's key. Fingerprints are irreversible; there is no way to run the process backwards and reconstruct the text.
What is searchable
- Everything the scan extracted — vendor names, invoice numbers, line-item descriptions, and the rest of the structured data pulled from each document.
- The filename, matched separately from the content.
Results are annotated by where the hit came from: filename, content, or both.
How queries behave
- Search-as-you-type from 3 characters. Word prefixes are indexed, so "hetz" already finds "Hetzner".
- Multi-word queries must match all words — "hetzner march" finds documents containing both, not either.
- Accents and umlauts are folded — searching "muller" finds "Müller".
- Very common filler words are excluded from the index, so they neither help nor hurt a query.
This is exact word- and prefix-based search: it finds the terms you type, in the documents that contain them. It doesn't guess at synonyms or related concepts.
Search on mobile
The mobile app has a global search overlay reachable from any tab. A few mobile specifics:
- It searches documents, invoices, and partners in parallel in one query.
- Document hits that matched inside the scanned content show a "Matched in content" label under the result, so you know why a result appeared even when the filename looks unrelated. This label is currently mobile-only.
- Amount search accepts both decimal styles — "29.90" and "29,90" both work.
- Your last 10 searches are remembered on the device for quick re-use.
Reliability and privacy notes
- If the search key is ever unavailable, search quietly falls back to filename-only rather than showing an error — you may notice fewer results, never a broken page.
- When you delete your account, your company's search key and all stored fingerprints are purged along with your documents. Without the key, the index was already unreadable; after erasure it's gone entirely.
Troubleshooting
A document I can see isn't found by content search. Content search covers what the scan extracted. If a document was uploaded but its scan didn't complete (or it's a document type with little extractable text), only its filename is searchable. Open the document and check its processing status.
Two characters return nothing. Expected — search needs at least 3 characters.
I searched a word I can see in the PDF but got no hit. The index is built from the scan's extraction, not from a separate copy of the raw PDF text. If the scan misread that word, correct the extracted data — the fix makes it findable.
Related
- How our encryption works, in plain English — the bigger encryption picture
- Scan with the mobile app — where the extracted data comes from
- Duplicate uploads and statement repair — finding and preventing double documents
- Where is my data stored? — storage location and residency
- Download all receipts as ZIP — getting documents out in bulk