Snap Apps

Getting started

Snap Apps turns a photo of a receipt or tax invoice into a checked, categorised record — and, once you confirm it, a balanced ledger entry. It is built for Australian sole traders, tradies, and the accountants and bookkeepers who look after them.

What it does

Three things, in order: it captures the original photo and keeps it untouched as your legal record; it extracts the supplier, amounts, GST, and line items from it; and once you confirm what it found, it posts a proper double-entry transaction to your ledger. Every posted transaction keeps a link back to the original image, for as long as you're required to keep it.

The shortest path to your first posted transaction

  1. 1. Photograph a receipt. See Capturing receipts well — a flat, well-lit, uncropped shot gets the best result.
  2. 2. Wait for extraction. A model reads the fields, and a set of deterministic checks — ABN checksum, GST arithmetic, date sanity — run over what it found before you ever see it.
  3. 3. Review or accept. If every field was confident and every check passed, it is auto-accepted and a draft transaction is already proposed. Otherwise it is needs review, with the specific thing that failed shown against the field. See Confidence and why a field was flagged.
  4. 4. Confirm. Confirming is the one action that posts a balanced transaction to your ledger. Editing changes data; confirming moves the books. See The ledger and what "posted" means.

Personal and business

A workspace is either personal or a business. Personal spending never mentions GST, an ABN, or a tax invoice — none of that applies to it. Business workspaces carry the full GST and BAS machinery described in the rest of these docs. If you do both, you track them as separate workspaces rather than mixing them in one place, which is what keeps your BAS clean.

What actually happens to the photo

The original image is kept exactly as captured — it is your legal evidence, and the ATO only accepts an electronic copy that is a true and clear reproduction of the original, so it is never recompressed or edited. A separate, normalised copy is derived for the extraction model to read; that derived copy has any location data stripped from it before it goes anywhere. See Data handling & security for the detail.