Getting started

The free coverage assessment

Before you integrate anything, the assessment shows you exactly what your current gambling block is missing, measured on your own transactions. By default your analyst runs it inside your environment with one command; nothing leaves your bank, and any file with personal data is refused before it is read.

What it tells you

The disguised gambling your block cleared. The specific charges that got through because they billed under ordinary category codes, with the amount involved.
Your exposure to the disguise codes. The share of your traffic running through the merchant codes disguised gambling hides behind, traffic your block cannot tell apart from gambling.
Behavioral signals. Accounts showing gambling-shaped spending, surfaced for review.
Suspected merchants we do not yet track. Merchants in your data that resemble known gambling operators, as research leads for us to investigate, never as automatic verdicts.

Every finding is evidence, not an action. The assessment changes nothing in your systems. It is a measurement you use to decide whether the gap is worth closing.

Two ways to run it

The assessment kit (default). One signed folder, run by your analyst on your own machines. python3 assess.py --sample proves the whole pipeline on bundled synthetic data in about a minute; python3 assess.py your_extract.csv runs the real thing. The feed signature is verified before every load and the run fails closed. If a file is refused, it is refused unprocessed and left untouched; it is your machine and your data. Nothing is installed, no network access is needed or used, and you choose what, if anything, to share back. The aggregate summary is enough for the follow-up conversation.
Anonymized extract. If you prefer, you send an anonymized transaction export. It is processed locally on our hardware, never uploaded to any cloud or third party, and deleted within 30 days under a data-use agreement; refused files are deleted at intake. An NDA comes first either way.

What to export

An anonymized export with, at minimum, the statement descriptor, plus any of amount, date, outcome, category code, and a random account token. No personal data of any kind. The intake scans the raw file for personal data before it parses anything and refuses the file if it finds emails, government IDs, phone numbers, card numbers, or personal or account name columns. In the kit, a refused file is left untouched on your machine; on the extract path it is deleted at our intake.

example columns

descriptor, amount, date, outcome, mcc, account_token

Getting a useful read from a sample. Disguised gambling is a small, concentrated fraction of card traffic. A small or random export can come back with few direct hits; that is a clean spot-check, not proof the exposure is absent. For a fuller picture, send a targeted extract of the disguise category codes over a quarter. We will tell you exactly which.

← How it works

Shadow mode →