YARDA Connect
Your transactions scored, with no integration project
No technical team available, no provider to bring in? Send a file of transactions from the console, or let YARDA read a view of your database. Every transaction is scored as if it came through the API, and the alerts arrive in the same queue.
By file
One file, four steps, no development
An export from your system is enough: the one you already produce for your checks. Nothing is imported until someone on your team has confirmed how the file is read.
Drop the file
CSV, separated by commas, semicolons, tabs or vertical bars, or JSON. The encoding is recognised, whether UTF-8 or Windows-1252. An Excel workbook is saved as CSV first.
Choose the purpose
“Analyse and alert” to have recent transactions reviewed, or “Add to the history” to give your models the past they learn from.
Check how the columns match
YARDA suggests a column for each expected field; you confirm or correct it. The date format is detected, and any ambiguity between day and month is pointed out. A preview shows the first rows as they will be read.
Start, then follow
The import is followed live: rows read, analysed, added to the history, already known or refused. The analysed transactions open in the console, and their decisions download as CSV.
Refused rows, explained
Missing reference, unreadable amount or date: every rejected row is listed with its reason and downloads as CSV, to be corrected in your system.
No duplicates
A transaction already analysed, however it arrived, is not analysed a second time: it shows as already known.
Pseudonymised on reading
The file’s identifiers of people and devices are replaced with pseudonyms before anything is stored, as with the API.
From the database
YARDA reads a view of your database, at your pace
For a continuous feed with no development, your team creates a view of the transactions and a read-only account. YARDA reads the new rows at a regular interval, one minute by default, and scores them.
PostgreSQL, read-only
The connector works today with a PostgreSQL database. It reads the view you name, with a read-only account, in a session itself opened read-only, and in limited batches.
A protected connection
The connection is encrypted by default, with certificate checking if you wish. The account’s password is encrypted by YARDA and never shown again.
The same settings as a file
The same column matching, the same choice between analysis and history. An ordering column tells YARDA where it stopped, and you choose to start from new rows or from the whole view.
Under your control
The connector is set up with the YARDA team. Your administrator sees its state and its last read, and can pause or resume it, giving a reason.
Then
The same product as through the API
Whether your transactions arrive by file, from the database or through the API, they go through the same analysis and feed the same console.
The same alerts
Models, rules and typologies apply in the same way. Alerts join your analysts’ queue, with their reasons.
Silence gets noticed
The Data health page shows the state of each feed. If nothing arrives beyond the delay set in your policy, an e-mail warns you.
History that serves
The months of transactions you add to the history are used to train your models and to replay your rules before you switch them on.
Frequently asked questions
- Which columns must my file contain?
- At least a transaction reference, a date, an amount, the sender and the receiver. The console shows the required fields for the chosen purpose and accepts your own column names: you match them once.
- Can I start with a file and move to the API later?
- Yes. The imported history and your settings stay in place; only the way transactions are sent changes.
- Who sets up the database connector?
- Your team creates the view and the read-only account; the YARDA team enters the connection and checks it with you. Your administrator can then suspend it at any time.
Start with a file
The free diagnostic starts from an export of three to six months of your transactions, with no development at all.