Payment screening

Use case · Payment screening

The name in a wire is free text.

Originator and beneficiary fields arrive typed by a human somewhere upstream: truncated, romanized, reordered. The payment clears in seconds, and the screen has milliseconds.

2 sidesof every payment get screened: originator and beneficiary, both free text.SWIFT / Fedwire fields

500pairs scored per API call, so one wire's aliases fit in one request.Onoma batch limit

mslatency budget. Screening cannot be the reason a payment is slow.Payment ops reality

Where the check runs

Inline, on every payment, against lists that changed this morning.

Payment ops screens each transaction's parties against sanctions lists before release. A false hit stops a legitimate wire and a customer calls. A miss is a regulatory event. The margin between them is the quality of one name comparison.

Today

String rules at wire speed.

Free-text fields mean no canonical spelling ever arrives.
Stopped wires from false hits mean ops queues and angry customers.
Tuning down sensitivity to release the queue opens the miss window.
Every list update means re-tuning by hand.

With Onoma

One score, one threshold, at speed.

Calibrated scores let you set a release threshold once and keep it.
Single-pair calls fit an inline latency budget; batches handle repair queues.
Evidence output gives the ops analyst the why in one glance.
Consistent scoring across list updates. No re-tuning cycle.

In the flow

Score both parties before release.

Inline single-pair calls for the payment path.
Batch calls for repair queues and retro-screens.
Stable thresholds. Auto-release below, auto-hold above, review the band.
50 req/s per key, higher on request.

$ curl https://onoma.kineticpartners.org/api/v1/match \
  -H "Authorization: Bearer sk_live_..." \
  -d '{ "a": "HUSSEINI MOHAMMED A",
        "b": "Мухаммад Хусейни" }'

{ "score": 0.90,
  "verdict": "match",
  "evidence": {
    "script_pair": "latin/cyrillic",
    "normalized": { "a": "husseini mohammed a",
                    "b": "mukhammad khuseini" } } }

Inline, on every payment, against lists that changed this morning.

String rules at wire speed.

One score, one threshold, at speed.

Score both parties before release.

Screen every payment without slowing one down.