DeepID vs point solutions
Most banks assemble fraud and identity checks from several vendors, one per check. DeepID puts sign-up, login and the session on one SDK and one decision. Here is how the two approaches differ, and when each fits.
On this page
Side by side
| Question | DeepID | Point solutions |
|---|---|---|
| SDKs and security reviews | One SDK for sign-up, login and the session, reviewed once by your security team. | One vendor per check, each with its own SDK and its own review. |
| Where the decision is made | One decision per moment, every signal weighed together, the reasons listed. | Each tool returns its own score; your team or your FRM joins them up. |
| A coached payment | The call, the shared screen, dictated typing and the payee are read in the same session. | Each signal arrives alone, if at all, and each passes the real customer. |
| A new phone or SIM | A fresh check against what was bound at sign-up, then a cool-off. | Often a new code to the same number. |
| What reaches your FRM | The decision, its reason codes, the message shown and the policy version, under one id. | Separate alerts, in separate formats, from separate vendors. |
| Keeping what you have | Keep your KYC provider and start at login or the payment. | Add a tool for each new check. |
Why the session matters
Point solutions each answer one question well: is this document genuine, is this face live, is this the registered phone. The fraud that grows now passes each question on its own, because the real customer is answering them, on their own phone, while someone on a call tells them what to do. What gives it away is the combination, read in one session.
When a point solution is enough
If one journey needs one check, and nothing else in the session matters to the decision, a single tool can be the simpler choice. Many banks also keep the KYC provider they have and start DeepID at login or the payment.
How to decide
Run the comparison on your own traffic. A free retrospective on your own history and a shadow proof on one journey show what each approach catches, against criteria agreed in writing first.