Who Sift is
Sift is an AI-powered fraud prevention company that has been fighting fraud since 2011. It scores the risk of users and their actions from 0 to 100, with a separate score for each kind of fraud you are preventing, and its products cover payment protection and account defence.
Its models learn from the decisions you send back and from data across Sift's global network of sites and apps. Online businesses use it to protect payments and accounts, which is why it is one of the fraud engines our Fraud & risk module connects to.
What it does in your platform
Fraud & risk
Scores account and payment events with Sift's models and its global network, with a separate score for each kind of fraud. The score and the result of any Sift Workflow come back with the event and join your own rules in one decision.
How the connection works
Sift's JavaScript snippet and mobile SDKs collect device data in your apps. The module sends each event to Sift's Events API and asks for the score in the same call, and decisions made in Sift come back by signed webhook.
Device data is collected
Sift's SDK in the app, or its JavaScript snippet on the web, records the device and how the customer uses it.
An event is sent
The module sends the login or the payment to Sift's Events API as one of Sift's standard events, such as a login or a transaction.
The score comes back
In the same call, Sift returns its score from 0 to 100 and the result of any Workflow you set up to run on that event.
The module decides
It merges Sift's score with your own rules: allow, ask for a second factor, hold for review or block.
Decisions flow both ways
An analyst's decision in the backoffice can go to Sift's Decisions API to teach its models, and decisions made in Sift come back by webhook, signed with your key.
Next to other providers
The module can call Sift in parallel with Sardine, SEON or Feedzai, or next to Fingerprint's device intelligence, and merge their signals into one decision. If Sift is slow, the decision falls back to your own rules within a set timeout.
Your rules, thresholds and case queue live in the module, and Sift's Workflows can sit next to them. Replacing Sift later does not change your apps, and past decisions stay in your records.
When Sift fits best
A strong fit when
- You want models trained on data from many online businesses, not only yours.
- You want a separate score for each kind of fraud, from payments to account takeover.
- Your team wants to automate decisions with Workflows, without code changes.
Also worth a look
- Sardine or SEON, when you also want AML screening or monitoring from the same provider.
- Feedzai, for banks, issuers and acquirers that want fraud and AML on one platform.
How we get you live
The contract
We help you get your Sift account and contract in place, set up for your products and the kinds of fraud you face.
Events and Workflows
We map your platform's events to Sift's, and set up the Workflows and thresholds that turn scores into decisions.
The keys
The API key and the webhook signature key go into your platform's secrets and nowhere else, so every webhook is proven to come from Sift.
A full test run
The whole flow runs in Sift's sandbox, which keeps test events away from your production scores, before your first real customer.
