Sift

Service Overview

Sift is a US fraud-prevention provider positioning itself as a fraud platform for digital businesses, framed around a more abstract idea: digital trust. It does not acquire payments, extend credit or absorb chargeback losses. It sits above a merchant's account system and transaction systems and outputs a judgement on whether this account and this action should be trusted.

The platform splits into 3 parts: Payment Protection makes real-time fraud decisions in the payment step with the goal of approving more genuine transactions; Account Defense blocks account takeover in real time; the AI agents line, part of the Fraud Attack Defense Suite, addresses large-scale automated attacks. Alongside those sit Sift Score API and Premium services, which round out its main delivery surface.

It covers 6 industries: B2B and B2C SaaS, e-commerce, iGaming, travel and transportation, food delivery, and finance and fintech. That list explains the positioning - all of them have account systems, continuous user behaviour, and losses that are individually small but frequent. Plain guest-checkout retail is not the centre of gravity.

The official proof points come in several forms: 700+ brands in use, a Sifters community of 1,500+ fraud fighters, customer testimonials citing a 99% new-user pass rate at Patreon and 6x ROI at Paula's Choice, and an annual Digital Trust Index report. These are best read as vendor statements, and individual testimonials cannot be extrapolated to another business.

Key Strengths

• Account-centric judgement fits platform businesses: transaction-level scoring is discrete - one order at a time, mutually disconnected - while account-level scoring is continuous, accumulating device fingerprints at signup, location jumps at login and address changes at checkout into one trail. For subscriptions and platforms, losses usually come from that trail breaking rather than from one anomalous order.
• Fake accounts are caught at registration: bulk registration is one of the most overlooked loss sources - it produces no chargebacks but consumes marketing budget, pollutes user data and inflates acquisition cost. Account Defense brings that layer in, while most payment risk tools never look at signup.
• You can take scores without the whole platform: Sift Score API lets teams with an in-house decision engine consume scores without migrating to its console. For teams with limited engineering bandwidth that still want to avoid single-vendor lock-in, this lowers migration cost in a practical way.
• AI agents handle automated attacks: credential stuffing, automated ordering and bulk arbitrage are high-volume and fast-moving, outpacing hand-written rules. Agent-driven detection and response compresses reaction time from "a human notices and edits rules" to near real time.
• Industry baselines are reusable: 6 industries mean common attack patterns and normal-behaviour distributions already have reference points, so thresholds need not be guessed from zero. For a newly formed risk team, the time saved is real.
• A broader industry view exists: reports like Digital Trust Index carry vendor bias, but they provide cross-industry attack trends that are more useful for annual planning than looking only at your own data.

Product Ecosystem

Payment Protection

Real-time fraud decisions in the payment step, returning approve or block with the goal of lifting genuine transaction pass rates rather than simply tightening thresholds. Fits merchants that treat false declines as a real loss line.

Account Defense

Covers registration, login, password reset and contact-detail changes, blocking account takeover and bulk registration in real time. This is the clearest difference from payment-only risk tools.

AI Agents (Fraud Attack Defense Suite)

A series of agents that identify and respond to automated and emerging attack patterns, upgrading detection from manual rule iteration to automated execution.

Sift Score API

Exposes risk scores by API for in-house decision engines. Suits teams that already run a risk platform and only want to add one external signal.

Premium Services

Configuration tuning and strategy review support, with scope defined by contract. For teams without dedicated fraud staff, this largely determines whether the platform actually gets running.

Limitations

• Both error directions must be priced in: approving a fraudulent account can later produce multiple chargebacks, card-network fines and abused promo budget; blocking a good user costs this conversion plus the whole lifetime value behind it. Both sides sit with the merchant because only the judgement is delivered.
• The account lens does nothing for guest checkout: if most orders are placed without login, Account Defense and fake-account blocking are barely usable and you are buying only Payment Protection, at a clearly worse value ratio. This mismatch is the most commonly overlooked selection error.
• Data sharing and cross-border compliance are not cheap: account-level scoring requires continuous registration, login, device and transaction data. Cross-border operations trigger GDPR and personal data export assessments, with separate paths where export applies, and compliance review often outlasts integration.
• Cold start depends on on-site volume: scores only discriminate with enough behavioural history. New businesses, low-traffic sites and categories with very narrow behaviour paths start near generic rules, and the model premium buys little.
• Duplicate spend against an existing risk stack: teams already running acquirer controls, in-house rule engines or identity verification must re-sequence decisions. Without that, the same user gets blocked twice and both experience and conversion drop.
• No coverage of regulated obligations: AML transaction monitoring, onboarding review and sanctions screening are out of scope. Finance and fintech customers with licensing duties still need a separate compliance system.

Use Cases

• SaaS and subscription businesses (★★★★★): long account lifecycles and heavy losses at signup and trial make account-level scoring and fake-account blocking a direct fit.
• Marketplaces and communities (★★★★☆): both sides hold accounts, so fake orders, bulk registration and account takeover coexist and one score covers several attack surfaces.
• iGaming and online entertainment (★★★★☆): accounts carry high value and abuse incentives are strong, making real-time ATO blocking essential - confirm local regulatory requirements first.
• Food delivery platforms (★★★☆☆): account protection is useful, but subsidy abuse and courier-side fraud work differently from payment fraud and need extra strategy configuration.
• Mainland China operations (★★☆☆☆): usable when a mainland entity runs cross-border business, but sending account and device data abroad must follow the Personal Information Protection Law and the applicable export rules, so a compliance assessment comes first. Purely domestic business does not need overseas services of this kind and has domestic equivalents with no export step.
• Guest-checkout retail sites (★☆☆☆☆): least suitable. Without a login state there is no account trail, so more than half the capability is unused and you pay for features you cannot deploy.
• Institutions requiring AML and licensing compliance (☆☆☆☆☆): least suitable. No AML transaction monitoring or watchlist screening in the product, so regulatory requirements cannot be met.

Pricing

Main cost drivers are event volume (registration, login, payment and password-reset calls), account scale, how many modules are enabled, whether professional services are included and the integration approach. Sift does not publish standard pricing; headings and final rates follow the current official quote and contract. Pass-rate and ROI figures in official testimonials are individual cases, not a pricing basis.

FAQ

• How is Sift different from conventional payment risk tooling?
  The object of judgement differs. Conventional tools read single-transaction features; Sift reads the account trail - signup device, login location, ordering behaviour and credential changes connected into one line. In practice the account lens blocks fake accounts and takeover at the front door, while the transaction lens can only judge once an order arrives.

• We mostly sell to guests. Is it still worth it?
  Not for full deployment. Without a login state there is no account trail, Account Defense and fake-account blocking sit idle, and the remaining Payment Protection is not meaningfully different from a dedicated payment risk tool. Unless you are moving toward membership or subscriptions, a lighter option comes first.

• Are the 99% pass rate and 6x ROI on the site reproducible?
  Not by extrapolation. They are results for specific customers in specific categories with specific traffic structures, heavily shaped by audience quality, promo cadence and the starting risk baseline. The right approach is back-testing on your own historical orders or a small canary rollout, watching false declines and fraud rate move together before going full.

• What data has to be provided?
  Registration, login, device fingerprint and payment or order events. More complete data produces sharper scores, but it has to be balanced against compliance: define the minimum necessary field set, sign a data processing agreement and assess whether the transfer constitutes personal data export. Cross-border assessment cannot be skipped.

• Will it absorb our chargeback losses?
  No. It delivers decisions and scores, and chargeback losses stay with the merchant. If the requirement is transferring chargeback risk, look at providers offering a guarantee or recovery-based pricing. The two can be layered, but billing and responsibility boundaries must be accounted for separately.

Source: original 16IDC content. Product capabilities, supported countries and fee definitions follow the official documentation and your contract.