Sift vs Signifyd (2026)
The decision is whether the vendor pays for its wrong approvals or your own team owns every call. Signifyd puts money behind the orders it approves. Its contract credits the merchant for an eligible chargeback on any order it recommended accepting. Complete Chargeback Protection stretches that cover to chargebacks that are not fraud. Since its summer 2026 release, Signifyd can also send a borderline order to a step up, such as 3D Secure, a human reviewer or a slower Deep Thinking model, rather than decline it. On Shopify, merchants can now take its decisions with the cover or without it. Sift sells no cover. Its scores feed workflows your team configures, and its contract leaves both the decision and the loss with you. In return your team gets the controls, from review queues to outside reviewers at TTEC or Riverty. Sift also reaches past the store, guarding signups, logins and money movement in consumer and fintech apps. Pick Signifyd if you sell online and want chargebacks paid for. Pick Sift if fraud spans accounts and payments across an app and your team wants to run the defenses itself.
- Your growth team cares about good signups as much as bad ones. Patreon reports accepting 99 percent of new users with Sift, and Remitly halved its manual review rate.
- Account takeover costs you more than card fraud. Rently reports 65 percent less account takeover on Sift, whose Account Defense watches logins and sessions as well as payments.
- Your review queue needs extra hands. Sift lists TTEC and Riverty as manual review providers that supply labels when a customer asks for them.
- Your security team wants the supply chain in writing. Sift's public addendum names its subprocessors, from Snowflake and Databricks for analysis to TeleSign and Twilio SendGrid for verification codes.
- You have to show how sensitive data stays out of decisions. Sift says it keeps protected class data out of its global model and that no single data point decides a score.
- You want every chargeback off your books, not only fraud. Signifyd's Complete Chargeback Protection reimburses eligible chargebacks that are not fraud as well, on the orders it approved.
- You would rather rescue a borderline order than decline it. Signifyd's step ups send it to verification, 3D Secure, a human reviewer or a slower Deep Thinking model that reads behavior as a sequence.
- AI shopping agents are placing orders on your site. Signifyd's agentic buying update sorts those orders into three kinds, from agent assisted to agent native, so they are not flagged by mistake.
- Returns are where you lose money. Signifyd's Returns Assurance scores a return the moment a shopper files it through AfterShip, and Shopify merchants can refund good returns once the carrier scans the package.
- You want a result from a brand you know. Philips raised its conversion rate from 40 percent to more than 75 percent on Signifyd and cut its chargeback rate from 5 percent to under 1 percent.
This comparison is published by AI FinTech Index, an independent research platform that publishes independent ratings of AI vendors for financial services. Sift and Signifyd are each graded against the same capability taxonomy, from each vendor's own public materials and the regulatory record, under the AI FinTech Index verification standard. No vendor pays for placement, and no vendor has reviewed this page. How this evidence is graded
Plain facts
| Sift | Signifyd | |
|---|---|---|
| Primary category | Fraud Detection & Transaction Risk | Fraud Detection & Transaction Risk |
| Founded | 2011 | 2011 |
| Headquarters | San Francisco, California, United States | San Jose, California, United States |
| Website | sift.com | www.signifyd.com |
Side by Side
Select any grade to read the note behind it.
| Axis | S Sift |
S Signifyd |
|---|---|---|
| AI Centrality | ||
| Autonomy and Oversight Model | ||
| Model Risk Management and Transparency | ||
| Operational and Outcome Evidence | ||
| AI Safety and Data Stewardship | ||
| GLBA and Data Privacy Posture | ||
| Security Certifications and Trust Center | ||
| Regulatory Status and Licensure | ||
| AI Governance and Bias Disclosure | ||
| AI Liability and Recourse | ||
| Model Supply Chain Disclosure | ||
| Core Systems and Integration Depth | ||
| Deployment Model and Data Residency | ||
| Commercial Transparency | ||
| Institution and Segment Coverage |
Product updates
The latest logged changes from each vendor, newest first, each linked to its source.
Sift
-
Product / capability
Sift released Review Agent, which reads a fraud case and gives the analyst color coded risk and trust signals, a short plain language explanation and a suggested action of Block, Accept or Watch.
Source
The short version of each
Sift
Sift is a fraud platform that digital businesses run themselves. It scores signups, logins, payments and disputes against a network Sift puts at more than a trillion events a year from 34,000 sites and apps, drawing on more than 16,000 signals. Payment Protection and Account Defense lead the product line, and Dispute Management prepares chargeback responses. Fraud teams set the rules that act on each score, and they can bring in outside reviewers from TTEC or Riverty. Teams connect through a public API, with Okta for console sign on and Tray.ai to link internal apps. Named results include Remitly's halved manual review rate and Rently's 65 percent drop in account takeover. Sift's trust center lists CCPA, DORA, the EU AI Act and GDPR without saying what it does under each. The main caveat is the contract. Sift puts no money behind any decision, makes no promise of accurate results and leaves the customer responsible for what it decides.
Source: AI FinTech Index, 2026
Signifyd
Signifyd decides whether to accept online orders and pays merchants back when an approved order turns into an eligible chargeback. Retailers such as Philips, Ferguson Home and Hot Topic buy it, and payment providers get a platform of their own. Its products now follow the shopper from first visit to return, across Session Intelligence, Account Integrity, Purchase Optimization and Returns Assurance. The summer 2026 Spark release added step ups for borderline orders, recognition of orders placed by AI agents and promo abuse rules a merchant can test on past orders. Merchants pay a percentage of each approved order's value, and an order declined for fraud costs nothing. Connections run through public APIs or plugins for Shopify, BigCommerce, Adobe Commerce and Salesforce Commerce Cloud. Chargebacks flow in from Adyen, Braintree, PayPal and Stripe. A SOC 3 report covering July 2025 to June 2026 is free to download. The main caveat is what sits outside the credit. The terms give no other assurance of accuracy and cap Signifyd's liability at six months of fees, and Signifyd publishes no error rate or fairness result.
Source: AI FinTech Index, 2026
Common questions
How is Sift different from Signifyd?
Sift and Signifyd both score online activity in real time on data pooled across many businesses, but they sell different promises. Signifyd decides each online order and credits the merchant for an eligible chargeback on any order it approved, so the vendor carries the cost of its own misses. Sift returns a risk score and leaves the decision to the customer's own workflows and analysts, so the customer carries the cost. Signifyd's buyers are online retailers and the payment providers that serve them. Sift reaches beyond the store to consumer apps and fintechs, guarding signups, logins and payments. A retailer that wants chargebacks paid for leans toward Signifyd. A team that wants to set its own thresholds across an app leans toward Sift. Graded by AI FinTech Index against the same capability axes from each vendor's own published materials, verified November 5, 2026. No vendor pays for placement.
Which is cheaper, Sift or Signifyd?
Neither Sift nor Signifyd publishes a rate, so neither can be called cheaper before a quote. Signifyd charges a percentage of the order total on each order it approves and nothing for an order declined as fraud. The percentage depends on the products bought, the merchant's vertical, order volume and average ticket, and Signifyd promises not to win business with a low introductory rate that rises later. Sift's one public figure is on its AWS Marketplace listing, where platform access sells at $10,000 per contract unit for 12 months, with no definition of a unit. The fees also buy different things. Signifyd's fee includes chargeback cover, while Sift's buys software and leaves chargeback losses with the customer. Graded by AI FinTech Index against the same capability axes from each vendor's own published materials, verified November 5, 2026. No vendor pays for placement.
How does Signifyd's chargeback credit work, and what does Sift offer instead?
Signifyd's terms of service credit the merchant when an order Signifyd recommended accepting draws an eligible chargeback. The credit is capped at the order amount plus standard processing fees. The merchant must report the chargeback within 7 days, keep its fees current and receive a reason code Signifyd's policy covers. Signifyd's product page says fraud chargebacks are repaid within 48 hours, and Complete Chargeback Protection adds eligible chargebacks that are not fraud. Sift offers no credit. Its terms make the customer responsible for its own decisions. Sift sells Dispute Management instead, so a team can build and send its own chargeback responses. Graded by AI FinTech Index against the same capability axes from each vendor's own published materials, verified November 5, 2026. No vendor pays for placement.
How do Sift and Signifyd explain decisions and check their AI assistants?
Each Sift score arrives with the signals that drove it, and Signifyd's Explore, Investigate, Act view lays out why an order was approved or declined. Both also put a generative AI assistant in the analyst console, and both keep a person in charge of it. Signifyd's Console AI Agent turns plain language into order searches, charts and draft policies, and nothing applies until a user approves it. Sift's Review Agent suggests a decision with its reasoning, and its Risk Activity Agent summarizes a user's sessions with a note to verify before acting. The AI FinTech Index points out that neither vendor names the model that drafts these suggestions. Signifyd says only that its provider is a subprocessor barred by contract from training on merchant data. Sift names Google as the generative AI supplier for its session summaries, but not the model behind the Review Agent.
Does my data train models for other Sift or Signifyd customers, and where is it kept?
Yes, with both. Signifyd's privacy notice says shopper data from across its merchant network improves its models, and it commits never to pass one merchant's shopper data to another. It adds that data enrichment providers and authorization partners, such as banks, card networks and processors, may use shared data to train predictive models of their own. Sift's terms say customer data is commingled with other customers' data, and its global model trains across the network. Sift does not identify a customer as the source. Sift holds end user data for four years at most. Neither describes a way to opt out. Sift hosts on Google Cloud in two US regions, while Signifyd names Amazon Web Services and Google Cloud without a region. Signifyd's SOC 3 report is free to download, and Sift's ISO 27001:2022 certificate and SOC 2 Type II report sit behind an access request. Graded by AI FinTech Index against the same capability axes from each vendor's own published materials, verified November 5, 2026. No vendor pays for placement.
Related comparisons
Other published head to head assessments involving these vendors or their closest peers. The full set for this category is on the Fraud Detection & Transaction Risk page.
Neither Sift nor Signifyd publishes a false decline rate, an error rate or a validation report, and neither shows how decisions fall across groups of shoppers. Customer figures on both sides, from Remitly at Sift to Philips at Signifyd, give no period or method. Both learn from data pooled across customers, and neither describes a way to opt out. Neither states a regulatory position, though Sift's trust center lists DORA and the EU AI Act without saying what Sift does under either.
Signifyd owes written answers on four points. Which provider runs the Console AI Agent, which enrichment and authorization partners may train models on data it shares, what its unpublished subprocessor list contains, and how long are shopper profiles kept? Sift owes written answers on three. Which model runs the Review Agent, what one AWS Marketplace contract unit covers, and what does Sift do under DORA and the EU AI Act?