Sift

Sift is a digital trust and safety platform — software that helps online businesses detect and prevent fraud and abuse. Founded in 2011 and based in San Francisco, the company was originally known as Sift Science before shortening its name to Sift. Its core job is to look at the activity flowing through a business — account signups, logins, payments, and other actions — and judge, in real time, how likely each one is to be fraudulent or abusive, so the business can block the bad and let the good through. It does this largely with machine learning, models trained on large volumes of activity to spot the patterns that distinguish legitimate users from fraudsters, chargeback abusers, and fake-account creators. In practice, a business sends Sift the signals around a transaction or event, and Sift returns a risk score and a recommendation, which the business uses to approve, review, or reject the action.

Sift matters as an example of the digital trust and safety category — the set of tools businesses use to fight online fraud without also blocking their real customers. That balance is the whole challenge. Fraud costs money directly, through stolen goods, chargebacks, and abuse, but over-aggressive fraud controls cost money too, by rejecting legitimate customers and adding friction that drives them away. A platform like Sift tries to thread that needle, using data and models to catch more fraud while wrongly rejecting fewer good users. It is used across e-commerce, marketplaces, fintech, and other online businesses where strangers transact and the risk of fraud is constant. For a marketer or operator, Sift is a concrete instance of the infrastructure that sits quietly behind a smooth checkout or signup, deciding which visitors are trustworthy — a function most customers never see but that shapes their experience.

Sift is one of several well-known fraud-prevention and trust-and-safety platforms, and it is worth distinguishing it from its peers rather than treating them as interchangeable. Companies such as Forter and Signifyd operate in the same broad space of online fraud prevention, and to a casual eye they look similar — all use data and machine learning to score transactions and stop fraud. But they differ in emphasis and model. Some rivals, for instance, center their offer on a guarantee, taking on the financial liability for fraudulent chargebacks they approve, so the merchant is reimbursed if a transaction they cleared turns out to be fraud. Sift's identity centers more on being a broad trust-and-safety platform covering many types of fraud and abuse across the user lifecycle, from account creation to payment, rather than being defined solely by a chargeback guarantee.

The distinction matters because the right tool depends on the problem. A business whose main pain is payment fraud and chargebacks might weigh a guarantee-based provider that shifts liability; a business fighting a wider range of abuse — fake accounts, promotion abuse, spam, account takeover, as well as payment fraud — might favor a broader platform positioned across the whole user journey, which is how Sift presents itself. The general point is that fraud prevention is not one product but a category with real differences in scope, model, and commercial structure, so lumping Sift, Forter, and Signifyd together as the same thing obscures the choice that actually matters. When comparing them, the useful questions are what kinds of fraud each targets, whether it offers a liability guarantee, and how it balances catching fraud against wrongly rejecting good customers — not which name is most familiar.

Sift fits into a business as one layer of a broader trust-and-safety and risk stack, not as a complete answer to fraud on its own. It provides risk scores and recommendations, but the business still has to decide the thresholds — how strict to be — and to tune the balance between blocking fraud and admitting good customers for its own tolerance and margins. Set the controls too loose and fraud slips through; set them too tight and legitimate customers get rejected, which is its own costly failure. Like any machine-learning system, its judgments are only as good as the data and patterns behind them, and fraudsters adapt, so the models must keep learning. Used well, Sift is a powerful tool for scoring risk at scale; used carelessly, it can either wave through fraud or quietly turn away real customers, both of which cost the business.

Reading Sift accurately also means separating durable facts from shifting details and avoiding promotional gloss. The stable facts are that Sift is a San-Francisco-based digital trust and safety platform, founded in 2011, formerly called Sift Science, that helps businesses detect and prevent online fraud with machine learning. Its exact customer list, pricing, and product lineup change over time and should be checked against the company's own materials rather than asserted. The failures to avoid are treating Sift as a magic fraud eliminator rather than a scoring tool that needs tuning, confusing it with peers that work differently, and ignoring the false-positive cost of rejecting good customers in the rush to stop fraud. The discipline is to see Sift as one configurable layer in fraud prevention, understand how it differs from alternatives, and weigh both kinds of error — fraud let through and good customers turned away.

Sift takes its name from the verb sift, to sort the good from the bad by passing through a sieve — what its models do to online transactions.

Etymology: source.

Search interest for this term over the last five years:

View interest-over-time on Google Trends →

Areas of marketing where sift is a core concern: