Berbix
Summer 2018AcquiredConvert more verified users. Instantly & accurately validate user IDs.
Pressure-test this opportunity
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Convert more verified users. Instantly & accurately validate user IDs.
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Berbix turned the hardest step in online identity proofing into developer infrastructure. Founded in 2018 by former Airbnb Trust and Safety operators Eric Levine and Steve Kirkham, it let a company capture an ID and selfie, inspect the document, detect fraud, and return a result through an API.
Berbix produced a strategic exit. Socure bought it in June 2023 for approximately $70 million in cash and stock. The deal exposed the limit of Berbix's position: document verification could be excellent and still remain one step in a buyer's larger identity, fraud, consent, and decision workflow. Socure joined the component to an identity graph and orchestration platform, then launched Predictive DocV 3.0.[1]
Image 1 / 1
Eric Levine and Steve Kirkham came to Berbix with unusually direct preparation. At Airbnb, both worked on Trust and Safety and helped build Verified ID, an early attempt to give hosts and guests confidence about who was on the other side of a transaction. Initialized Capital's Garry Tan described them as the team behind that feature when his firm led Berbix's $2.5 million seed round.[2]
The Airbnb experience supplied the product thesis. In his post-acquisition account, Levine wrote: "While we leveraged many of the best identity tools available at the time, we found that even the best left us wanting something with more reliability and better user experience."[3] Berbix would make the document check immediate, automated, and easy for a developer to add.
The founders incorporated that idea in 2018 and joined Y Combinator's Summer 2018 batch.[4] The first product owned the complete capture flow. It checked image quality while the user was still present, identified the document type, decoded machine-readable fields, compared the document with known jurisdiction formats, and flagged convincing fakes. A customer could add standard ID checks with a few lines of code, then add a selfie and liveness step when the risk justified more friction.
Cannabis became an early wedge because online and delivery businesses needed to verify age without putting a clerk in front of every customer. The product also fit car rental, financial services, marketplaces, and nonprofits distributing funds. When the pandemic pushed more formerly in-person checks online, demand rose. Steve Kirkham told TechCrunch: "The inability to conduct traditional identity checks in person has forced organizations to move online for innumerable use cases."[5]
That demand supported a $9 million Series A led by Mayfield in August 2020. Initialized, Y Combinator, and Fika Ventures also participated. Berbix planned to hire across product and sales, localize the service, and keep improving fraud detection as forged documents evolved.[5]
Berbix sold an identity-verification session. A customer created a transaction through the API and sent the end user into a hosted or embedded capture flow. The user chose an ID type, photographed the document, and could complete a selfie or liveness check. The service returned extracted fields and a verification result. The published mobile demo shows that sequence without the customer having to design its own camera or review interface.[9]
The first product advantage was capture quality. Berbix checked blur, glare, obstruction, and document framing before submission so the user could retry immediately. It then classified the document, read its barcode or machine-readable zone, and compared fields and visual indicators. Initialized claimed in 2019 that the system could find sophisticated fake IDs without routing every transaction to a person; that was an investor claim, not an independent benchmark.[2]
The forensics engine became the strategic asset. Eric Levine later explained that the patented method compared human-visible and machine-readable elements, tested whether portraits and barcodes looked legitimately printed, and searched for inconsistencies within the document's data. He said those extraction and classification systems were deeply integrated into Socure DocV after the sale.[7]
The customer experience involved policy as well as recognition. A cannabis commerce flow documented by Blaze sent users a link by text or email, asked for the front and back of an ID, captured a selfie, and offered a manual-review path when a user withheld automated facial-recognition consent.[10] Berbix wrapped its computer-vision model in a finished verification workflow.
Its 2020 list price made that infrastructure accessible: $99 per month included 100 standard checks, then checks cost $0.99 each, with selfie and liveness checks available separately.[5] Stripe-like developer ergonomics reduced integration work. The hard part remained the changing adversary: every new document type, capture device, and forgery method created another condition the system had to understand.
Berbix fit companies that had to trust a remote stranger before handing over money, access, goods, or regulated services. Cannabis sellers needed age checks. Car-rental operators needed to confirm a driver's identity. Marketplaces and gig platforms needed to keep banned or fraudulent users from returning under a new account. Fintechs and nonprofits needed to validate applicants without requiring an in-person visit.
The buying center varied by vertical, but the job was consistent: convert a legitimate user quickly, catch a fraudulent one, and retain enough evidence to explain the decision. Berbix's developer-first product was strongest when document verification was the only missing step. A buyer with complex KYC, sanctions, device, identity-graph, case-management, and consent requirements needed more than the Berbix response.
No credible public figure isolates Berbix's serviceable market during 2018–2023. Broad identity-verification forecasts mix document capture with databases, biometrics, compliance, authentication, and fraud decisioning. Using them as Berbix revenue potential would overstate the opportunity.
The operational demand was clear. TechCrunch reported that Berbix's daily verification volume in August 2020 equaled its monthly volume a year earlier, according to Kirkham.[5] Yet the public record does not disclose transaction count, customers, revenue, retention, or gross margin. Growth direction can be established; scale cannot.
Berbix competed on two related axes: document-forensics quality and completeness of the identity stack. Human review and older verification vendors could inspect documents but introduced delay and variable accuracy. Developer-first vendors such as Persona and Onfido packaged capture, workflows, and checks. Larger platforms such as Socure could combine document evidence with phone, email, address, device, watchlist, and identity-graph signals.
Berbix was well placed on the first axis. Its automated capture and document-specific forensics reduced user delay, and its patent suggests real technical depth. The second axis favored consolidation. A document decision can reject a fake card, but it cannot by itself tell a lender whether the submitted person, device, phone, bank account, and address form a coherent identity.
The current Socure product makes the distinction explicit. Predictive DocV webhook payloads now include reason codes, parsed document data, device intelligence, and workflow decisions.[11] Its mandatory consent step covers photos, biometrics, retention, and automated facial recognition.[12] Customers can export consent identifiers, timestamps, versions, and language for audit.[13] These surrounding controls explain why the component had more value inside a platform than alone.
Berbix charged for access plus usage. In 2020 the entry tier cost $99 per month for 100 checks; additional standard checks cost $0.99, and selfie or liveness checks added fees.[5] That model aligned revenue with verification volume and let a small developer start without an enterprise contract.
Public funding includes the $2.5 million seed and $9 million Series A, though the public record does not establish whether every reported figure is incremental. Revenue, gross margin, cloud cost per check, manual-review share, customer concentration, and burn were not disclosed. A burn estimate from current YC headcount would be misleading because the listing may reflect a later snapshot and omits compensation, compute, insurance, and legal costs.
The approximately $70 million purchase price was several times the named financing, but that ratio does not reveal investor or founder returns. Ownership, preferences, debt, transaction expenses, and the cash-stock split by holder remain private. The defensible conclusion is that Berbix created strategic value; the payout distribution is unknown.
The strongest public growth signal is Kirkham's 2020 comparison: Berbix processed as many verifications in one day as it had in a month one year earlier.[5] TechCrunch also named the Family Independence Initiative as a customer use case during COVID, when remote identity checks helped the nonprofit screen applications.
The customer record spans cannabis, marketplaces, car rental, and financial services, but the available examples do not establish repeatable economics. No audited transaction volume, customer count, annual recurring revenue, retention, or false-accept rate was located. Socure's willingness to pay approximately $70 million and claim an 18-month roadmap acceleration is the clearest external signal that the technology and team mattered.[1]
Berbix solved the document. Enterprise customers bought an identity decision. That difference became more important as buyers combined KYC databases, fraud models, device signals, watchlists, workflow rules, consent, manual review, and audit evidence.
Levine stated the constraint plainly after the acquisition: "But despite the strength of our identity document verification solution, it was only one component of a complete identity stack."[3] The team could have expanded horizontally, but each added signal demanded new data, models, compliance work, and enterprise integration. Socure already owned those layers.
This is the structural mechanism behind the exit. Better document forensics increased Berbix's strategic value while reducing the odds that document verification alone would remain the system of record. The more important the component became, the more valuable it was to a platform that controlled the final decision.
Biometric verification brings consent, retention, disclosure, and jurisdiction-specific obligations. The 2022 Mahmood order illustrates that exposure. The plaintiff alleged Berbix collected and analyzed facial geometry through SilverCar's rental flow and failed to obtain required consent for certain uses and disclosures. The judge denied Berbix's motion to dismiss, finding the claims plausible at that stage; the order did not determine that Berbix violated BIPA.[6]
The team had product responses, including consent and manual-review paths. The problem was broader than a checkbox. A regulated buyer must know which notice appeared, which version applied, whether the user agreed, what data moved to which processor, and how a later decision can be reconstructed. Current Socure documentation treats those records as first-class outputs.[12][13]
Regulation therefore reinforced the full-stack advantage. A larger platform could spread legal, policy, reporting, and integration costs across more products and customers. Berbix could keep deepening document accuracy, but every enterprise deployment still carried workflow obligations outside the model.
Document verification is a permanent contest with forgery. Berbix invested in localization, data extraction, fake-ID detection, and liveness. Its patented virtual-authentication approach examined indicators that visually convincing documents could still get wrong.[7]
The work did not end at acquisition. Socure's 2026 updates include new injection and deepfake detection, cryptographic validation of California ID barcodes, cross-transaction face matching, and revised consent language.[8] A small vendor must fund that cycle while supporting SDKs, devices, document formats, and policy changes. A platform can reuse its device intelligence, graph, workflow, and sales organization around the same document engine.
Calling Berbix a failed startup would distort the record. The company raised institutional capital, showed rapid volume growth, built patentable technology, and sold for a disclosed strategic price. Socure said the integration advanced its roadmap by 18 months, and Predictive DocV remains active three years later.[1][8]
The narrower judgment is more useful. Berbix proved that developer-friendly document verification could be a valuable company, then exited before trying to build every adjacent identity layer. Its technology survived because the acquirer had the distribution and product breadth the component lacked. Whether the founders and investors maximized their financial upside cannot be determined from public evidence.