The Team That Made Airbnb Safe For Guests is Bringing Full Automation to ID Verification
Initialized led a $2.5M seed in Berbix to bring real high quality software to an error-prone bug-ridden process that is the keystone of online markets
Trust and reputation is a foundation of modern commerce. Plainly, you just really need to be able to make sure who you’re doing business is who they say they are. Greater transparency into another counterparty’s record and history can make it easier for strangers to do business together.
Airbnb has shown that when you create a software platform that puts those two sides of a marketplace together, you can create a multi-billion dollar opportunity by making something better, faster, and cheaper.
The team that launched Airbnb’s Verified ID feature is making it possible for everyone to do the same.
Steve Kirkham and Eric Levine helped build that at Airbnb on the Trust and Safety team, and now they’re on a path to bring that to every business on the planet. They are world experts at enabling one of the most fundamental forms of safety— the physical safety of where you lay your head at night— which has resulted in a brand new way to travel that has touched over 150 million people and over 2 million people staying in an Airbnb room per night, more than any hotel chain in the world.
Trust and safety matters a lot to online marketplaces, and this team is going to bring that level of safety to billions more people through their next company Berbix.
Real software for ID verification, not just web forms and a front for Mechanical Turk.
The gold standard for identity as of today is still the upload of government issued identification including passports and drivers’ licenses. Even after uploading these documents, most companies still use verification by a human being. What results is often a kludgey, multi-step process that typically involves web forms and human verification steps that are error prone and sometimes let fake IDs through.
Using humans is expensive, error prone, and literally defeats the purpose of having a process at all. What’s more, since you can’t really know if you are letting false positives through, you don’t actually establish the trust that is central to doing business online.
How it works: The thousand paper cuts of doing it right
Berbix actually provides the full flow for how to get an ID verified on a fully automated basis. Here are all the things they have to do to get it right.
- Instant image quality verification
You need clear and unobscured images to be able to do this, and you need to be able to tell that immediately so that you can let users upload a better one right away. Delay equals a lost customer. - Automatic ID type identification
Users often get confused and upload the wrong kind of thing: military IDs, school IDs, work place IDs and credit cards. Nobody else does this instantly. - Fake ID blocking
Berbix is also the only one with no human-in-the-loop that can actually flag and find fake IDs immediately: flawless ones that otherwise would confuse or trick humans are found this way - Deep specific data on IDs at the state and nation level
The platform knows about specific state-issued IDs, by nation, province and state, and because it’s codified in software, the system gets better over time and doesn’t make false calls like a human would.
Trust and safety is the cornerstone of the Internet
The world’s largest bazaar operates open and freely because identity exists. In cryptocurrencies, it’s no mistake that one of the biggest issues is countering Sybil attacks: it’s a way of operating multiple identities actively at the same time to undermine the authority and power of reputation systems. While crypto solutions do it by making it costly (e.g. proof of work or proof of stake consensus algorithms), the non-crypto world can thankfully rely on state-issued IDs.
Berbix takes the simplicity of the offline world’s identification system and brings it online. We’re proud to be a part of Berbix and can’t wait to see the kinds of new marketplaces that can thrive because it exists.
