Korva is a scroll of the internet’s best ideas.
Software could benchmark you against peers
One place, isolated views
First-generation SaaS kept each customer isolated. The second generation went multi-tenant, which made it cheap for the vendor to stand up a new customer and cheap for a customer to try something. That generation has not fully used what multi-tenancy makes possible. The data from every customer now sits in one virtual place, and each customer's experience is still isolated, customer by customer.
The same data could be abstracted and anonymized so that every company sees how it is doing relative to its peers and what the benchmarks are. A cross-customer view like that would carry real network effect value.
Fraud signals across customers
Stripe Radar, the fraud product, shows the beginnings. Fraud coming from one IP address on one customer's API makes it probably fraud when that same address shows up at another customer. That is a small example, and there is a lot of room for SaaS companies to anonymize data across customers and improve each customer's experience.
How do vendors anonymize pooled customer data?
Common methods remove identifying details, combine figures into groups so that no single customer stands out, and require a minimum group size before showing a number. Some systems also add small amounts of statistical noise, a technique called differential privacy. The goal is that a benchmark reveals the range of a peer group without exposing any member. The US Census Bureau applied differential privacy to its 2020 census data.
Why does shared fraud data work so well?
Fraudsters tend to reuse stolen cards, devices, and addresses across many merchants. A signal seen at one business therefore counts as evidence at the others. A card tested for small purchases at one site often shows up within minutes at another. Stripe's documentation describes Radar as learning from payments across its whole network.
Why do pooled benchmarks grow more valuable with scale?
Each added customer sharpens the comparison, because the pool can be split into narrower peer groups by industry, size, and region. A small pool might only support one broad average. A large one can show a company how it stacks up against firms very like itself. Credit bureaus built their business by pooling data that many lenders reported.
More from Invest Like the Best with Patrick O'Shaughnessy
