How we used AWS to build our identity and liveness system

Identity verification and liveness checks are central to lending, and any lender who misses it will likely end up running a charity organization instead of a profitable loan business.

A loan is a legal contract between two people – I’m giving you money, so you’ll pay back on xyz date with interest. Now, for you, as a lender, to go into such a contract, you must be confident of who you’re dealing with, and that’s where identity verification comes in — making sure that Dara, who has applied for a loan is indeed Dara and not masquerading as someone else.

Also, to ensure that Dara who claims she can pay back x amount of money in xyz timeframe, can indeed pay back.

Why we built a liveness system

Identification verification is done with documents and data the customer provides. For example, when a borrower provides their NIN, BVN and other means of identification, you, as the lender, should be able to use their picture to confirm the trueness of their identity. This means at the point of onboarding, a lender must be able to verify authenticity.

Sadly, humans will be humans and often try to weasel their way through your many checkpoints with stolen identities. Typical. Luckily, this challenge has been solved with liveness platforms like — Onfido and SmileID. 

While liveness platforms effectively solve the problem of stolen identity, most of these services are quoted in dollars. Unfortunately, we are in Africa and constantly battle with depreciating currencies. This means that if Onfido charges 40 cents, and 1 dollar equals 1,500 naira, it will cost 600 naira for a single verified borrower.

Therefore, a large ecosystem like Lendsqr, which caters to up to 1.7 million borrowers, would spend roughly 1 billion naira.

In Nigeria, 1 billion naira is a lot of money for anyone. The real question is, what about non-performing loans or borrowers who apply multiple times?

As a leading loan management platform that provides liveness and identity check measures for over 3,600 lenders, we can’t stand to place this heavy financial burden on our lenders, or they’ll run at a loss.

As people who’ve previously excelled at building internal infrastructures, we asked ourselves, “What if we build our liveness platform? What would it take?”  

Featured read: How we dumped SendGrid, built Mailer and solved our email delivery troubles.

Building our internal identity and liveness system

Lendsqr is an ardent AWS stack user; while we are also tethered to the Microsoft Azure stack, most of our complex tech is with AWS. So, we crafted and designed our process around AWS using Amazon Rekognition.

One of the significant features of our identity system is the image compare feature that utilises face recognition APIs. It allows our lenders to detect, analyse, and compare borrowers’ faces for various use cases, including verification, cataloguing, and general public safety, which runs at the cost of 1000 image checks for one dollar.

The most affordable you can find. It means that even when the dollar is 1,600 naira, we’re spending #1.60 per image check. Automatically, Lendsqr spends about 2.7 million naira for 1.7 million borrowers. This is a huge disparity (which we’re grateful for) from the 1 billion naira originally. As for the liveness part of our system, we use the AWS liveness SDK, which comes at a reasonably lower price.

More advanced ML / AI-based checks are more expensive, but they are still at least 1/10 cheaper than the cheapest service we could have signed up for.

Ultimately, our identity and liveness checks summed up to about 2 naira per customer. For that reason, we didn’t have the heart to charge our lenders anything. 

Featured read: How we built our URL shortener (Monstrator) as a replacement for Bitly.

Saving a ton of money with cost-effective solutions 

In an era where financial prudence is often underrated, we’re loud propagators of building cost-effective solutions to solve and fix our troubles.

These solutions have proven to be a major boon for our bottom line, resulting in substantial savings of millions. Why should we settle for a more expensive alternative when we can achieve similar, if not better, outcomes at a fraction of the cost? What could be better?