ML At Scale: Why We Visually Audit Millions Of Websites.

Compute-optimized servers, talented engineers.

In June of 2019, several of our web design developers started asking us for an automated way to visually audit the quality of a website, based on the initial view of the landing page. As growing agencies and freelancers, they were interested in branching out from small, localized clients to more established customers with outdated, but high-traffic websites. Sifting through thousands of domain lists manually is not something that they were willing to budget for, both time and resource-wise.

Our engineers combined their input and designed a sandbox environment with the combination of our in-house domain database and machine learning algorithms.

The result of that commitment is that we trained a state-of-the-art model that can segment websites based on their visual representation, to a 96% accuracy. Our clusters sort millions of URLs into two groups. Group One are those that look great and do not any improvement. Group Two consists of those that need a little TLC, a complete makeover, or maybe just a great CTA added. Those are the websites that we serve on our API. The end goal is to make sure these automated audits generate simple yet effective metrics that you, as a developer or prospector, can use to jumpstart the crucial aspect of communicating your services to a potential client.

Sure, you will probably say that the context of what is outdated is subjective. Agreed. However, there is a universally implied understanding of what a broken or outdated website looks like. Disorganized, cluttered, bad colors, textually-dominant - you get the idea. That is everything what our models are trained to detect. Additionally, all training is constantly performed on a myriad of different categories of websites, so that the model can understand and properly identify quality across all industries and types. Think local pizza places, metal fabricators, skin care clinics, and software apps.

So why do we sift through millions of websites? Because it drastically decreases the amount of time you need to prospect new clients and increases the time that you can spend making websites great. These businesses are hard at working providing their services to customers. Help them make their online presence work for them. Get started today and start improving the web.

Pavel Sazonov - Founder of SiteAffirm