How HumbleWorth Works

HumbleWorth estimates the value of a domain name on a domain name auction site like Godaddy Auctions. The analysis ignores the underlying business value and focuses exclusively on the domain name's value - analyzing the words and structure of the domain.

HumbleWorth builds on the work of talented researchers from the likes of Microsoft, Eleuther, and most importantly Huggingface, who have built massive, freely available language models.

By starting from these models, our models have gained an understanding of human language on an internet scale. We based our pricing model on language models because domain names are essentially just words and phrases that you can own.

We refined our language models based on data from a variety of sources, including dropped domains, domain names used on active websites, domain marketplaces, and a 20-year history of more than 3 million domain name auctions from our friends at DNPric.es.

However, predicting a domain's value alone tends to be volatile. Instead, we developed a fairly novel machine learning model that predicted domain prices at various levels.

Different parts of the network work together to "vote" on their prediction. We combine these votes to form a probability curve determining the probability of a domain selling at different prices.

Compared to other appraisal tools, this context-rich information is far more helpful.

Our machine learning models constantly improve by adding new data from the ever-active domain auction marketplace. Our predictions will improve over time as we identify new pricing trends.