WebKite Blog

Periodically we publish thoughts and ideas about he state of the web and the state of structured data publishing. Topics range from small business marketing to cost efficient advertising.

When Machine Learning Meets Online Advertising

 

Since watching Star Trek reruns with my father, I have always had a deep fascination with machines that exhibit intelligence. In my youth that meant robots, of course, and the AI behind those machines. But in my graduate studies at Carnegie Mellon’s Language Technologies Institute (in which I actually did get to work with this robot – fun!) I learned to eschew the term “AI” for the more prosaic and precise science of “Machine Learning.”

Machine Learning is simply the discipline for the study and application of “decision systems that improve over time,” where “over time” is actually shorthand for “with new observations.” As simple and powerful as that is, it has wide application, and has both borrowed from and contributed volumes to its parents: computer science and statistics. Machine Learning sidesteps a lot of the intractable philosophical issues of Artificial Intelligence while remaining quite pragmatic and highly useful. It’s what’s behind all modern speech recognition, recommender systems a la Netflix, and self-driving cars for starters.

At WebKite, we are harnessing Machine Learning to grease the wheels of retail commerce.

There’s a famous quote by an 19th-century retailer John Wannamaker: “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” Over a hundred years later, this is still a huge problem for today’s retailer. But deciding how to spend advertising dollars is another great case for Machine Learning.

Using Machine Learning and automation, we built a platform we call Radius. Radius plugs into the search engine marketing (SEM) ecosystem and learns to make better decisions about how to spend those advertising dollars. For example, ad bidding is adjusted higher or lower for search terms that result in better or worse customer conversions. We get to help retailers advertise to people who are ready to buy their products, and not waste their money otherwise. We get to see those retailers get new customers at a fraction of the costs of other advertising alternatives, and I get to develop a machine that exhibits a form of intelligence.

Inventory-based advertising is a better way to advertise

Since we're generating specific ads based on a business' inventory, ads are only shown to customers that are ready to buy those products.

This mitigates waste and ensures that you only get clicks that have a better chance of leading to a sale. Inventory-based ads help you skip past the bartering and get right to the sale. Meaning you can spend less time maintaining ads and more time taking care of your business.

Start Using WebKite