tinyclues

Behind the scenes, our platform incorporates the latest advances in machine-learning. Rather than looking for naive categories ("this product is for middle-aged men") or deterministic rules (retargeting), our system bases its predictions and recommendations on adatabase-wide consolidation of all the relevant bits of information. Our technology finds the tiniest clues and picks up low frequency patterns:
-the hidden meaning behind a pseudonym is no secret to us: for example, if you picked a pseudonym including the name "Yoda", we can predict what other pseudonym you most likely considered. 
-by analyzing linguistic patterns, we can guess the gender of most email addresses, even lovelybutterfly@example.com or dragonkiller@example.org 

Behind such results, advanced algorithms sift through high-volume raw data, detect correlations and learn the implicit semantics within. What we do with emails and pseudonyms, we also do with all the data available in your database, from user data to IP logs, from purchase graphs ("who buys what") to navigation patterns. 

Another series of mathematical tricks allow us to package this multitude of patterns into reliable predictions and actionable insight.

So when you are recommended a product, it might be because of a completely different item you clicked on a year ago, because of your first name, because of your zipcode, because of your browser-of-choice — or, more likely, because of a subtle combination of many such facts.