Context is required for Machine Learning to understand Human Behavior

Ok someone takes a photo of a building right in front of them - what is the name of the building?

Unless that building is the Taj Mahal - it could take a huge library of photos (akin to Streetview's database) - and a lot of processing time - to work out what and where that building is. Then - if the building happens to be a generic block - similar to thousands - you have no chance.

However, it you know - the location - where the photo was taken - thinks become much simpler - and a hacker without access to a super computer (or Amazon's redshift) - will be able to solve this problem.

The key here is context.

The same applies to human behavior,

We can look for patterns - in someone's browsing history and search terms - however those patterns become more accurate/reliable and powerful when we have their clickstream + context:

• Their location;
• Their actual income (as opposed to census data)
• Examples of their shopping bill;
• Their relationships;
• Their Gender;
• Their work history and occupation.

Context - is therefore what the Market Research industry can add to machine learning models on insight.