User profiling for personalized apps

Question

I am trying to create a taxonomic user profile for a personalised recommender system. thought this system we can make user behavior assumtions such as:

• User A from country A is interest in Parks 20% more than Restaurants
• User B from country A is interest in Restaurants 10% more than Parks

We can even tell if the Restaurant serves italian food (for example). How can i assume that User B likes the Restaurant A because the serves italian food?.

Is that possible? and how? Can a **Persona** improve the recommendation system/engine?

Answer 1

The short answer is that it is possible, but you need to be aware of the limitations of what user analytics can tell you about them.

What you are suggesting seems to be using analytics to build a recommendation engine that probably uses some AI or machine learning algorithm to continually improve its accuracy. While this is useful for providing general recommendations, I think that the extent to which you want to use this (based on the type of data that you are collecting) might require more than just pure analytics.

Of course, this will probably be more accurate if you are trying to predict the same or similar types of behaviour (e.g. Park or Restaurant related), but it will be more difficult once you have to extrapolate this information and work out if they will be more interested in Theme Parks or Takeaway Shops.

Analytics is usually the starting point for you to focus on specific areas of user behaviour research, rather than the answer for all the questions you have about the user behaviour. This is because analytics is best suited for finding the common pattern and trends in a large dataset (to provide a certain confidence level in your predictions), while user research tries to reveal the various factors and context that influence individual decisions by observing the actions and behaviour that is abstracted away by the data collected.