When should I update a recommendation engine?

Question

[I asked this on StackOverflow and was told it would be a better fit here]

I am including a basic recommendation engine in a very small project for my final exam. I understand the code and the math but I am not too certain as to when I should update the values.

In this project it would be fairly simple to update in real time but in the real world would that not take an extreme amount of cpu time? Would it make more sense to run the algorithm on a schedule?

In the project I am using linq2entities, storing the purchases in a table that I use to calculate the correlation coefficient. I only have a couple of thousand users in the db but I am trying to get an idea of what would be more efficient in practice.

Answer 1

After thinking about it I came to the conclusion that a schedule would be best.

On a small scale app updating in real-time would not be difficult although once a week would probably be best.

For large scale applications I think once a day would be optimal, allowing the best possible use of the recommendation system.

Really should have been obvious from the get go.