# How to model engagement score over time

I am trying to model an individual's 'engagement' with a document over time.

Example: A user receives a document at time t, they then download it at t+1. They could then come back to the same document and download it again (t+2). This second download action is seen as an indication of increased engagement and therefore the weight of this action should be more than the weight of the first.

• I'm look at downloads over a 28 day period.
• The total score for the user against a document should be the sum of all these download actions at certain times.
• I want the score for a user's engagement for a particular document to be on the scale from 0-1.

Any ideas for how to model this?

Could happen that you have just a single variable model, where numberOfdownloads is the main variable. You would have to definite what is an Engaged reader, say someone who downloaded more than 40 times in 28 days period (you should check what the data says to choose the number that defines Engaged), for all others below this value, you can call them notEngaged, then changing them into 1 and 0 respectively, you can build a simple one variable logistic model that will give you a value between 0 and 1.