Website Optimization with machine learning.

I'm looking for assistance in conducting the initial research on machine learning applications that support website owner (publisher). Answering a question: How could a publisher benefits from the deployment of machine learning? Publisher here, is a website owner that provide content to the web, for example, websites for surfers, cars owners, people looking to buy properties, etc. . This is profoundly different to e-commerce case, so the most common aproach - building a recommendation engine for e-commerce website - are not directly applicable here. Additionaly the research should cover a case when subsription model is not available. So, we should also make an assumption that website visitors don't loging to the website and don't provide data to be used in the next session. I've already researched how to gather data about website visitors (tracking data, fingerprinting, real-time visitors behaviour on a website). Now I want to find how to use this data to bring value to the publisher. The main focus on the research should be on approaches based on customer's level data (as opposed to aggregated data). Answering the question: how we could influence visitors/customers behaviour by treating them individually in a way that allows us to get more revenue? This challenge has a lot in common with "Real-Time Bidding', but without bidding itself. Instead of bidding, we look for optimal actions from the publisher perspective. It's mostly about modelling the customer journey while he/she is traversing through the website's subpages, proposing an offer or invitation to chat or showing different content. Right content for the right customer at the right time. Like for the traditional approach for modelling customer journey but constrain to tracking customer real-time website behaviour. The following area should not be considered: Real-Time Bidding, Heat Maps, recommendation systems. Well-documented (preferably by publications) examples of success stories in this area are appreciated. I prefer to divide research into several steps and move to the next level base on the previous results. Within the first phase, I'm looking for the answer in a crude form, just the essence: bullets with the topics plus links to the source materials. This will be a base for further research and the description. About me: I'm Data Scientist. I'm looking for a support in my research.