Maybe you can clarify what you mean by 'values of a person' which the recommendation engines are considering. You mention measurable by actions, behaviors. Wouldn't you consider Amazons recommendations engine behavior based. They bought the book, wrote a review, others rated the review, etc. etc.
To use Pandora as an example, they do a ton of metrics to measure what a song sounds like, however their most valuable metrics and the reason the system keeps getting better is the collaborative filtering Pandora receives from the thumbs-up/thumbs-down of users.
So, unless you can give an example of other values, I'm not sure I'm understanding that part of your question.
As far as a business idea, many have tried to sell access to a recommendation system to run on customer data. I think most of these companies fail because recommendation systems are difficult to measure, and look like a high cost to customers. There are lots of recommendation algorithms available which can fairly easily be built into most solutions.
You mention strands which I haven't looked at in a long time, and now is a 'digital log for active people'.
Take a look at echonest. they've been around for quite a while, and may be one of the better music recommendation companies out there, but I wouldn't exactly say they are hugely successful.
I think there are lots of 'good enough' implementations of a custom recommendation engine that can be made by most companies who are already creating value with products or information.
If you have an idea or a market you are looking to build a recommendation engine for, I'd suggest building the entire company, and not just the recommendation engine. Recommendations are not a unique proposition in most markets.