You would never know unless you learn from others’ experiences.
A website where anyone can share a habit they have, how they started it and why, and what were the surprising results. For instance: after drinking coffee for one week without any sugar, it tastes better and having it now with sugar is just not as good.
As a good measurement of what was the right answer.
Allow users to donate how much they want (maybe even less than 1 dollar cent), based on their satisfaction of the result they got in the SERP. This will give a better indication as to what is a really useful answer to a specific query.
If you need to make a decision and don’t have enough information.
Ask a lot of people (using your social networks for example) what they would do, then do the opposite. The “common knowledge” is what get the majority to where they are – average. You have a much better chance of getting better results, doing the exact opposite of what they would do.
For many of the things you search the web for, search engines will have an algorithm to show you what they think are the best results for you. Some try to deduct from information they have on your behavior (search history, clicks, etc.), what would be personally best for you. They also offer some simple options for more advanced search, like similarity or boolean operators. Wouldn’t you like more control in an easier way for better results?
Allow the user more control over what he thinks is best. Allow him to say “I put more weight on anything by or connected to the following people when you are returning results for this search,” or “don’t show me anything that relates to an individual concept or a subject.” All of these options should not rely on me deciding what keywords are relevant or not, but more on a conceptual level. The algorithm should be able to deduce in a useful way what belongs to these concepts, people, situations, etc., without you having to “spell it out.”
To get ideas on what is working and what’s not.
Private members only club. You can buy a membership for full price, or get a growing (depending on your contributions – the number, quality and popularity of experiments you publish) discount if you publish your tests and results. Published experiments can be anonymized, but the club operator checks all tests for validity.