Search is evolving fast. With search engines growing smarter every day they will be facing dilemmas they didn’t have before. Some choices have to be made when developing their algorithm, but it probably will result in a considerable amount of brain crunching to make those choices. I will discuss some of those dilemmas here.
Relevancy or broad matching?
Understanding which words have a similar meaning gives search engines the possibility to show results about topics similar to the search query (broad matching). This gives the user a broader spectrum of information and therefore a wider choice in results. But this also means that some results will not be as good as when search engines only show results for exact matches. For example most people see apes and monkeys as the same kind of animals. So whether I’m searching for apes or monkeys when I’m looking for an image doesn’t really matter, so broad matching is permitted. But when I’m looking for information on were apes live it does matter, so broad matching is not permitted. Search engines will have to make a choice whether they are going to prefer exact matches, and therefore better results, or use broad matching to show more different results, and therefore give a more complete list of results including results that don’t match the search query as good as exact matches.
Ranking real time
With the introduction of real time results the question of ranking these results was a hot topic. Since real time results can’t be qualified on historic data they have to be ranked with other data in mind. There are a few possibilities here. One is ranking them on factors belonging to the sender. The sender has historic data and can therefore be qualified. But historic data of users doesn’t say anything about the quality and relevancy of the specific message. Another way to rank results could be based on the quality of the specific message, based for example on retweets and replies. But it takes more time for a search engine to qualify a message before showing it in the search results, so they aren’t really real time anymore. Search engines have to make a choice here whether they are going for real ‘real time’ results based on sender quality, or they choose to go for qualitative better results but not completely real time.
Interpreting user behavior
Whether user behavior is going to play a role in ranking websites isn’t a question anymore. How to interpret user behavior is a matter that’s going to be full of dilemmas though. Search engines only have access to user data within the engine itself to determine the quality of search results (yes, I know that’s somewhat naive). So when users click on a search result and then return to the search engine to click on another result, it says something about the quality of the search result. But what does it say when a user returns from a search result? Couldn’t they find the information they were looking for and are they trying another result? Did they find the information they were looking for and are they going to verify or compare it with another result? And what does it say about a search result when a user doesn’t return or enters a new search query? Is that a good or bad signal? User behavior is probably going to play a more important role in ranking websites but identifying and qualifying behavior will be a very tough job for the search engines.
Determining user intent
Search engines are pretty far evolved in offering the correct results for users based on their intent. Whether users are searching for local companies or the latest developments about current events, search engines can offer specific information targeting that need (e.g. local listings or real time results). However, to determine the user intent search engines need much more information than just a generic search term. For example, when I’m searching for “Albert Heijn” (a Dutch supermarket chain), how would a search engine know whether I’m searching for a local store, the latest news, the latest stock price of my shares or just general information about the company?
They could use historical data. But historical doesn’t always tell everything about my next search. Search engines can only determine my current intent by looking at the information I’m providing at that moment. So when I’m using generic search terms I’m not providing the information search engines need. Google seems to realize their intelligence to offer search results based on intents has been developing faster than their users have come to specify their intent. That is why they changed the interface of their search engine, to make it easier “to find exactly what you’re looking for”. Somehow search engines need more information about user intent to provide the right results out of all the different information they can offer about a topic. How they are going to get that extra piece of information will be a great challenge.
The road to perfection
As you can see, getting better doesn’t always make things easier. Search engines will have to overcome a lot of issues to keep getting better. Besides the four I highlighted here, there will be more dilemmas on the road to perfection. Do you know some others?