One of the most frustrating elements of any Internet listening project can be setting up the search terms that will result in the most relevant data. If you’re not careful, valuable time can be wasted on a bad query while irrelevant results pour in. (For an example of bad queries, look at this post.)
Your search is more than just a search. It’s the first step of a research project. Our job is to make your project successful, so we’ve had a lot of experience setting up and testing search queries to give clients the best possible statistics.
Here are the steps you need to take to define and refine your query:
1. Know what you’re looking for—have a defined topic.
If you went into a library and told the librarian you were looking for stuff on history, the first question any good librarian would ask is “What kind of history?” Otherwise, you’d be poring through the entire breadth of human history just to find information on poodle skirts.
It’s all about narrowing down the field. If you want to find out whether a used 2003 Ford Expedition truck is a good buy or not, you’re one step ahead of the game because you’re looking for a specific make, model, and year. A search for “Ford” isn’t going to do you any good. You have to define what you are looking for specifically because the search won’t do that for you. It gives you everything.
2. Think like a machine—narrow your topic into a concept.
Let’s say you want to run a query about politicians who put a lot of political pork in the bills they submit to Congress. A machine doesn’t know what “pork” means. Machine language would happily spit back recipes for baconated grapefruit with the same enthusiasm as information about a spending bill. You’re liable to get a lot of irrelevant results unless you bear that in mind.
3. Search for what people are saying, not what you want them to say.
This is a common pitfall. If you are looking to define what customers who frequent movie concession stands are saying about the products there, don’t search for “concessions.” Chances are, few of the relevant URLs will contain this word. Use common language. You’ll get better results if you think about how people actually talk.
People are more likely to be talking about the popcorn, hot dogs, or Cokes that they had at the movies. They may also talk about prices. All of these are important words to consider.
Another example: Use “laid off” or “lay off” before you use “unemployed.”
4. Find them and learn their language.
Sometimes you have to do a little research before you start. You wouldn’t go to a mommy blog to find out what people think of the public option. Do a test queries to find out where the conversations about your topic are happening.
Now it’s time to find out specifically what’s being said and how it relates to your concept. Read the blogs that cover your topic. Even better—use a tool like semantic analysis to learn the language that your targets are using. Now that you know the lingo, it’s time to refine your queries.
5. Know that you may not find what you’re looking for.
Expectation management is important. As shocking as it may be, there probably aren’t a lot of people talking about your original Cap’n Crunch/James Bond fan fiction or the rubberized oscillating spammer widget that you marketed to the Amish.
That could mean many things. Are you not getting the right results because you haven’t used the right terms to find them or because the results you want aren’t there? If nobody is talking about your widget, look at it as an opportunity! Running a query on popular Internet widgets might reveal to you that the Amish aren’t real big on computers—and that’s valuable information.
{ 1 trackback }