One of the problems about Twitter is that it was never meant to be used for data mining. It displayed conversations of those you followed, and you could see whenever someone mentioned your username or sent your a private message. In fact, Twitter didn't even have an integrated search engine until it bought one from Summize in July, 2008, 2 years after they had launched their service. Fast forward to today, where the most recent stats last year hint at Twitter being one of the largest search engines that exist and you get a sense of how important searching for data on Twitter has become in a mere 3 years.
Just by looking at Twitter, however, it's seemingly impossible to find a lot of data that you might be looking for. That's where Twitter's APIs come in. Instead of creating everything themselves, they let 3rd party companies come up with thousands of applications for the needs of any given user (of course, recently Twitter has begun to buy back some of these strategic 3rd party apps like Tweetdeck...).
Here's a simple task to prove my point: I work for a company and am targeting Twitter users who live in a certain area. How do I find them? I decided to experiment with an actual city and show you the different results I get with what limited capabilities Twitter has and 5 ways to search for people by location that I recommend.
For the purpose of this exercise, I used an actual case study from a client who needed to target users in Guadalajara, Mexico. Any Twitter tool must be universal in scope, and the limited numbers of users there when compared to a major metropolitan American city made it a perfect test case scenario for this blog post.
Let's first start with Twitter to set the standard for our 5 alternative search methods. How do you search for people by location on Twitter? You simply enter the city name in the search bar, and you will get the following:
While the primary results deliver tweets that have the same term, notice that the right-hand side shows "People-results." It only lists 3 people who happen to have Guadalajara as part of their username. However, when we hit "view all" we are now shown the following screen:
The screen being shown has actually forwarded us to the Who to Follow search result for Guadalajara in the bios of users. Unfortunately, the result was very limiting in terms of numbers - and there were no filtering options.
CONCLUSION: 20 Users Found, No Search Filters, No Data Outside of Bio. Not Usable.
Now let's compare how the 5 alternatives stack up: (and make sure you read all 5 to get a comparative sense of what each delivers that the other doesn't)
Some of you might not have heard of this tool, from the creators of Tweetspinner, but if you haven't you may be missing out on a very valuable application. FollowerWonk allows you to scan bios for terms, location, full name, as well as a range in number of Twitter followers or friends. There's a few other tricks you can use FollowerWonk for, but here are what the results looked like:
It might be hard to see, but that's because FollowerWonk provides you with a plethora of data: screen name, real name, location, "wonk score," tweets, friends, followers, and how old the Twitter account is. Furthermore, the search could be sorted by "relevance," "follower count," and "friend count." Not only that, the number of results that FollowerWonk delivered was the most of any tool tested: 10,823!
CONCLUSION: 10,000+ Users Found, 3 Sort Filters, 5 Data Sets. Best of the 5.
Twellow is one of the most famous user directories, so you would expect to see some great results here. While the numbers, although fewer than FollowerWonk, were impressive, the data displayed was little and had no sorting options:
All of the users who were associated with Guadalajara, presumably as their location, were listed in descending order by their numbers of followers. It's nice to see the number of followers, but with no ways to sort or other data shown, the tool lacked some of the sophisticated features that FollowerWonk had. Search results returned were 6,105.
CONCLUSION: 6,000+ Users Found, 0 Sort Filters, 1 Data Set. Twellow gave a decent number of results and had a nice user interface, but lack of sort filters and data provided means it is not the ideal solution out of the 5.
Another service that you might not have heard of is Tweepz. Tweepz just has one simple search box to enter a search term, but the results and data that displayed was impressive:
In addition to the expected data, we can see their numbers for followers, following, and updates. In addition to sorting by relevancy, followers, and following similar to FollowerWonk, we can further refine search results by more narrow ranges of numbers of followers and following, language, as well as "extracted entities" or keywords that presumably frequently appear in names or bios that appear on this list. Tweepz even provides us an RSS feed to further work with this data. Unfortunately, while the results were impressive, the number of users that were actually returned were less than half of FollowerWonk: 3,978.
CONCLUSION: 3,000+ Users Found, 4 Sort Filters, 3 Data Sets, 4 Refinement Filters. Despite lack of results an excellent alternative to WonkFollower.
So, if Twellow only had 6,000 users found, you would figure that the granddaddy of all Twitter profile directories, the mighty WeFollow, would have more results, right? WRONG! By far the most shocking result was that WeFollow only returned 70 results. Perhaps like so many older Twitter applications that have come and gone, WeFollow is on its way out? Maybe it is only strong in North America the United States? That being said, WeFollow did deliver one type of result the others did not: "influence":
With only 70 user results, the data is not useful at all. However, the ability to sort by influence, in addition to number of followers, is interesting. Of course, there is (and never will be) no single way of measuring influence, so the number here is as good as the Wonk Score on FollowerWonk. Search results: 70.
CONCLUSION: <100 Users Found, 1 Sort Filter, 1 Data Set. Not Usable.
The above 4 methods give you a good idea as to what types of results, data, and sorting filters you can get with 3rd party applications. But if Twitter bought Summize, where we can see that search engine in action? The answer is the 5th method below:
This is the "hidden" search bar that Twitter never integrated at is into their service and instead have you searching down the URL (http://search.twitter.com) to access it. Now, this search engine is only going to display tweets, but it gives us the ability to find who is tweeting and what they're tweeting about in a certain location. Once you select the Advanced Search button on the right above you get a whole host of fields for which you can perform your search:
While this can undoubtedly be used for many a search purpose, we are interested in using the "Places" filter which we can also try to search within x miles or kilometers of the distance away from that city. Typing Guadalajara into the "Near this place" and including the default 15 miles displayed this result:
This data can be useful for a number of ways in a combination with the results from the other 4 results above. We can do searches seeing how often people we found are mentioned in Guadalajara, who they are tweeting with, and what they are saying. Twitter also displays the range of our search on a map, allows us to search for particular language of tweets, will translate the tweets into English, as well as show us the Trending topics. Applying the Twitter Advanced Search to a portion of the results we find in, say, FollowerWonk will allow us to build a targeted list of people in a locale which we can further utilize the many fields in Advanced Search to find the exact subject matter of tweets that are our target.
The beauty about the above tools is that they are all free. Sure there are robust social media monitoring solutions that can help you filter and sort the data in a number of ways for a fee, but don't underestimate the power of free tools to help you mine Twitter to look for people by location.
What other tools are out there that YOU use to look for people by location?