Our Twitter
search
Flickr
search
  • IMG_6494
  • IMG_6493
  • IMG_6491
  • IMG_6490
  • IMG_6489
  • IMG_6488
  • IMG_6487
  • IMG_6486
Blog-Author_Location

Social media listening tools today simply do not accurately report author location, even though they often say they do. Here’s why.

When you look at your social media dashboard — or listening tool, or river of news — most of the data you see in your tool was gathered or created in one of the two following ways:

  1. Either it was pulled directly from the metadata attached to each post (for example, the publish date, the author’s name, the URL), or
  2. The post was analyzed and interpreted in some way by the software.

But location data is different. In most tools, the physical location of the author is actually determined through a hybrid of the above approaches, as follows:

  1. Website registration information: The tool might determine the location where the site is registered through a reverse lookup of the site’s registration information. Such an approach is not reliable when the property is registered to one entity and authored by another — or authored by several others, in the case of a group blog. Further, the vast majority of social media conversations occur within social networking domains and platforms, rather than on blogs registered to individual users.
  2. Top-level domain: The domain “extension” that can indicate the country of registration, for example: “.ca” in Canada. First, many of the flaws in the above bullet also apply to this technique. In addition, a forum is hosted on a Canadian domain certainly will not restrict participation to Canadians only; even if they did, there is a vast difference between an author in Quebec and author in the northern reaches of the Yukon Territory.

As a result, automated author location analysis is not very reliable.

As more companies rely on listening tools for business decisions, it’s vital that we all understand the true quality and level of reliability of the data in our monitoring tools. The simple reality is that most people assume that the data in their listening tool is reliable, and that’s simply not always the case — especially for author location data.

While location-based services like Foursquare and Gowalla can provide reliable author location data, and Twitter enabled geo-location reporting months ago, very few people use such those services. In fact, less than 1% of all Twitter users report their location when tweeting, and Pew recently reported that only 4% of online adults use any location-based service. Further, those users who use location-based services submit a small portion of their social content with location data attached.

For now, we should all place very little faith in the geographic data within automated tools until self-reported geo-location becomes more prevalent. In fact, we suggest that all of our clients do the following when considering the use of author location data in social media analytics and reporting:

  1. Only rely on geographic data that is (a) self-reported or (b) verified by a human analyst.
  2. Ask your listening provider how their tool determines author location, and the extent to which the information is truly reliable.
There is 1 comment for this entry

I agree that location can be a hard thing to determine. IP address aren’t accurate (if you looked up mine from work it will say I’m in Mountain View California, but I’m really sitting here at my desk in Toronto, Canada).
Our Sysomos software actually gathers location info from the information that users disclose. That means that all our information is gathered by what the author has told in their bio, location space on Twitter, etc. We never guess. People can lie about their location (although I’m not sure why anyone would really do that), and there are also people that don’t disclose their location, so their data may not be 100%, but we’ve found this to be the most accurate way to find location so far.

Cheers,
Sheldon, community manager for Sysomos

11:53 am November 10th, 2010
Leave a Comment
Name (required)
Website

CAPTCHA image