J Hackett

Recently I've been interested in the real value of a Like on Facebook. I see it breaking out into two distinct value contexts; 1. The value of a Like as a reliable data point, and 2. The value of a Like in terms of user engagement (traffic, purchase conversion, advocacy, etc...). Both have different implications regarding the best way to use them to align with strategy. Being a programmer I'll speak to Likes as a data point as I think their value in terms of user engagement is a bit more subjective and more dependent on the strategy being employed.

The reliability of Likes as a data point breaks down into two different categories, the value of total likes as a measure of interest in a particular object (brand, product, service), and the Like as a indicator of a particular user's interest in said object (music, books, products, brands, etc...). Unfortunately the former is misleading as it's not a measure of unique users, but is actually an aggregate of a number of interactions with the object being "liked." The actual number that appears is a roll up of unique likes, comments on likes, likes of those comments, shares of a link to an object, and sharing the link via a Facebook message. I realize that this is a design decision by Facebook to ensure objects have higher amount interactivity. The catch is end users interpret this number as an exact measure, when it is not exact. Actually to add another layer of complexity it isn't a real time value, but matriculates in over some increment (acceptable and expected considering the number of users interacting with the platform. This accuracy issue is a problem when creating applications or experiences that intend to use the like as an exact measure (it's actually against FB Policy to use them as a voting mechanism in promotions or contests). Another consideration that I learned about the hard way is if there is a reason to game the accuracy of a like button malicious users can do so with browser hacks like greasemonkey scripts.

As per likes being a meaningful measure of defining what a user actually likes it is a much bigger miss in my opinion. The barrier to like something is so low that i don't think many end users' think hard about it before pressing the like button. I doubt they're thinking this is a thing or service that really helps define my interests and I want my friends to see it. With the change to the function of the like button to act more closely to how share button worked users may start to think (a little) more about what they choose to like since it takes greater prominence in their friend's newsfeeds. I think this is a great measure to improve how users utilize the like button, but there is another critical flaw. My biggest grip with likes is there is a low integrity to the data. It is on the various developers implementing open graph to ensure the accuracy of the meta data being associated with an open graph object. I've seen a high level of variability in how well this is implemented across the web, and there is no normalizing of the data. The data that is returned via the Open Graph API is also very basic for example this an excerpt of the output from my likes.

      {
         "name": "Bud Light",
         "category": "Food/beverages",
         "id": "54876245094",
         "created_time": "2011-07-05T13:57:13+0000"
      },
      {
         "name": "Grey Goose Vodka",
         "category": "Food/beverages",
         "id": "83969398808",
         "created_time": "2011-06-28T16:30:52+0000"
      },
      {
         "name": "Nike Football",
         "category": "Company",
         "id": "51212153078",
         "created_time": "2011-01-26T16:36:25+0000"
      },
      {
         "name": "Oreo",
         "category": "Food/beverages",
         "id": "114998944652",
         "created_time": "2011-01-26T16:36:18+0000"
      },
      {
         "name": "Golden State Warriors",
         "category": "Professional sports team",
         "id": "47657117525",
         "created_time": "2011-06-16T16:51:49+0000"
      },
      {
         "name": "Smirnoff",
         "category": "Wine/spirits",
         "id": "216299650110",
         "created_time": "2011-01-25T23:28:42+0000"
      },
      {
         "name": "Cointreau",
         "category": "Food/beverages",
         "id": "192869992789",
         "created_time": "2011-01-24T22:05:22+0000"
      },        
      {
         "name": "Converse",
         "category": "Retail and consumer merchandise",
         "id": "23402039579",
         "created_time": "2011-06-01T19:11:29+0000"
      },

This data doesn't really tell a very deep story about who I am, nor does give very much detail regarding what these likes are outside of giving a basic category. The categories seem to be poorly organized or not always logical "Cointreau" is listed as Food/Beverage, but "Smirnoff" is Wine/Spirits. This would obviously cause confusion if you are building business logic that depends on these categorizations being a dependent and accurate data point. It would probably be better if these items could have multiple categories or if there was a built in hierarchy for them to improve accuracy. If you make additional calls to the Open Graph API you can get additional details for each of these items, but that can be cumbersome if there is a significant number of items and if you are asking for likes for multiple users. It would be great to get some more detail in the initial call, possibly adding the URL as that would maybe allow for sorting by items from a specific source. For instance sorting all items liked on a specific website, this would be great for sites with large amounts of content and then sorting content based on the users' past likes (there are obviously other ways to achieve this, but this would make things simpler).

There is opportunity to enhance things; I think likes can be amended to improve their value. For me the key missing piece is the relevance of this like to the user, do they spend a lot of time on the fan page, do they interact with the brand a lot, or did they just do it to take advantage of a promotion? It would be interesting to add a measure of how much someone likes a specific item, maybe giving users the ability to rank their likes in a way that tells a more meaningful story about who they are. Getting greater context for the relevance of that like to a user would allow us to create really interesting experiences catered to the users interests in a more accurate way. Another opportunity to increase value is to add some system for users to police the data integrity, similar to how Foursquare uses their "super users" to police and correct data issues on their platform. It would be great to allow users to flag data as wrong, suggest corrections, and or merge duplicate items.

I'm certain that Facebook will continue to evolve likes as the platform grows and expands. It's just not quiet there at the moment and if they address it in the near future likes could really become a much richer and accurate data point. I'd love to hear others thoughts as I'm sure there are some different perspectives I've missed.

Update: I opened a bug on Facebook's Bugzilla aimed at helping improve "likes." This is the first of a few I will likely open on this topic to help push them towards resolving some of these issue.
http://bugs.developers.facebook.net/show_bug.cgi?id=19069

I'm still questioning the direction Facebook (www.facebook.com/places/) has taken moving into the location check ins space that Foursquare (www.foursquare.com) currently dominates. It makes sense why Facebook would be concerned with locations and building features around that dimension of the social graph. However I feel like they came a bit late to the game and didn't really deliver anything compelling or interesting. Actually I find their mobile strategy confusing since both the iPhone and Android apps are such poor user experiences in comparison to the full web version of the platform. I'm certain there will be a lot of work to iteratively flesh out Facebook Places, but at the moment I don't see its value add. It's hard to play catch up versus being the innovator who gets there first.

I think there is a limited subset of the population that is a viable market for location check in based apps and the majority of those people seem very happy with Foursquare for the time being (I'd be interested to see the actual usage numbers for Places). The game element of Foursquare is a significant motivator for interaction with their platform. Getting rewarded with mayorships & badges is why people use Foursquare, or at least that was my initial and continuing motivation for playing. It's also rather interesting to see a detailed history of where I've been, but that is a secondary feature for me. Even the deals element is much lower on the priority scale, although I think that's a factor of businesses not knowing what to do with it at this point. At some juncture I think that this could be brought to the forefront and serve as a means of enticing those not interested in the game element into using the platform, but I'd be surprised if Foursquare would mess with the priority they give to the game side of the experience.

The smart and natural progression of Facebook Places seems in the direction of localized deals this could be really interesting if it is married with credits system and the ability to make payments for things right from your phone (NFC). So the battle between Places & Foursquare in my mind is really deals versus games, which are not really directly competing products. I think that the two companies will focus on taking their offerings in directions that will not be geared to directly compete. People will use Foursquare for its game elements and push experiences into Facebook/ Twitter to spread around their social graph. I see a path where a check in through Foursquare will also be a check in to Facebook Places and allow you access to the deals both in Place and Foursquare. These two products from my stand point are not now nor will be direct competitors, and they may even be symbiotic.

Of course a lot of this is built on my assumptions, feel free to poke holes in my thinking or agree.

I've been tinkering with a simple Facebook app using the Open Graph API. The challenge was posted in a creative concepting meeting to find which friends a user is actually closer to in real life, this is difficult since people tend to friend a wide set of people on Facebook and there are a great deal of interactions that don't effectively having any mapping to the importance that person plays in the user's real life.

Initially I was going to try to focus on the number of news feed interactions between two people and I created a simple test application to study the results, I simply pulled the feed and ordered friends by the number of interactions they've had over the past few weeks. This obviously yielded mixed results since I have a number of people I'm not necessarily close to in real life interact with my news feed. This is mostly professional contacts remarking on various articles I post into my feed on a daily basis.

After tinkering a bit I decided that physical proximity would be the best way to determine a more meaningful relationship with another person, since you don't interact in person as often with casual Facebook friends. The data from photos seemed like the simplest way to determine people you've actually been in close contact with and those that you spend time with most frequently. This assumption assumes that you have a fair number of photos that you are tagged in with other friends and that you share that data.

So I amended my initial test app to also include users you are tagged with in the same photo, instead of throwing out the news feed data I choose to add the photo data and give it a stronger weight. This covers a result set being generated even if you only have a small set of tagged photos. To add another level of interest I also added comments on photos to the equation since people that are good friends seem to comment on group photos.

The end result is roughly this:

newsfeed comments = n
photo tags = p
photo comments = c
gravity score = g

n*1+p*5+c*3 = g

The results so far seem to be fairly accurate for most people that I've had test it, although I'm sure there is a lot that can be done to improve the accuracy. I'm really interested to see if I can use the new friend's page data to make this smarter or if that will simply make my little experiment obsolete. I'm eager to also integrate places data as that becomes more popular, but at the moment I don't think it gets used enough to justify adding it.

Have a look at the project and let me know what you think Social Gravity. FYI the code is a bit rough and sloppy.

Ok I've finally decided to add a blog to my site, since I have no shortage of opinions these days on social media & digital strategy.  An interesting side note is that this blog is a wordpress install mix into codeigniter. So check back soon I'm sure I'll have a lot to say. On the post detail page I've included a Facebook comments widget so give me some feed back if you feel so inclined.