CNN and Facebook have partnered together to report discussions on Facebook involving the presidential and vice-presidential candidates. The variables have not been released, but they are using Facebook Insights, a simple analytics tool where you can measure the number of Likes, Shares, Virality and ultimately total Reach. Basically, CNN’s Facebook Insights is presenting how many people are talking about any specific candidate at any given time.
It’s an interesting overview of which different demographics are talking about which candidate. In fact, the vast majority of those talking about the candidates is over 45 years of age, and conversations across all demographics tend to be roughly 50/50 for the Democrats versus the Republicans.
To say that any one candidate is succeeding more than the other, or that a candidate is most likely to win, is nearly impossible. Not because the conversations nation-wide are sticking roughly to that 50% line, but because the Facebook tool provides no insights into sentiment; the most important aspect of understanding the conversation.
The tool is an interesting insight into the popularity of each candidate. This in-of-itself may be an indicator of ballot box success, or might be if we knew the number of Facebook users who actually vote. In reality, it would be silly to assume that mentions in a conversation is an indicator of positive perception or intention to vote for any of the candidates. Is it good that no one is talking about Biden compared to the other three candidates? It might be yes, it might be no. The point is, no one knows. Since when did someone talking about someone else become inherently positive? Following mentions without understanding the sentiment of the statement will lead to flawed results.
Back in January 2012, Facebook teamed-up with Politico to measure the volume of keywords in status updates, postings, and comments, as well as the frequency of negative emotional keywords in the same. Apparently this was a first for Facebook to do data sharing of this kind publicly. Perhaps not surprisingly the results were not clear and were without any significance when correlated to the real results of the primaries. Even Politico’s own conclusions were mired in doubt as to the real value of the data, probably due to the fact that of the voting population on Facebook, not everyone is able to or does vote in presidential primaries.
When analyzing the Insights page on CNN, we’ve found that there is no contextual information that would assist in understanding the data. For example, we checked-out national data on females between the ages of 25-34. Over the last 7 days, the Democrats were clearly more popular amongst this demographic versus Republicans. However, the nation is in a sea of red over the past 12 hours.
The speculation? The Republican National Convention just finished and people are obviously talking about Clint Eastwood’s stellar performance. Does this lend to any information about voting intention? Not at all. Also, neither CNN nor Facebook provide any contextual insight into the changes on this page. A list of news headlines corresponding to the date range you’re viewing would have been sufficient for adding some context (much like Google provides for its Finance and Trends graphs), but sadly, none of that.
One of Facebook’s own told me that the data is coming from a direct pipeline, is accurate for the visualizations, but went on to state that they “didn’t feel comfortable doing sentiment analysis because of the error margin.”
With Facebook’s meteoric stock decline and general grumblings about the efficacy of its advertising platform, it would be nice to see something actionable from them. Are they selling their “insights” to polling firms, agencies, politicians, or news outlets? If they are it’s not resolving any doubts as to the value of their data. Facebook is simply not utilizing their data fully. They need to improve their simplistic analysis tools so that people can derive actionable and meaningful conclusions.
If Facebook really is the data behemoth that gave investors wet dreams – real information about real people – and they can’t effectively mine that data, then perhaps the declining worth in Facebook, especially its stock, is fully warranted.
Here’s an idea: to prove its worth to the world and restore some confidence in the company, Facebook needs to predict the 2012 presidential election, state by state.