I’m just as skeptical (and maybe more so) than others when it comes to Facebook. But some posts this week crossed the line for me. It all stems from headlines run on certain social sites reporting an infographic (shown below) generated by SodaHead.com that shows that users are growing weary of technology.
Here’s a sample of headlines from the blogosphere:
INFOGRAPHIC: Facebook’s Popularity Slipping?
Is Facebook’s Popularity Falling On Its Face? [Infographic]
Public growing weary of popular technology (infographic)
The problem? The data behind this isn’t randomized or diverse enough to draw such a broad conclusion. It’s a sample of 1300 users of SodaHead.com, the majority of respondents being over 25. Drawing from a homogenous group amplifies the sampling error of any data that is collected. It’s like drawing a conclusion about the Presidential race by polling only Californians – it’s a big group of people, but not at all representative of the viewpoints and interests of people in the other 49 states. It also should be noted that the polling was done post-IPO where user sentiment appears to be colored by their fluctuating stock performance.
In speaking with SodaHead founder Jason Feffer, they are aware that is inherent sampling error to their methods . So why would prominent bloggers treat this information to make sweeping generalizations? I suspect that Facebook + infographic + conflict is probably a good recipe for getting people’s attention. As Reuters demonstrated a couple of months ago, it’s important to approach these opinion polls with a healthy dose of skepticism. People oftentimes misrepresent their behaviors when asked. This is nothing new: John F. Kennedy won his Presidential election over Richard Nixon with a minuscule majority of the popular vote, yet a year after he took office 60% of people said they voted for him.
A more reliable indicator of user sentiment (measured by time spent on site) towards Facebook would be Comscore or Nielsen. Their sampling methods generally focus on tangible action and can reduce the sampling error (most times) to as low as 2%.
As a post-script – the original draft of this post misrepresented some aspects of SodaHead’s business, which is actually quite unique. I made some assumptions about the content on SodaHead based upon an initial glance into the site, and demographics informing their polls based upon traffic to their site. SodaHead actually places their more sensational polls on their landing page (which I understand, since I still get a fair amount of search traffic from my posts on Alyssa Milano and Kim Kardashian) and much of their polling is done off of the site, so drawing conclusions about their content from the landing page is inaccurate as are conclusions about poll respondents based upon site demographics. I’m really grateful to Jason Feffer for taking the time to reach out and explain how my assumptions about his company were wrong, and to introduce me to a company that has established itself by taking a great idea outside of the box. More about that in a future post….