Why Facebook Debunked Princeton’s Study

And, How Their Debunking Was Also Flawed.


Facebook unfriended Princeton after a research study published on January 17, 2014. Facebook retaliated with humour in a “Debunking Princeton” post that was reported by Forbes contributor Pascal Gobry, however, Mr. Gobry did not carefully research why epidemiology was appropriate and valid as a modelling approach to viral marketing, instead calling it Bad Science. His reasons were 1) the appropriateness of the model and 2) his comparisons of valid proxies.

Facebook and Mr. Gobry’ used an “all or nothing” approach to try to disprove a hypothesis, “Madonna is dead because there have been less [sic] articles about her” and concluded with, “who knows? In reality, nobody.” For these two reasons, Mr. Gobry’s analysis fails despite being a lecturer at a pedigree school, HEC Paris.

Facebook’s analysts—Mike Develin, Lada Adamic and Sean Taylor, however, did not address the real issue of Facebook’s declining audience, and instead attacked Princeton—a red herring. In “Debunking” Facebook presented arguments using “Google Search for academic schools” as a proxy for student enrollment, is not valid, whereas Princeton’s use is valid and backed up by additional research.

Specifically, School enrollment is not a social or viral activity—not epidemiological in nature. There are limits on student enrollment and these institutions have been in existence for centuries. One cannot project future performance based on limited data without reviewing 200 years of historic data because the time slice is too small. So, Facebook’s analogy misses the mark.

Further, Facebook’s analysis of student enrollment to attack Princeton was not reliable “Unfortunately, in investigating this, we found a strong correlation between the undergraduate enrollment of an institution and its Google Trends index.” The graphic depicted a plot with a regression line, and an equation with R2=0.54537. This R2 means that about 54% of the error is accounted for by Facebook’s regression model and not statistically significant, so the model is not good to predicting enrollment (within a confidence interval). So, Facebook is wrong here, too. The subsequent analysis of Princeton enrollment (declining to zero by 2012) is then based on a series of incorrect conclusions.

Still, Facebook’s response was funny, and gained support of its users and loyal users: extreme arguments prove a point—that “socially negative behaviour” or information, if the Princeton study was taken this way, may a social consequence, just like a virus, can prematurely destroy another firm’s goodwill and momentum. Some will read the study and leave because of that news, while others will defend Facebook, like Forbes contributor Gobry.

As a neutral observer, there may be some truth and a moderating effect on the study and Forbes’ support of Facebook even if the defense analysis is wrong.

The term “viral marketing” was coined by Harvard Graduate Tim Draper and Jeff Rayport to describe Hotmail’s social distribution mechanism in 1997 and diffusion of information similar to epidemiology observations in the real world. And so the model is appropriate. Microsoft also used this term often in its strategy planning.

So, in undertaking further research, I point to new study by K. Sohn (Ph.D.), Gardner (PhD) and J. L. Weaver (PhD), “Viral Marketing—More than a Buzzword” from the Journal of Applied Business and Economics, vol. 14(1) 2013, which confirms “The epidemiological model offers a solid foundation and a number of good insights for understanding” virtual marketing (VM).

This report relies on substantial and significant studies in Viral Marketing such as “Viral Marketing in the Real World” Watts, DJ and Peretti. J (2007) published by Harvard Business Review, and “A Viral Branching Model for Predicting the spread of Electronic Word of Mouth,” Marketing Science, Vol. 29. No.2.

Looking over time at Facebook’s dispersion over time graphically, using Google’s search mapping tool, is like watching a communicable “social virus” spread. You should try this yourself using the Time Ranges and Search Terms features concurrently, and click the “View change over time” button to see what happens.

Facebook concludes by ridiculing the Princeton study through their own flawed Google trends analysis, “Google Trends for "air" have also been declining steadily, and our projections show that by the year 2060 there will be no air left.”

Again, search for air is not a good proxy for the actual oxygen in the atmosphere because air is not a contagion or a socially or virally transmission activity. They extrapolated data from 2006-2013 and the planet has been in existence for 4.5 billion years, and oxygen for at least 2 billion years ago. (Actually, Hypoxiaconglomerates is just not an attractive idea to me.What Facebook must do to keep its audience from further declines?

Teens Migrate to Twitter and Instagram, and Teens Don’t Want To Hang Out With Parents (15-19) as reported in Huffington Post. Where is this happening? The US? Latin America? Europe? Africa or Asia?

Facebook needs to “immunize” the rest of its base against defection, and to maintain growth. Zuckerberg might be correct in a monopoly US teens: the fallout may be elsewhere. He may be wrong. He’ll have to look at the data.

Once the cool kids leave, their snooping parents might follow to make sure there are no bullies or sex offenders on the prowl. Followers may follow, too. Difficult to predict unless you have a reliable model or insights: Facebook should look at the epidemiological model as a tool and plan to immunize itself: like develop a flank brand or platform strategy and re-architect its brand experience. After all, it touts deep analytic insights to its advertisers and have invested heavily in this technology.

What Should Start-ups Do?

In conclusion, opportunity knocks. Social Media Platforms that compete with Facebook act aggressively by developing switch behaviour applications to lure developers and consumers away—Pinterest, Snapchat, Myspace.com, AOL Music, KakaoTalk, Tumblr, Qzone (China), and V Kontakte (Russia).

Start-ups unsure of where this is going should place make hedges (develop for both platforms and see who wins). This was a chapter from Adobe’s playbook in the Apple-Microsoft war in the ‘90s. Writing for Facebook apps and widgets is always a good option, for now.

Start-ups who are loyal to Facebook, may want to defend their strategic partner, (albeit for additional leverage, a better deal or bragging rights).

Some start-ups will be opportunistic and try to capture dissatisfied Facebook users immediately if they compete. Twitter can acquire start-ups like Twitmusic, and other applications to integrate into its platform. Aligning with Pinterest for female teens is a good solution.

For larger firms like Zynga, who has distanced itself, they are in a better position to renegotiate a better deal after 2015 when their agreement ends with Facebook. If you’re a start-up, what’s your best strategic and action plan?

It just depends.


Al Leong is formerly a writer for TechVibes. He earned an MBA from the University of Toronto in Global Management, Strategy Consulting and Innovation. And, a certificate in Product Management from MIT/Sloan. If you found this article useful, go ahead and “recommend” or share it. And, feel free to view some other articles in this collection.


"Face-to-Face with Mark Zuckerberg '02". Phillips Exeter Academy. Phillips Exeter Academy. 24 January 2007. Retrieved 26 March 2013.



Al Leong, Blockchain Web3 CMO/CEO, Board Advisor
I Love Charts

Award-winning blockchain CMO, Advisor, Projects: Adobe, Apple, IBM, Microsoft, Intel, Disney, Sony, Siemens, Microsoft, IDC, Microsoft. #DeFi #Web3 #L1 L2s #DEX