Facebook should’ve kept its research quiet
By now you almost certainly know about Facebook’s little foray into experimental psychology. Working with two researchers from Cornell University, Facebook selected 689,000 users randomly from its billion-plus user base and then added an algorithm to each of these user’s Facebook pages that blocked either positive or negative content for seven days
Then they sat back, presumably stroking the figurative white cat in their lap, and monitored the posts that these unwitting subjects created in response to the demonstratively good or bad news they were experiencing. The results of the experiment, published in the March issue of the Proceedings of the National Academy of Sciences, appear to confirm that the relative positivity of the media that a person is exposed to influences the frequency and degree of optimism they then communicate to others.
But the real impact of the research has far exceeded these original published findings. There has been an outcry about the way Facebook brazenly manipulated so many of its users with no apparent concern for their general wellbeing. On Sunday the data scientist responsible for the experiment, Adam Kramer, finally apologised for the study. “Our goal was never to upset anyone”, he explained on Facebook. “In hindsight, the research benefits of the paper may not have justified all of this anxiety.”
By the letter of the law, Adam Kramer really had nothing to apologise for. Every one of the more than half a million research subjects he experimented on, like the rest of Facebook’s user base, agreed to the experimental use of their data when they started their Facebook page and digitally signed its Data Use Policy. Running an experiment on its users is also hardly something Facebook needs to feel red-faced about. Kramer is part of a gigantic team of data analysts at the company who regularly run experiments on unwitting users as they access the website and interact with others.
And Facebook is not unusual in its experimental zeal. Major supermarkets have long varied prices from store to store to isolate the ideal price and promotional activity to sell their wares. Supermarkets don’t necessarily need to know why a BOGOF on frozen peas works better than a 20% Extra Free offer, they just need to know that it’s true and execute accordingly.
Similarly, there is not a coupon book or catalogue in the UK that does not mail out multiple, alternate versions with differing offers, formats and even paper weights to examine which document “pulls” best. Ask your favourite charity about their direct mail experiments next time you want a nice shock about the prevalence and popularity of consumer experimentation.
Every major ecommerce player from Amazon to Zappos is equally culpable. You might only experience one webpage when you visit your favourite etailer but in reality hundreds of different variations are being offered at any one time. The variants help the company decide a plethora of issues. Should the page be blue or red? Ten or twelve point font? Top right corner? Experimentation and the ability it confers to test, learn, alter and improve explains much of the success of ecommerce in this country.
Even manufacturers, who typically find the business of experimentation a lot trickier to initiate than their retail and online peers, regularly get in on the game. For years Procter & Gamble pre-launched its British products in Carlisle to test the response of consumers to new products or new product offers prior to any national launch. Not one Cumbrian probably ever realised their soap powder preferences were being fed back to a central database at P&G’s Cincinnati HQ – exactly how the company likes it.
At Marketing Week, we too are no strangers to the benefits of experimentation. Each week we launch hundreds of variants of this column – some with swearing, some with deliberate factual errors – to isolate the perfect formula. Even the photo before you is the result of years of experimental testing. I am actually a 75 year old West Indian woman but our experiments confirm that this fat, white bloke in a suit works best on our target reader.
The key lesson from two decades of real-time consumer experimentation is to do it but keep the research quiet – something Facebook singularly failed to do last week. Given Facebook’s current image as a faceless manipulator of people’s data one might have imagined that of all the companies currently experimenting on users the social media company would have been especially careful with its approach.