Customer Psychometrics

“Psychometrics, sometimes also known as psychography, is a scientific attempt to ‘measure’ the personality of a person. The so-called Ocean Method has become the standard approach. Two psychologists were able to demonstrate in the 1980s that the character profile of a person can be measured and expressed in five dimensions, the Big Five: Openness (how open are you to new experiences?), Conscientiousness (how much of a perfectionist are you?), Extroversion (how sociable are you?), Agreeableness (how considerate and cooperative are you?), and Neuroticism (how sensitive/vulnerable are you?). With these five dimensions (O.C.E.A.N.), you can determine fairly precisely what kind of person you are dealing with—her needs and fears as well as how she will generally behave. For a long time, however, the problem was data collection, because to produce such a character profile meant asking subjects to fill out a complicated survey asking quite personal questions. Then came the internet. And Facebook. And Kosinski.”

In 2008, With a fellow Cambridge student, Kosinski created a small app for Facebook called MyPersonality that asked users a handful of questions from the Ocean survey and they would receive a rating, or a “Personality Profile” consisting of traits defined by the Ocean method.193 The researchers, in turn, got the users’ personal data, which soon amounted to millions and millions of reviews.193 “It was, literally, the then-largest psychological data set ever produced.”193

In the ensuing years, Kosinski and his colleagues continued the research; “first surveys are distributed to test subjects—this is the online quiz. From the subjects’ responses, their personal Ocean traits are calculated. Then Kosinski’s team would compile every other possible online data point of a test subject—what they’ve liked, shared, or posted on Facebook; gender, age, and location.”193 Once the researchers dug into the data, they discovered that amazingly reliable conclusions could be drawn about a person by observing their online behavior.193 For example, “men who ‘like’ the cosmetics brand MAC are, to a high degree of probability, gay,” which isn’t that surprising, but there are other interesting findings, such as one of the best indicators of heterosexuality is liking Wu-Tang Clan.193 Also, people who follow Lady Gaga are most probably extroverts, while someone who likes philosophy is probably an introvert.193

In the ensuing years, Kosinski and his team continued, tirelessly refining their models. “In 2012, Kosinski demonstrated that from a mere 68 Facebook likes, a lot about a user could be reliably predicted: skin color (95% certainty), sexual orientation (88% certainty), Democrat or Republican (85%),” explains Grassegger and Krogerus.193 Level of intellect, religious affiliation, alcohol-, cigarette-, and drug use could all be calculated as well, something that a casino company might find quite interesting as there are strong correlations between alcohol and/or drug abuse and problem gamblers.193 Employee Facebook pages could be scanned to screen out problem candidates as well.

As Kosinski continued refining his model, he discovered that with a mere ten “likes” as input, his model could appraise a person’s character better than an average coworker.193 With seventy, “it could ‘know’ a subject better than a friend; with 150 likes, better than their parents. With 300 likes, Kosinski’s machine could predict a subject’s behavior better than their partner. With even more likes it could exceed what a person thinks they know about themselves,”193 which is a pretty frightening thought.

The day Kosinski published his findings, he received two phone calls, both from Facebook; one a threat to sue, the other a job offer.193

Since the publication of Kosinski’s article, Facebook has introduced a differentiation between public and private posts so the data isn’t as easily accessible now.193 In “private” mode, “only one’s own friends can see what one likes. This is still no obstacle for data-collectors: while Kosinski always requests the consent of the Facebook users he tests, many online quizzes these days demand access to private information as a precondition to taking a personality test.”193

Kosinski and his team are now adding variables beyond Facebook “Likes”.193 Offline activity is now traceable and “motion sensors can show, for example, how fast we are moving a smartphone around or how far we are traveling (correlates with emotional instability).”193

Flipping this idea on its head, Kosinski speculated his research could become a search engine for people.193 By using all of this data, psychological profiles could not only be constructed, but they could also be sought and found.193 For example, if a company, or a politician, wants to find worried fathers, or angry introverts, or undecided Democrats, these profiles could be uncovered in the data.193

To Kosinski’s chagrin, one company he had been partnered with—Cambridge Analytica—was involved with Donald Trump’s 2016 presidential election.193 Cambridge Analytica bought up extensive personal data on American voters: “What car you drive, what products you purchase in shops, what magazines you read, what clubs you belong to.”193 Voter and medical records were purchased as well.193

In America, detailed personal consumer data is available for a price and Cambridge Analytica snapped it up and the company crosschecked these data sets with Republican Party voter rolls and online data such as Facebook likes. 193 Ocean personality profiles were built from this data and, from a selection of digital signatures there suddenly emerged real individual people with real fears, needs, and interests—and home addresses.193 Today, Cambridge Analytics has assembled psychograms for all adult US citizens, 220 million people, and they have used this data to influence electoral outcomes, as was seen with the 2016 U.S. Presidential election.193

“Trump’s conspicuous contradictions and his oft-criticized habit of staking out multiple positions on a single issue result in a gigantic number of resulting messaging options that creates a huge advantage for a firm like Cambridge Analytica: for every voter, a different message,” explains Grassegger and Krogerus.193

Mathematician Cathy O’Neil notes that Trump is like a machine learning algorithm that adjusts to public reactions.193 On the day of the third 2016 presidential debate, “Trump’s team blasted out 175,000 distinct variations on his arguments, mostly via Facebook,”193 an astounding number of unique ads. “The messages varied mostly in their microscopic details, in order to communicate optimally with their recipients: different titles, colors, subtitles, with different images or videos” were utilized, explains Grassegger and Krogerus.193

Small towns, city districts, apartment buildings, and even individual people could be targeted, explains Grassegger and Krogerus.193 Blanket advertising—the idea that a hundred million people will be sent the same piece of marketing collateral, the same television advert, the same digital advert—is no more, note Grassegger and Krogerus.193 Micro and personalization targeting has reached the point where companies can advertise to a market of one.

Cambridge Analytica separated the entire US population into 32 different personality types, and focused their efforts on only seventeen states.193 “Just as Kosinski had determined that men who like MAC cosmetics on Facebook are probably gay, Cambridge Analytica found that a predilection for American-produced cars is the best predictor of a possible Trump voter.”193 Among other things, this kind of information helped the Trump campaign focus in on what messages to use, and where to use them, perhaps even what channel to use them on.193 In effect, the candidate himself became an implementation instrument of the model.193

As Grassegger and Krogerus note, the first results seen by Das Magazin were amazing: psychological targeting increased the clickthru rate on Facebook ads by more than sixty percent. And the so-called conversion rate (the term for how likely a person is to act upon a personally-tailored ad, i.e., whether they buy a product or, yes, go vote) increases by a staggering 1,400 percent.”193

Now, what does all of this mean for an IR? How can they use “Likes” to gain a deeper understanding of their patrons? Well, potentially, by analyzing Facebook Likes, an IR could predict how open, conscientious, outgoing and neurotic an individual user and/or patron was. It could be as simple as doing a Facebook graph search of “Pictures liked“ or “Videos liked” and/or “Stories Liked” with the patron’s name. In addition to predicting a user's personality, these tests could estimate a user/patron's age, relationship status, intelligence level, life satisfaction, political and religious beliefs, and education.

An IR’s HR department would also find these personality test results interesting as matching a candidate with jobs based on their personality might make more sense than the current scattershot approach HR often takes in hiring. These personality tests could also reveal troubling traits that should not be ignored.


[i] Grassegger, H., Krogerus, M. December 3, 2016. I Just Showed That The Bomb Was There. Das Magazin

© 2017-2018 Intelligencia Limited. All Rights Reserved.



Rua da Esrela, No. 8, Macau

Macau: +853 6616 1033



505 Hennessy Road, #613, Causeway Bay

HK: +852 5196 1277