Ad targeting by menstrual cycle at Facebook
Facebook is testing timing ads to its female users’ menstrual cycles. The social network’s data scientists have uncovered that women are more likely to buy fashion items during ovulation. Facebook now offers select marketers an opportunity to show the ads during that period. Early tests show an impressive 23% increase in CTR and even stronger, 27% uplift in sales.
Though uncovered through multivariate regression analysis, the results probably won’t be surprising to evolutionary biologists. Indeed, during ovulation, the only period when a female can get pregnant, she would subconsciously try to be the most attractive, to ensure mating with the males carrying the best genes. Nature entices her to buy more expensive fashion and beauty items.
The roughly 30-day periodicity in CTR, conversion ratio and sales volume data was clearly visible. Tiding the bump to ovulation was trickier. Even trickier was to determine, for a given Facebook female user, when the ovulation takes place. The crucial data came from Max Levchin’s Glow pregnancy app which pinpoints ovulation via basal body temperature and cervical mucus test. Facebook and Glow data teams, in cooperation with the OBG researchers at Stanford University, set up a year-long opt-in study whereby 7,000 women allowed a tie-in between their Glow and Facebook apps.
With the exact day of the women’s menstrual cycle being determined by Glow, Facebook data scientists attempted correlating it with the user’s activity on Facebook, shopping-related or not. Thousands of features have been tested, of which a few dozen were found most relevant and were incorporated into the statistical model predicting in which day of her cycle the woman is.
Some of these features, e.g. more Likes on images of babies and kittens or spending more time perusing photos posted by their male friends, were rather intuitive. Also not surprisingly, posts by men who, in turn, have more female likes and followers happened to enter the formula with higher weight. Fewer smiles and more negative posts during the PMS days were not very surprising either. Other features such as higher attention to political news related to ISIS beheadings, wars and unemployment or spending more or less time playing Candy Crush looked completely illogical and unrelated, but demonstrated strong correlation with certain days of the cycle.
Eventually, a machine learning model based only on the woman’s online activity that can map the cycle with the astounding accuracy of 96% of Glow has been developed!
One of the first applications of this model was the propensity to shop during ovulation. More surprisingly, one can target ads aimed at men based on the phase of the menstrual cycles of their partners. Indeed, there is a roughly 12% uplift in CTR on ads promoting men’s fashion, motorcycles, fancy cars and similar items during the days when their female partners had their periods. (The beasts seem to be looking elsewhere…) Ads for beer, alcohol and X-rated movies worked better during the PMS days. Restaurant and theater bookings by men went up around their partners’ ovulation.
The applications of this truly breakthrough model go way beyond ad targeting. First and foremost, it is a “software-only Glow”. It can help millions of women to lower the odds of unwanted pregnancy or to conceive.
Dating is another obvious killer app. The previous statistical analyses at OK Cupid, Badoo and if(we) dating sites indicated that women are more receptive to pick-up solicitation during ovulation, and less so during PMS days. Knowing when to message the women to maximize his chances would be a boon to men.
And, surprise-surprise, there are indication that this is already happening. A London-based dating website Badoo is offering a premium service to its paying male customers promising exactly that: give us the name of your female interest and we will tell you when she would be ovulating or having her PMS, so you would or would not message her then, on Badoo, Tinder, OK Cupid or many other dating sites, as well as on Facebook. Moreover, the service is promising integration with a number of smart watches and Happn, the emerging dating service for physical locations. Happn allows users to see online dating profiles of other users around them, at the same physical location (a bar, nightclub or cafe). Badoo for Happn promises to send a message to the men’s smart watch with pick-up line ideas based on the woman’s online profile, and showing whom of the women around he has more chances with.
Badoo keeps mum on where they got the data. However, there is a strong suspicion that the Facebook-Glow-Stanford research code depository on Github was hacked and the formula stolen. Indeed, a notorious Ukrainian hacker Petr Kislodrischenko has been boasting on the underground Russian hackers’ forums that he had hacked the account, and he’s been offering its content for sale to the highest bidder. It sounds like Badoo has won the auction…