If a user who hasn't bought a washing machine recently is 0.1% likely to be interested in a washing machine ad, and one who has is 0.2% likely to be interested, then they may be making the right call, even if for most people that second number is 0.0%. A few outliers can really skew those kinds of things.
Yeah, it might be a strange statistical artifact. Perhaps someone is buying washing machines for an entire dormitory and isn't satisfied with the first test model. Such people, while rare, might be valuable enough to push new washing machine ads on everyone.
That, people with multiple laundry rooms (somewhat-large McMansions with two or more floors may have these) who buy one to see if they like it, then buy another (same or a different one), people with enough money to buy washing machines for someone else (or "this one's OK but I'm open to buying another if it looks better, and giving this nearly-new one to my kid/parent/friend" or whatever), et c. With those kinds of low-frequency purchases it doesn't take a lot to make one category of buyer twice as likely to buy a second time as the population at large is to buy one in the first place—even if the notion seems silly to most people seeing the ads.
