Bad Research: Popular Sex Search Terms

People like sex. They like sex so much, they spend a lot of time searching for it online. Go figure. (You can tell I’m about to delve into really highbrow, heady stuff here…)

Researchers Ogi Ogas and Sai Gaddam recently published a book, A Billion Wicked Thoughts, detailing their analysis of 400 million searches they collected from the Dogpile search engine. Of those 400 million searches, 13 percent (55 million) were for erotic content.

How did those 55 million searches break down? Let’s find out… but let’s also look at the methodology of these researchers to see if their findings are worth the paper that they are printed on. (If you think not, you’re probably right.)

So here’s what people on Dogpile search for when it comes to sexual interests. Note that the terms below are the general category of search for that interest, which is inclusive of all sorts of permutations of the terms. These permutations (such as “tits” for breasts) are not listed below; use your imagination.

1. Youth – 13.5 percent
2. Gay – 4.7 percent
3. MILFs (Mother’s I’d Like to F***) – 4.3 percent
4. Breasts – 4.0 percent
5. Cheating wives – 3.4 percent
6. Vaginas – 2.8 percent
7. Penises – 2.4 percent

Garbage In, Garbage Out

There’s an old saying in computer programming — GIGO: Garbage In, Garbage Out. It applies equally well to any scientific endeavor, which is only as good as the data you choose to analyze. If you start out with a dataset of questionable generalizability or value, you may find yourself drawing conclusions which have little connection with reality.

In this instance, there’s a huge problem with the research data these researchers compiled. They don’t come from Google or even Bing. They come from a little-known search engine called “Dogpile” which isn’t even a search engine. What Dogpile is is simply an aggregation engine of search results from Google, Yahoo and Bing (since Bing now provides Yahoo with their search data, I’m not sure why there’s still this differentiation).

This is not the same as a search conducted on Google through Google.com, or a search conducted on Bing through Bing.com. You actually have to go to the Dogpile website to get these results — results that formed the dataset for the current researchers. If you do a search on Google.com, your search would not have been analyzed by these researchers (which makes sense, since Google and Bing don’t make the data it collects on searches readily available to researchers).

What probably makes Dogpile little-used is the fact that it intermingles sponsored search ads with the organic search results with virtually no visual cue it is doing so. Tiny print at the end of each search result lets you know whether it is a “sponsored” result or not — e.g., an ad. In a search for “depression” on Dogpile, 14 of the first 20 search results were ads — not exactly something most ordinary people would put up with for very long.

People looking for content online have long decided to abandon the use of search engines that try and intermingle advertising with actual results. The reason is simple — people will click on an advertisement when they’re interested in the product or service being offered. They don’t like to be tricked into clicking on what they thought was a search result, only to find out it was an ad in disguise.

So who uses Dogpile? Who knows, but it certainly isn’t likely to be a mainstream Internet user. While over 150 million people use Google and 90 million use Bing.com, Dogpile’s measly 2-3 million people per month pales in comparison and is far less than 0.05 percent of the total search engine market.

Can you conduct a survey on such a tiny dataset and try and use smoke and mirrors to make it seem like you actually did the same sort of ground-breaking research that the Kinsey Institute did in the 1950s and 1960s? You sure can.

For instance, Ogi Ogas and Sai Gaddam said they analyzed 400 million Internet searches. But compare this number to the 3 billion searches conducted each and every day, according to Hitwise, an online analytics company. Suddenly 400 million — while seemingly an impressive number in a vacuum — looks far less impressive when placed into some sort of data context. 400 million searches is the equivalent of what’s conducted in about 3 hours. In one day.

Context is, of course, everything when it comes to datasets, especially when those datasets are likely to be biased in ways you never bothered to investigate. In this instance, the dataset is biased by the use of the Dogpile search engine — a tiny, niche search engine that is more likely than not used by a certain subset of the population that differs from the rest of the population.

So take this list with a grain of salt. It’s interesting, but I’m not sure it’s reflective of the general population. And it’s certainly not worth buying a whole book that delves into this flawed dataset.