AdSense Literary Experiment

BACKGROUND

For just over a year, we’ve been using a secret Gmail account to read and track submissions. It’s pretty simple: People email me submissions, I forward them to Gmail, and all the editors have the infinite storage and ultimate searchability at their fingertips. We can then use the labels to keep track of submissions as they move along the editorial food chain.

Google makes $4.2 billion a year, and a lot of that revenue comes through AdSense. If you ever Google anything — and I’m sure you do — you’ll notice the small classified-like ads in a column to the right. Those ads are targeted to the words you search for, so in theory, the ads are relevant to something you’re interested in. Advertisers get to pick their own target words, and pay a few cents per click. Rattle is actually one of those advertisers, though I’m still in the trial stage of deciding whether or not it makes “sense” for us.

These little ads appear next to all messages read in Gmail, and so every single submission we read at Rattle comes with its own set of targeted advertising (though I seriously doubt either of us have ever clicked). Furthermore, since most of the text in a poetry submission is the poems themselves, you might say the ads are targeted to the poetry.

HYPOTHESIS

I’ve always wondered if we could save some time and read the 20-word Google ads instead of the much longer poems. (Hypothetically, of course — we’re always going to read every word of every submission we get!.) But as an editor, I can’t help but wonder if the Google ad bots might be decent poetry editors themselves. Glancing at the ads every once in awhile, I actually think they can.

EXPERIMENT

The largest Rattle Gmail label is “Rejected.” I like to say the submissions are “returned”, which sounds a lot better in the spin room, but whatever you call it, that’s the label you don’t want. I won’t say how many submissions have landed there, but it’s thousands. The best you can hope for as a submitter is “Printed”, which means that after Megan and I have each read the submission, we both felt like it was worth printing out in hardcopy and talking about amongst ourselves and with Alan. It’s less than 5% of email submissions that wind up with the “Printed” label.

I’m going to collect a random sample of 5 submissions from each of these groups, and record the top Google Ad that appears next to it. If we can see a pattern emerge to distinguish between the two groups, we’ll know AdSense is on to something. The test is whether or not you can place a randomly selected submission, based solely on the ad.

At first I’ll make this a double-blind experiment for those following at home, and keep the labels a secret.

DATA

GROUP A:

Handcrafted Garage Doors / Custom finishes and accessories / Free quote. Guaranteed quality.

Eco Structural Technology / Green PreFab Steel Systems LEED / Cert- Commercial Residential

M+B Gallery / Contemporary Art Gallery / Specialized in Photography

Cheap Cowboy Boots / Cowboy Boots Sale Top Brands / Up to 50% off Justin Ropers

Teach Children thru Books / Children’s books that teach life lessons / Teasing,Peer Pressure,etc.

GROUP B:

Publish your book / at a price you can afford from $335, / + editing & marketing

John Wayne Ringtones / Ringtones, Screensavers, Wallpapers / in 30 Seconds to Your Cell Phone.

Book Printing in Asia / Self-publish high-quality fine art, / photography and large format books.

Publish Your Poetry Free / Get Your Poetry Published Online. / Join Helium’s Poetry Community Now!

Publish Your Poetry Free / Get Your Poetry Published Online. / Join Helium’s Poetry Community Now!

[Not a misprint, the last two of the randomly selected ads really were the same.]

ISOLATED X:

Reading Worksheets / Get Printable Daily Lesson Plans. / Find Free Worksheets & Resources.

ANALYSIS

Do you have a reasonable guess as to which group ad X belongs in? If you do, the hypothesis is confirmed.

It might be obvious already that Group A was submissions from the “Printed” label. Group B was “Rejected.” Ad X is another “Rejected.”

“Printed” ads seem to be more specific and unique, from prefab steel to cowboy boots. The only ad that might be considered bland and mass-marketed is the one for teaching books, but even that includes the specificity of children.

“Rejected” ads seem much more generic, and even lean a bit toward scams and spam. The only interesting subject in the group is “John Wayne”, but even that is a pop-icon, trying to sell a ringtone. It’s also the only one of the five that doesn’t contain the word “publish.”

CONCLUSION

Why do the ads for the better poems look so much more worth looking at? Why do the lesser poems generate bland ads? Even though this experiment confirms my assumption going in, I’m still surprised at the clarity of the results — five plot-points is such a small sample to be discerning a recognizable trend. Could the Google ad bots really have developed AI and an ear for poetry?

I don’t think so. The truth is, good poems have something to say — they’re about something interesting. The better poems contain a more unusual and precise vocabulary, and so they trigger the targets of advertisers who are more unique. The lesser poems tend to speak in generalities, and so there’s no linguistic content unusual enough to overcome the vocabulary of their standardized cover letters. Almost everyone mentions where they’ve been “published” or thanks us for considering their poems for “publication” — but with the weaker submissions, that’s all the little robots have to pick up on.

It’s going to be a long time before a robot takes my job — and you could always just Google Bomb your submission like a true spammer — but this little experiment should be an important lesson in one of the basic components of memorable poetry. Substance is key.

One thought on “AdSense Literary Experiment

  1. This is freaking hilarious (and, like most humor, there’s more than a kernel of truth in it). When I was working on my manuscript last week, I took some time to jot down keywords on the poems — sort of a “tagging” scheme that helped me see how they all fit together. Maybe next time through, I should gmail them to myself and sort them into sections by which Google ads they get. 😉

Leave a Reply to Anne Cancel reply