After I published the Dark Side of Inkitt, I started seeing more and more writers respond to and relay their own sour experience of Inkitt to me. Fellow writer/blogger Enchoseon wrote an article unambiguously titled The Grand Inkitt Scam. Enchoseon goes in on the people behind Inkitt. If nothing else, you guys should check it out just to see my appearance via the craziest of Photoshops.
Another writer, Michael Ampersant, contacted me directly after reading my post because of his experience with both Inkitt and coding to express his observations into what seems to be their algorithm, this legendary AI they tout as being able to predict bestsellers. It was a very enlightening conversation and you can view his observations to me and my followup response in their entirety on his website.
While I recommend you read both these writers’ posts, I’ll summarize and conjecture a sort of group hypothesis: Their algorithm, their AI, it’s magnificently craptastic. Let me run through the reasons.
- It was written by programmers who appear to have had no initial or very limited data starting out. This is a fairly reasonable assumption, seeing as the CEO is a programmer with an eye on publishing rather than, say a publisher with a penchant for programming. Like any programmer who believes in the power of machine learning, I imagine the programmers started the algorithm with their own hunches and assumptions of what readership for a bestseller would look like and let the AI learn as it’s relatively empty database began to fill. Ultimately it feels as if their hunches were derived with little to no knowledge of the publishing industry or how readers actually consume books.
- Inkitt relied on faulty means to gather more data. Empty database remember? To fill it, they held quasi-contests and promos and promises of pie in the sky to get writers to bring readers to their website. Once on the website, their database would start to fill based on reader behavior and Bango! Predictive Analytics Achievement award. It seems they didn’t account for one detail: Would readers across all spectrums use their website? It seems readers by and large don’t like to read via website… they’ve bought Kindles and Nooks and iPads just so they wouldn’t have to read books via website. Plus Inkitt’s got to rely on writers to drive readership, and writers aren’t exactly known for marketing ability just like programmers aren’t known for their publishing chops. The result seems to be a lack of participation in some genres and a wealth of participation in others.
- And now you’ve got Garbage In, Garbage Out. If you have malformed hunches and lopsided participation, your AI is going to learn bad. If you go to Inkitt’s “bestseller” page, you’ll notice that the overwhelming majority of their books are romance novels. Inkitt and their AI is predicting the next great bestseller is a romance novel. I’m not trying to crap on romance novels, I’m just saying it’s not taking into account the next Harry Potter or The DaVinci Code or any other novel not a romance. I wager that the only demographic that isn’t opposed to reading on the website are romance readers.
The Big Question: Are Inkitt Books Even Bestsellers?
They all have a nice yellow ribbon on the Inkitt website that labels them bestsellers, but by who’s definition? Certainly not the New York Times. It looks like they’re touting whatever top place they’ve held in Amazon’s ranks to claim this bestseller status. Appropriately named for this post, let’s look at Fake by Haley Ladawn.
As of the time of this post, this book Fake was at 819,150 in Amazon’s Best Sellers Rank, meaning Amazon considers 819,149 other novels to be selling better. These numbers are transient and temporary, fluctuating as sales pick up or decrease. My novel’s been in the Top 100 for it’s genre and I wouldn’t go anywhere close to calling it a bestseller like I’m giving up the day job. If you want to know how much of a difference one purchase can make, then go to any novel you’ve been looking to purchase,that one that maybe you’ve been putting off… if nothing comes to mind than I highly recommend this ultra-clever, adrenaline fueled fantasy adventure novel. Highly recommend. Anyway, check out the Amazon Best Sellers Rank before you buy. Then buy. Then check the number again. Yes. One sale makes that much difference.
Don’t get me wrong, more than total number of sold factors into that Amazon bestsellers rank. Some other factors is how many you’ve sold that day and how long you continue to sell over time. It only takes a few days to slide back a hundred thousand or more… if you haven’t sold many over time.
If you’re still on Amazon and wanting to experiment, type in LitRPG into your search bar. Pick ANY… I’m not kidding. ANY of these novels with LitRPG in their subtitle and you’ll see better sales ranks overwhelmingly to what Inkitt’s doing. Some of these books are also several years old. This means they’ve sold well and continue to sell well. The best even, like you know, a bestseller. The GENRE is wrecking the sales of others on Amazon, to include romance, and to especially include Inkitt’s romance line.
So the next best question is, if Inkitt’s awesome AI can predict the next bestseller, how come none of their books are even in this runaway genre?