From The Digital Reader:
Earlier this week Amazon released a statement tangential to the ongoing problem of cheating in Kindle Unlimited, but as with many of Amazon’s statements the bigger story is what was left unsaid.
To recap, Amazon has a problem with people gaming the system in Kindle Unlimited. This issue has been around since the beginning, and as Amazon changed the rules to thwart the cheats, the cheaters changed their tactics in what is no more or less than a game of whack-a-mole.
In the early days people were uploading really short books, and were getting paid for a loan when the 10% threshold was reached (with a 10 to 20 page book, this didn’t take long). Amazon fixed that by switching to a system where they paid based on the number of pages read by subscribers, and in response the scammers invented the book-stuffing con and started uploading really long books.
. . . .
Amazon announced this week that the jump to the end trick no longer worked:
In response to concerns we’ve heard from authors, we wanted to take a moment to clarify in more detail how we measure pages read to calculate the monthly allocation of the KDP Select Global Fund.
We have worked steadily over time to improve the fidelity of the KENPC system that measures the number of pages read. For the vast majority of cases, KENPC v3.0 records actual pages read with a high degree of precision. For the few remaining cases, such as very old devices, we employ several processes and technologies (both manual and automated) to accurately measure pages read. In addition, we regularly audit the pages read of top titles.
Our commitment to the fair allocation of the KDP Select Global Fund remains a top priority. That includes addressing attempts to manipulate our services. If you have direct evidence of these types of activities, we will review every single example provided to us at content-review@amazon.com.
There were reports last fall that this had been fixed, but I have also heard from authors who disagreed, so it is good to finally get a statement from Amazon.
But even so, while Amazon may claim to count pages read with exactitude, authors argue that the way Amazon measures a page size is less an exact science than a wild guess or perhaps a randomly generated number.
Authors are saying over on KBoards that it is possible to make minor changes to an ebook’s formatting and end up with significantly different KENPC. One said he could format his ebook fiver different times and get five different KEPNCs.
. . . .
The thing is, folks, if one page in Book A is equal to 1.2 pages in Book B and 0.8 pages in Book C then it really doesn’t matter how carefully Amazon counts the pages.
Link to the rest at The Digital Reader
Aha! Finally I have had a direct experience in this area: changing fewer than 50 characters out of 167K words (I fixed a few tiny typos) led to a loss of 4 ‘pages’ in the KENPC for my novel. And about half of those changes were a character name – of the same number of letters.
Weird.
I even had to debate with myself whether to recheck the Createspace cover spine. I didn’t, and the proof copy looked no different than before, but that was odd.
Might have been where a little space made the difference of needing an extra line here and there.
As to the name change, same letters can still be different space as ‘eye’ and ‘ill’ if your font closes the gaps.
The stated goal of Amazon’s KENPC algorithm is to removal formatting from the page count calculations.
Sort-of.
KENPC uses a standard virtual page that it “renders” the text to. What typographical elements and rules it applies is unknown–margins, line spacing, font and size, indents, etc–but one would expect widow and orphan control and full justification to be part of the virtual rendering. The added spaces from justification and added lines from widow/orphan control, across tens of thousands of words could easily account for a variation of a couple of percentage points.
Some formatting of the text would be expected to have some minor effect on the virtual page count. Things like breaking up a paragraph, enabling enhanced typography, etc. All minor, though.
Now, if people are reporting more than a minor percentage then something else might be at play. Bugs in their virtual rendering engine would be the first suspicion. No code is ever bug-free and if enhanced typography and algorithmic hyphenation is part of the virtual rendering there is even more room for bugs and weirdness to creep in.
As the old saying goes, to err is human but to really mess things up you need a computer. 🙂
Interesting. All along I had thought Kindle was pegging the page counts to the print version of the books. I never thought of double checking. No wonder there’s so much uncertainty.
They used to do that to assign reference page numbers to some ebooks way back before KU. Still might but since most folks are used to “locations” it’s not the issue it once was.
Here’s a piece detailing the kind of variation different books get in “per page” word counts under KENPC 2.0.
http://rogerpacker.com/the-long-and-the-short-of-the-new-kindle-unlimited-kenpc-recount/
My suggestion is that Amazon adopt a modified Netflix rental model. That is, a tiered pricing system according to which you can “borrow” X number of books per month, with Y number of books out at a time, with an upper bound of Z number of pages (or total number of words). X and Y would mitigate against short page-count schemes, while Z would mitigate against long page-count schemes.
It also seems like they could lower the maximum number of pages. Currently, a book can be paid for a maximum of 3000 pages. How many books even come close to that many pages? It seems like 1000 pages would be a more reasonable maximum. And if someone has a longer body of work, they could split it into several volumes.
This would make it less worthwhile to put bundled books into KU, but maybe that would stop the most recent forms of scamming. And legitimate authors could just unbundle their books. It does not cost a reader any more to read individual books than a bundle of the same books, so why not block the scammers from using that approach?
How would a switch to a tiered model benefit consumers? The welfare of consumers will always trump the welfare of suppliers.
They should have left the payment method alone and just banned books under a certain size. At least cheaters would only have gotten about $1.30, instead of ten times as much now.
I had some shorts in, and I would have understood if that wasn’t allowed any longer. As it is, the scamming got worse, the payouts got bigger (hello, bonus money!), and innocent authors are still dumped on.
And some of us tried to warn Amazon about it all, but they don’t seem to care unless the press names and shames them. In the end, nothing will likely change.
What do you mean ‘under a certain size’? This isn’t KU1 anymore. The scammer books are the 1,000-page ones. Shorts aren’t the problem.
I think she is suggesting that they should have stayed with KU1, and just banned short books from KU altogether. But I am sure the scammers would find a way around that also.
The KU1 model is unsustainable. Oyster proved it.
Incremental payment is the only (even vaguely) long term viable model.
Quota-based systems might be short term viable but I doubt they will grow enough to prosper. Eventually they will fade.
KU 1 was an incremental payment model. The payment varied based on usage and the suze of the funding pool.
No, it wasn’t. It was all or nothing.
KU1 paid out in full after after 10% (just like Oyster) which is why they were flooded with short books and had to switch to per page payment to incentivize longer works.
The problem with KU is that it requires Authors to place an enormous amount of trust in Amazon. The only real protection for Authors seems to be the ability to enter and exit the program with relative ease. This protection did not exist in Audible’s romance package, though fortunately the last I heard Audible was allowing Author’s to exit the program even though not contractually obliged to do so.
KU,from my limited understanding, allows Amazon total discretion in setting the funding pool. Authors have no right to call for an Audit, and Amazon can tweak the system at will. Authors who don’t like it can choose to leave. This is why KU works whilst other unlimited type subscription models fail epically. Amazon has control over its expenditure. If you put your book in KU, you are trusting Amazon, pure and simple. If you’re not completely happy do your sums and pull your books out. But don’t forget that KU offers other advantages as well, and it may be better to be paid less for more “borrows” than more for fewer purchases. As an observer it seems to me that many authors use KU very effectively as a promotional tool, with some of their books in KU and others only available to purchase, but at a reasonable price.
Amazon does need to come to terms with scammers, and this is not an easy task. Amazon does need to do much better. Ideally Amazon’s algorithm for determining KENPC should be transparent. I’d imagine the argument for not making it so would be that it would assist the scammers, though I can’t really think how if the method is sound. For most books I think a word count would be perfectly equitable, with a set number of words deemed to constitute a page. There would obviously need to be some tweaking to take account of pictures, maps, illustrations etc. A picture would not of course be worth a thousand words, but may be worth a page.
Yes, those using KU are having to trust Amazon. If they don’t trust Amazon or don’t like the way Amazon is doing things they are free to leave KU and still sell on Amazon and elsewhere.
Most seem to be staying because despite scammers KU’s they still think it’s the best game in town. If they didn’t think so they’d leave.
We keep seeing/hearing whining about the scammers and how much they’re stealing from the KU pool. Does anyone have any ‘actual’ numbers we can see? May’s pool was 22.5 million, if scammers ‘only’ scammed away a million that’s less than 1%, most stores have to figure more than that for shoplifters. And like shoplifters Amazon has to figure out what can be done – without locking the store down to the point no one can shop there.
Before anyone says they need to ‘do more’, I’ll remind you that on these very pages we’ve seen what happens every time Amazon does more or tries something new; the ‘honest’ writers cry that Amazon is now hurting them for no reason.
It’s Amazon’s ball park, if you don’t like the rules or that the umpires are sometimes blind then grab up your ball, bat and glove and hit the other fields.
Oops, 5%, but still we have no real numbers to go by …
Last I checked, 1/22.5 is 4.44%.
And soon the new limits/changes will have ‘honest’ writers whining that Amazon is punishing them – and they’ve done nothing wrong.
The extra box of popping corn just came in – I’m ready! 😉
“And soon the new limits/changes will have ‘honest’ writers whining that Amazon is punishing them – and they’ve done nothing wrong.”
You keep saying this nonsense, Anonymous.
If we see and report a crime we do not expect to get arrested and locked up ourselves because the police are heavy-handed.
Why should we sit back and accept we are collateral damage just because Amazon prefers to use algorithms rather than real people to police its own rules?
The big scams here are happening in full view in the Top 100 and are being reported by David Gaughran and others Amazon knows well enough to take seriously.
If Amazon were serious about stopping these scams it would employ someone – maybe David himself – to clamp down properly.
But of course it is not Amazon losing money here, only the honest writers who play by the rules. Including the ‘honest’ writers Anonymous says are whining.
But of course it is not Amazon losing money here, only the honest writers who play by the rules. Including the ‘honest’ writers Anonymous says are whining.
How much is being lost?
Amazon decides what the payout is for KU. If all scammers disappeared, would the total KU payout go up, down, or stay the same? Would honest writers see their pages read go up, down, or stay the same?
Clearly if you flunked basic arithmetic at elementary school there is no point in trying to explain further. The answer is in the question.
Just carry on insulting honest writers and treating the scammers as entertainment.
Sorry, different anonymous there.
I had replied, breaking it down for you, but apparently PG thought I’d added too much snark for you (or maybe his other readers) to handle.
The bottom line is no matter how bad ‘you’ and others think the scamming situation is, it’s not bad enough for most writers or readers to pull out – therefore Amazon doesn’t see it as that much of an issue.
(Like that guy at the diner yelling that the food there is crap – yet he keeps going there to eat. 😉 )
I’ve never understood why they don’t just do wordcount.
Because then people would use a hundred little words where a couple big ones would have been better. 😉
And wordcount wouldn’t actually help anything if you think about it. The scammers would find the wordcount limit and make their books that many words – much as they’ve been doing page count. (Oh, and while Amazon can see what page you stopped at – how on earth would they know which word you stopped reading at? 😛 )
I was just wondering the same thing, but it occurs to me that books with pictures would be hard to quantify. How many words is a picture worth? My nonfiction titles have hundreds of illustrations.
“How many words is a picture worth?”
Classically, one thousand. This comes out to about three or four pages. This seems low if we are talking about custom-made illustrations. It seems high if we are talking about random images pulled off the web.
And then the scammers would add 5 or 6 small pictures to every page of their books….
Yup. There may be some clever fix around the various problems, but my dark suspicion is that the subscription model is inherently susceptible to scamming.
I think philosophically Amazon is going to have to decide that not all types of books belong in KU. I’d draw a line in and simply say “pictures don’t count.” Otherwise it would be ridiculously easy to load a file with a bunch of clear or 1 pixel images and game the system. Authors would simply have to decide if it makes sense for their type of book to be in KU based on their overall word count.