Skip to content

Some Semi-Random Observations

The observation-lemur observes.

I had not planned on releasing three very different books in the space of three weeks. In fact, one of them wasn’t supposed to be written at all! But the Evil Muse, and the success of what was supposed to be a set of stand-alone stories was such that I released a short urban-fantasy novel, the tenth Cat Among Dragons book, and the second Shikari book in three weeks.

So, what were the results? 1) A frazzled author.That’s because I also had a major concert I was performing in, plus some Day Job things that overwhelmed me to a degree.

2) The urban fantasy (Strangely Familiar) sold like wildfire. This is partly because Sarah graciously mentioned it on Instapundit, and partly because there seems to be a hungry market for urban fantasy without “whiny b-tches for main characters” as Dorothy Grant phrased it. There’s also no sex, although that may or may not be a selling point, depending on how the series is marketed. Humor and adventure sell in urban fantasy, as the success of the Alexi stories and now the Familiar stories shows. Take heed, all ye who write.

3) The Cat novel is moving very slowly. I expected this, because the series has been slowing down. The story turned dark, and a number of readers stopped enjoying the character and the series. Plus it is ten in the series, and lots of people will not want to begin with the end. And I didn’t advertise it, other than at my blog and at Sarah’s place.

4) The Shikari stories sell really well. Not as briskly as the Familiars book, but this is the second in a series, and adventure science fiction with younger protagonists (16 and 19, in this volume). Again, there’s a market for retro, Kipling-esque adventure stories out there for teen and adult readers. I have not advertised this one as much, but that’s not stopping it. Rory’s cover is also helping. *waves toward Australia*

5) According to the warning I got from the ‘Zon, I am perhaps selling too well and they think someone is committing fraud using my books. I will not get paid for those reads and buys. They did not say how many are fraudulent, so I won’t know for several months how badly this will hurt my payment. I sent a reply thanking them for the warning and pointing out that I do not advertise. This was more to establish a paper trail than to get anything done, because I am aware that Amazon is… less than instantly receptive to certain things. If I see a sudden spike in sales, I will tell Amazon, so I will be less likely to be blamed for the activity, now that I am aware of it.

6) I am not certain that having the three releases boosted sales. There may have been some spill-over, but since they are different genres, different series, and different shades of light, darkness, humor, and romance, I’m not certain they helped each other the way three Shikari books or three Cat books would have done. Several fans were thrilled at having three new books in as many weeks, but I can’t keep up that level of output. It wasn’t planned, I assure you.

7) There may be a fourth pure fantasy novel before Memorial Day, but if not, it will be in late June or mid-July. I expect it will move very slowly because of the summer doldrums.

Edited to add: Merchant and Magic will be available Tuesday, May 15, barring major problems.

18 Comments
  1. I genuflect at your output. One of these days I hope to rise to such levels of industriousness. I definitely enjoyed “Strangely Familiar” and can’t wait for more in that world.

    May 13, 2018
  2. Forrest #

    As one of your faithful readers, who bought both the new Shikari book and the new Familiars book, I am upset to hear that you may not be paid by Amazon. Please keep us informed, and let us know if there is anything we, the readers can do to help.

    May 13, 2018
    • I’ll find out in two months how many sales and reads were fraudulent. I’m hoping just a few, but I will let folks know in August what the difference between projected income and actual is.

      May 13, 2018
  3. Well *my* (Familiar) purchase was real, dagnabbit. Left a review, even (not under this name, mind.) And yes, humor sells, at least I find I’m more willing to go for it. Also, yes, the genre differing means different rates. The Familiars stuff is pretty much insta-buy or as close as finances permit for me. Cat-Amongst-Dragons.. is think on it some. Shikari? Well, more thought, really.

    May 13, 2018
    • RCPete #

      So were both of mine. Sample bits are gooood bait!

      May 13, 2018
  4. Fraud…Is this a KU issue?

    May 13, 2018
    • Yes. Through the page reads. I was trying to find one of the articles that explains how it works, but it’s crunch time at Day Job.

      May 13, 2018
  5. I used to work at Amazon, where I managed a team that wrote AI software (Actually we called it “Machine Learning and Natural Language software”; “AI” has a bad name with a lot of people, who think it means “cool-sounding stuff that doesn’t really work.”) I wasn’t on the fraud team, but I worked with them a lot, so I can take a few guesses as to what’s happening.

    I see all three books are available through Kindle Unlimited, so my guess is that that’s the only revenue they’re withholding. They’d only withhold revenue from actual sales if they suspected you didn’t really own the copyrights to the books, but if that were the case, they’d suppress them from the catalog too, and that’s clearly not the case. (I can see all three of them.)

    The way most production AI works (and, again, I don’t know what the fraud team is really using; I’m just basing this on what I’d do if it were my team) is that you collect information about confirmed valid and fraudulent events and then feed a dog’s breakfast of data about each one into an algorithm that trains a classifier. You then feed data about new, unknown, events into the classifier, and it takes a guess as to which are fraud and which are not. You then have your human staff put extra effort into the ones it thinks are fraud (but you still need to check at least some of the ones it thinks are valid). As you find more examples of fraud (and non-fraud) you periodically retrain the classifier, so it gets better over time. You don’t usually let it make final decisions; all it does it better direct the efforts of your staff.

    The key is what’s in that “dog’s breakfast.” (We call them “features,” in the sense that faces have features that humans use to tell them apart.) Obvious features are info about the author, info about the books, and info about the readers. Does the author only sell via KU? That is, does anyone buy his/her books outright? Does the author have any traditionally published books? Is the author a member of an association like SFWA? Has the author ever been investigated for fraud before? How many books a year does the author produce? How long are the books? Do they contain spelling errors? When readers “like” the books, what other books do they like? Are any of those books by authors suspected of fraud (or investigated for it)? How many books a year does each reader read? Do the readers ever buy “normal” books? You can probably think of a few others to add.

    The classifier will decide, based on the training data, which of these features are “predictive” and which are not. Most classifiers are smart enough to determine when features are only predictive in combination with others. And at the end it never says “yes” or “no”; it just gives a score.

    It’s harmless to have too many features, as long as there’s enough training data. The classifier will figure out that things like “how long is the author’s name” have no predictive value. Useless features just slow down the process.

    So in this case, I would guess that publishing three books in the same month is what pushed you over the edge. It’s also possible that people who commit real fraud have figured out that they need to boost the sales of a few authors who aren’t in on the scam, just to try to confuse the AI. So I’d also guess that some of your unexpectedly high sales really are fraud–even though you’re not to blame for it.

    Finally, just in case anyone misunderstood your article, when they finish investigating, you’ll get paid for all the legitimate reads and buys. They’ll only withhold payment on the ones they deem fraudulent. (I’m sure you already know this, but I didn’t want anyone thinking you wouldn’t get paid at all for all three books.)

    Sorry this happened to you. Hope it gets cleared up soon!

    May 13, 2018
    • Oh yes. I have no problem with not getting paid for fraudulent buys. I wouldn’t expect to get paid for them, honestly, and if I caught something going on, I would notify Amazon and send the money back (do NOT care to be an accessory to fraud or money laundering out of ignorance.)

      May 13, 2018
      • Draven #

        yes but them detecting a sudden surge of sales as fraudulent because, more than likely, Sarah instapundited you… is annoying

        May 13, 2018
    • Draven #

      WEll, I hope their ‘ai’ works better than what Youtube is using.

      May 13, 2018
  6. 23skidoo

    May 13, 2018
  7. Dammit, check the box!

    May 13, 2018
  8. ‘Kiplingesque’ – you have no idea how happy I am to see you use that description, because that’s the vibe I was getting while doing the cover, and periodically had to remind myself to show that this was also a science fiction setting! It is great to read that the book is selling well and the cover helps with that. ^o^

    May 14, 2018
    • TonyT #

      Kipling rocks…slowly reading Kim aloud to my kids

      May 14, 2018
  9. Congratulations!

    May 14, 2018
  10. elainethomp #

    I found Merchant & Magic on amazon yesterday and downloaded.

    May 15, 2018
    • Thank you!

      May 15, 2018

Comments are closed.

%d bloggers like this: