Fake Amazon Tech Book Reviews Are Hurting My Book
blogs.perl.org
blogs.perl.org
What's bad (really bad) here, is that not only are the books that are being ranked highly absolute crap, but they are crap in different categories. I.E. A "book" on swift, that was apparently "written", in a few hours, shoved onto Amazon, falsely ranked up - now ranks ahead of a 700 page opus on Perl, in the Perl category.
That's a total fail on Amazon's part, and really, really disappointing.
When it comes to reviews, the correct answer is that every user should have their tastes profiled and compared against the tastes of other users, with only users whose tastes in other areas counting strongly. I could easily imagine that this sort of thing might be difficult for Amazon to implement, though, depending on their infrastructure.
The bogus reviews had one or more of these aspects:
- bad reviews from reviewers with no other reviews
- mentions of the same competing book in my bad reviews
- 5 star ratings of that mentioned competing book (and no other reviews beyond those)
- 1-star reviews of another random book. This one was interesting, there was another book (Service Oriented Architecture for Dummies) that is entirely unrelated. But 2 of the reviewers also gave shitty reviews to that book as well as mine. There's no logical reason that the same people badmouthing my book would have also read, reviewed, and badmouthed that other book (totally different topic) within a few days. So they were clearly hired for a number of books at once.
So yeah, you'd think some of those patterns would be easy to automatically detect. Although I assume it's constantly a game of cat and mouse.
One thing I do like about their system is that you can click a link to read the reviews they consider "not recommended."
Here's what they say about it publicly: http://officialblog.yelp.com/2013/11/yelp-recommended-review...
This may not be the best place to bash Yelp, but if you're reading this, please don't pay Yelp for their exorbitantly-priced business service. It is absolute garbage.
And frankly, I don't have seem to have a lot of tastes in common with the hardcore Yelp users.
Maybe I was too generous by saying their system works "decently well"; though I don't think I'd say "complete crap"--unless my business had been burned by fake bad reviews, which it sounds like happened to you.
Perhaps this illustrates how subtle/difficult this problem is.
I.e. improving your resume, looking prestigious to potential consulting clients, or landing paid speaking gigs.
The way it works is simple. Publishers hope to sell about 5,000 or so copies of your typical tech book. That's enough to keep the publisher going, but not enough to really make money for them. The publisher keeps those "filler" books going to keep up street cred and hopes to sometimes land the "killer" book that makes them a nice profit.
I can't share the sales info for my book, but let's just say that I think the book is doing better than one might expect for its category.
If you want to write a tech book, you want to go with O'Reillly. My advance was much larger than what O'Reilly would offer, but they sell more tech books in the long run. However, O'Reilly's Perl book section is saturated and they generally don't take new authors.
What I did to reposition my book was a strategy I wish more tech books would take: I focused on jobs. Instead of teaching every esoteric niche of a language (though I covered it well), I focused very heavily on skills that employers look for. I've been doing this for a long, long time and am very familiar with employer needs, so I was well-positioned to do this. That has led to my book doing better than expected: here's what you need for a job!
I now have a successful Perl consulting firm (http://www.allaroundtheworld.fr/) and that book, while I can't prove it helped, definitely didn't hurt.
Oh, and my publisher has spoken to an Amazon rep and a contact I have in Amazon tells me that they're aware of the situation and are looking into it. I don't have much hope, but who knows?
From sponsored "impartial" reviews to niche items with way too many reviews and a mismatched helpful/not helpful or Q&A area, Amazon reviews are starting to lose their value.
I haven't really thought this through, so I'm open for arguments here.
There are a few legitimate reviews that indicate this book doesn't really have much content, and then 2+ dozen fake 5-star reviews praising the book highly and saying how much better it is than Munroe's book. If you look at the review history for any of the 5-star reviewers, they all seem to be reviewing the same products, and often they are reviewing 3-5 books per day.
I reported the reviews via feedback, but never received a response. Ultimately it's just disappointing to see this kind of thing happening.
https://www.amazon.com/gp/help/customer/contact-us
#1: Ignore this section, skip to #2.
#2: From the dropdowns, select: "More order issues", "Give Amazon feedback", "Other feedback topic". In the text box, enter "Catalog feedback".
#3: Click "Chat".
It took me around five minutes to explain that several of the books in the top 100 were incorrectly listed in the Perl category, and they happily accepted the feedback.
Including a direct link to the Perl category helps them tremendously:
http://www.amazon.com/gp/bestsellers/books/6134005011
And I also included a link to the post above.
Edit: Tell them specifically which books, in which top 100 position, are in the wrong category, so that they know which ones aren't Perl.
Most fake reviews are easy to spot. The simple test:
Could this review have been written about any other book in this category without changing a word?
I see fake reviews all the time for Kindle books written by indie authors. A mark of quality in a book is when there appear to be no 5 star fake reviews, but several written by real people, even if there are only 3 or 4 reviews.
Some fake reviews are harder to spot, though. For example, I suspect this account, which has been around for many years, is more than one person (perhaps a PR firm) hand writing unique reviews for each one:
http://www.amazon.com/gp/pdp/profile/A130YN8T37O833
All reviews are either 1 star or 5 star. There is content specific information but all stuff you can get from tech specs or descriptions - no sense the person actually used the product except generic intro paragraphs.
I think there is also fake voting on really good reviews. The review that I felt was the best review I ever wrote was downvoted more than any other review I've written. I don't really know if the downvoting was fake - maybe it wasn't helpful because I delved into too much technical detail:
http://www.amazon.com/review/R14QK0B7HRE5L8
However, most of the glowingly favorable reviews for this book have unanimous thumbs up, and the first 3 are written by "A Customer" which I'm guessing means the account of the reviewer has been terminated.
Unfortunately it would also reveal to customers that there are fake reviews, causing them to lose confidence in and lessen the value of reviews. They may then read reviews and subconsciously be evaluating if the comment is fake, not if the product is a good buy.
If you hid the "Irrelevant Review" button under a dropdown it might work better, since it wouldn't be there as a constant reminder.
* one or more incorrect facts
* goes on and on about one minor point while ignoring major stuff
* too brief
* reveals spoiler without a warning
Etc.
It's unclear if a crowdsourced flag would tip the balance of helpful vs spam in any useful way for categories where there's already a unhelpful level of spam.
I think you were probably being down voted for being overly technical though yes.
Personally, I'd prefer to have only reviews backed up with an explanation. If a book is good, tell me why. If it isn't good, I'd like to know why you think it isn't. That's the only way to make the next edition better.
I too hate reviews like that, but I think it's due to Amazon nagging customers to leave reviews for things that are purchased on the Kindle. I am beginning to suspect they have an app or form somewhere where you tap the number of stars and write your review in a text block.
In short: fake reviews and fake review upvotes very much exist and are WAY more rampant than you'd ever believe. It's an 80/20 or maybe 90/10 problem - most of the review content on Amazon comes from a small percentage of active users, and as soon as you become active on the site you start to get sucked into this sort of thing.
While not really related, I seem to run across a lot of books on Amazon that have a high number of 5 star reviews that, upon reading, are just bad. Seems to happen a lot with books by independent sci-fi authors with Kindle-only books...
Of course, "bad" is just my opinion.
Seems to be this JS file's fault: http://blogs.perl.org/users/ovid/mt.js
It's fairly easy to ignore a bogus review when you look at it, but the star system makes it too easy to take them at face value.
I've come to trust those items with all 4-5 star reviews far less than I used to.
That they don't have an efficient algorithm for this sounds more like they don't really care and never bothered with it.
Amazon has hundreds of thousands of products with tens of millions of reviews... correlating that with log history for each review will take a lot of time... not just to run, but to write any automated process, and work through resolving it.
It seems to me that Amazon seems to be pretty responsive when bot reviews are pointed out, and that may be, or at least have been a more effective strategy... But looking at an article a few days ago regarding twitter bot nets, and even seeing them try to draw me in... it's a very large problem all around.
Bad people will do bad things... as will misguided people. The bigger issue is the false positives... we've all read the horror stories of when a legitimate domain gets screwed by (insert popular domain registrar here) because of incorrect reportiong/reaction... or when a business' google apps is offline, and nobody can be reached at google... it happens.
In the case mentioned in TFA... it's probably prudent to ban the publisher in question. In others, the case may well be different.
It is a complex problem - but amazon has a serious advantage over other sites that have to deal with such issues (i.e. Twitter) - in that they have significantly more information on each user. I don't think Amazon is short on computing power either.
Taking into account order and browsing history, product review trends, linguistic similarities in review posts etc. They should be able to get very low error rates in identification.
Further unlike something like twitter feeds, it's quite possibly to silently de-prioritize abusive reviewers and associated products. Really, I'm quite surprised at how bad of a job they are doing - most of these cases are so blatant and obvious they should not require an author and a live representative to resolve.
We can see this happening in the case of SEO with many white hat sites employing black/gray hat techniques simply to maintain their current positions in Google.
It's not reasonable to expect Amazon (or anyone, really) to detect fraudulent behavior with perfect accuracy. I have to agree with the OP, though, in that they should be able to do a lot better. Many fake reviews are blatantly fake, and could easily be flagged by a relatively naive set of heuristics.
Yes, that's true, but I think this is a question of scale. Amazon is the world's largest e-commerce site (well, I think they're #2 after Alibaba now) and I've worked on some of the world's largest sites. Even "trivial" solutions become much harder at that scale.
I've gotten this on the iTunes app store quite a bit. Low ratings from people who don't review any other products (or if they do, it looks super spammy).
On top of that, the bad reviews tend to come in bunches (multiple in a day), are void of any actual useful feedback or complaints and are just vague & angry (sucks, ripoff, etc).
I've had decent luck with having Apple remove any that are WAY off topic, but it still sucks.
Google Play allows you to reply to reviews (which helps quite a bit as far as either helping the customer or correcting a misconception). I'm not sure what Amazon can change on that front.
Any type of system to respond would be useless for this particular problem.
I really can't say as much as I would like (there's stuff I can't share), but my publisher had a face-to-face with an Amazon rep and internal action was taken. Amazon's investigation is apparently over. The internal position seems to be "we're making money, there are words on pages, so there's no problem here." Amazon's investigation was short and sweet. Some bogus reviewers were removed, but Felicity — one of the worst offenders (http://www.amazon.com/gp/pdp/profile/A1NT2YXTUES4RW/ref=cm_c...) — is still there, despite the obvious fact that these are fake reviews. Many other obviously fake reviews remain. In fact, a new fake book with fake reviewers showed up. I genuinely do not know if this response is because of a careless employee or if Amazon discourages employees from shutting down profit streams.
It would not be to hard too for amazon. Use reviews that are a 'Verified Purchase' and highly 'helpful' reviews as a training/test set. Then machine learn a weighting to every review based on the (product, reviewer, other reviews, etc features...)
The outcome could actually hinder tactics like this for products that game the system much like how link farms hurt sites they were propping up when Google figured out how to stop sites from gaming a search engine.
I'm sure there's a lot that could be done with this, but some run of the mill NLP seems like it could at least help. I'm not sure the plausibility of this at a large scale, but it seems like an interesting problem nonetheless.
If this were a novel, economically hostile and unsubstantiated feedback would be considered economic warfare.
BUT that means you still need a critical mass of genuine reviewers who give your book positive reviews, and for technical books, sometimes that's hard to come by. One way to do this is to get lots of people reviewing your manuscript pre-publication. That has lots of other obvious benefits, too.
Reply to reviews, especially if a reader had difficulty with a code example and you can help them.
As an author, you make very little: maybe 5% of the cover price. So money is not an incentive. You are either writing because a. it looks good on your resume, or b. you really care about this topic and feel like this book should already exist.
Books are sold based on number of pages, not quality of content. The publishers know that if your book has 100 more pages, they can tack on another $5 - $10 to the price. So there's heavy incentive to produce a lot of content.
Readers don't want to buy multiple books: they want the one book that will cover all of their needs. So they will buy mostly based on the table of contents.
So if you want to optimize for your own monetary gain, the best book you can write is one that is big, and has an impressive table of contents, but took very little time to write. So we end up with books that have a very poor signal to noise ratio. For example: my book has code samples in Python. A couple of readers have asked me to write an appendix that shows them how to set up Python on their computer, play with the REPL, etc. I think this is totally useless. They can google and get up-to-date information, and the appendix will be out of date soon. But it is very easy for me to write and would pages to the book.
So as an author, the best job you can do is pump out a bunch of fluff and then pay for reviews on amazon. It is really frustrating to see, but that's how the incentives are set up here.
There isn't much reason to use a publisher, unless its a reputable one who fronts you an advance and knows how to market your book for increased sales.
I'm using a publisher because:
- They have the resources to market the book. The book's not done yet and I've already sold ~2000 advance copies. I'm trying to reach as many people with this book as possible, and it would have taken me a lot longer to sell 2k copies on my own.
- I'm trying to learn how to write a tech book. My editor is the rare breed that cares about making a good book, and has 20 years of teaching + writing experience. I have learned a lot from him.
(To be fair, they were awesome with me and I had a blast, but damn, 2000 advance copies?)
What's the topic of the book?
It is one of their better selling MEAPs :) I have also put a lot of work into it. I don't think 2k is typical.
I've already learned all the material listed in the table of contents, so I might not get much from reading it, but I wish I had seen that book 15 years ago.
I found a recently published technical book on Amazon, which had a couple dozen five star reviews. I read a few of them, and it seemed great. The author's biography said he worked at X, a very well-known software company. I bought a paperback copy, read it, and was disappointed. The book wasn't terrible, but it wasn't written very well, and did not contain the technical depth that it appeared it would on the reviews. The book turned out to be self-published, and the technical editor was the author's boss at X. The other editors were the author's family members.
I went back and looked at the reviews, and found that many of the five-star reviewers worked at X, or if I couldn't figure out their employer, they happened to be located in the same metro area as X's headquarters. One of the Amazon reviewers is even mentioned in the acknowledgements of the book.
I think all the reviewers had good intentions to help their friend and colleague, but I think it's still misleading, as you cannot expect a someone to give an impartial public review on their colleague's work.
All in all, the book was not total junk, but perhaps should have been a 3/5 star book instead of 5/5.
So I have the impression that Amazon is pretty good at weeding out good reviews from people you know, while still being complete rubbish at policing obvious spam reviews, and doing a terrible job of suppressing reviews from obvious trolls (my book has a two-star review from a guy who also gave "A Tale of Two Cities" a two star review, but you wouldn't know that unless you looked at their review history, and who can be bothered to do that?
As both an author and book-buyer, Amazon's review system is a complete loss. Every single book I look at has both good and bad reviews with almost no way to tell if they are based on standards at all relevant to my taste. Any review that doesn't say something along the lines of "My taste runs to X and this was a great example of X" or "My taste runs to Y and while this was kinda-Y-like it failed in these respects" may as well not exist.
What would be Amazon's best possible approach to dealing with this? Does there exist software good enough at distinguishing potentially fake reviews from real ones?
I've never published a dishonest review, but I have reviewed a product or two that I bought somewhere else (not on Amazon), but I certainly "bought the product."
Yes it still means a person can buy the book, then give it a bad review, then return it, however the effort to do this would be higher AND there would be a purchase pattern that begins to show itself.
When the purchase price is so low, astroturfing with "verified purchase" reviews looks to be quite feasible.
I wonder if Amazon has enough info to create a Googlesque filter bubble, so reviews are per user weighted to account for reviewer\user interest overlap.
I learned this from fiction, but it applies to nonfiction too. First of all, books are categorized by what you put in the keywords section in KDP, if they're self-published, or by whatever your publisher put in there if you're not. Let's say you write a book about Perl and cooking. It may outsell pure Perl books because more people like cooking than Perl. It might be a great cooking book but a poor Perl book - or maybe you're lying about the Perl thing, and it's just a cookbook. But it'll top the Perl category as long as Amazon believes you that it's about Perl, and among the Perl books, it will have the highest sales rank.
Unfortunately, what I bet is happening is that this book is legitimately somewhat about Perl, and the author tagged it as such in the keywords. But its sales aren't coming from winning the Perl category alone; they're coming from that and PHP and beginning programming and generally from being ranked in multiple categories.
Your best bet to get your category back is to try to convince Amazon it's not about Perl at all. Good luck - I didn't look close enough to see if that's a reasonable claim or not.
1) Visit Fiverr [1]
2) Repeat
Sadly, I don't see very many ways to overcome these kinds of review factories. Maybe Amazon should start doing sting operations on these services.
[1] https://www.fiverr.com/search/gigs?acmpl=1&utf8=%E2%9C%93&qu...
I got curious if the people reviewing these items are genuine buyers or otherwise, so I click through to one guy's profile (verified purchase). Turns out he posted hundreds of one liner reviews for a plethora of esoteric products(all verified purchase) all on Jan 12, 2015.
They are out there.
This is a big problem for more than just niche books. When I tried to find a new fiction book by browsing categories I found the same books listed in every major category, i.e. it's difficult to believe the same book ought to be listed in Sci fi, horror, self help, history, and current events.
The other thing I notice is the titles are Kindle. Why are dead-tree products, arguably a more liberal format, mixed with a e-book products that are proprietary and closed?
I'm not sure they can. There are things Amazon seems able to do well, but this certainly has never been one of them. And I don't know that anyone else at half Amazon's scale or larger has done much better.
Lol. chuckle
It basically analyzes reviews to detect if they are fake or not.