My last post was an open letter to Amazon head Jeff Bezos. Today I got a call from an “Executive Customer Relations Specialist,” a nice young man who explained what he could find out about the incident. We talked about their review problems for some time, and I continued thinking about it long after (as usual.) So of course I couldn’t resist writing them a memo:

[This message is intended for Bnnnn Bnnnn, ECR, but I have no direct email for him.]

Bnnn —

Thanks for taking time out to talk today. Our conversation has triggered a lot more thought, and since I now know you may actually be able to convey some of what we’re thinking out here back to the people who decide these things, I am writing in the interest of improving Amazon and self-publishers’ cooperation.

Why You Should Listen To Me – Qualifications

[Various accomplishments and credentials omitted] I was one of your earliest customers, perhaps back around 1995, and I have been a big supporter, doing an increasing amount of business with you over the years. I managed the investments of [tech guy], founder of [bubble-era company name], during and after the dotcom bubble era, and came away with [large sum of money], so my book earnings in retirement are not critical to me — unlike some of my fellow authors, who depend on Amazon sales for much of their income, and are too afraid to speak up to you in public. In my role as portfolio manager, I became quite familiar with e-commerce and marketing.

Market Trends

Amazon is a key actor in the restructuring of publishing and general retailing, disintermediating and lowering the cost while increasing selection of goods. Books are especially suitable for this, since they are not commodities — each book and its reader/buyer have a different relationship, and there’s no simple linear scale like star ratings which can predict how satisfactory the book will be for that reader. Formerly publishers treated books like produce — with a big marketing buildup, often with paid or negotiated display space in bookstores, and a short shelf-life, with the unpredictable return rate adding to risks and costs. This led to fads and copying of trends, flooding the market with books similar to previous bestsellers and shutting out some quality books that were less commercially promising.

We understand why legacy publishers want ebook prices to be very high, often higher than print — they control the print model, which can be very profitable still, and want to slow the disintermediation restructuring, which leads to a world where they have a lesser role. This is damaging the quality of books produced by their system, and their low-paid lower level employees keep out a lot of fresh new perspectives, especially if they’re not in agreement with their politics.

The future is with online groups of readers who cross-recommend books to each other. Online communities are not likely to promote crap books like those that make up 95% of Amazon’s new books catalog. All of us online have our own reputations to guard, and we don’t push stuff on our readers that they are unlikely to like. Amazon needs to encourage this kind of community.

The Problem of Crap — Book Discovery in an Age of Excess

We all know you have too much bad material on Amazon — quite reasonably, how could you judge quality with an automated process?

When I started relying on Amazon for books, you emphasized your algorithm for recommending suitable books based on the ratings of the reader for other books they had read. This has disappeared, and you recommendations are useless now. I relied on the list of new science fiction books you made available, and bought from it based on author’s reputation, “institutional” reviews, and lastly customer reviews. This is not working at all now, as there are far too many new books, some of the best writers are self-published, and the Big 5 push less quality stuff; most small and self-pubbed writers can’t afford the delay and expense of Kirkus reviews, and the other big reviewers refuse to review us. The gatekeepers have narrowed the gate so much that many quality writers can’t get through. Meanwhile, voracious readers of genres like Mil-SF can’t find enough to read, and often end up buying low-priced books that are barely literate to keep their habits going. Genres like Romance and Mil-sf are a big part of your ebook sales, and are very poorly served by the Big 5 — not enough books in some niches are legacy-published to satisfy reader demand. This situation will only get worse as your review policies make it harder for us to meet that demand by making it harder for us to get new books off the ground with enough reviews to round out the customer’s knowledge of whether that book is likely to satisfy.

Misuse of Customer Data – Violating the Customer’s Trust

I respect Amazon as one of the most ethical companies, working hard to make the customer’s lives better. This image was gained by being exceptionally careful to fight for customers when short-term profits might have been easier by agreeing to supplier’s demands. You risk that image, though, when you use private customer data in the interest of anyone but that specific customer — it simply does not fly that you can trace customer’s relationships and injure them as a result of your desire to protect the review system’s integrity. Using customer data to inform that customer about products they might interested in is fine; using it to erase the reviews they spent time and effort to share on your system is damaging to customer trust and results in strange advice like “don’t send de minimis gift books via Amazon to your literary group, it will forever bar them from reviewing your work.” It’s quite creepy and damages your image, the kind of damage that eventually leads to government regulation or antitrust issues.

Better Ideas – Some Problems and Solutions

Problem — fake reviews, generally paid-for. If anyone can review a product at Amazon without proof of purchase, then there really is no way to stop this entirely. The two ends of the problem can be addressed: 1) any account that has generated unusual numbers of extreme positive or negative reviews can be warned, then barred from reviewing if it doesn’t stop, and 2) reviews from verified purchases can be given display and rating priority, which I understand you’re rolling out now.

Problem — Helping new authors get real reviews, because they absolutely have to have a way to distinguish their good product from the flood of crap. You are actively damaging us now, but you could implement a system allowing a book’s author to send out codes redeemable for review copies, then let the reviewer use that code when entering their review to give it the flag “This item was reviewed in return for a certified review copy.” This neatly solves two problems: our difficulty getting DRM’d review copies to reviewers, and your fear that we will only select favorable reviewers. The reviewers can be watched and rated using your account system, so that those who are good get higher weights, and those who regularly trash everything or five-star everything get low ratings. Currently NetGalley has this market sewed up, but at high prices most authors can’t afford — $150 or more per book. You should displace them and give more support to the small and self-publishers that are your future suppliers.

Problem — review “skew.” Where an author’s successive works get higher and higher ratings, because as a known quantity, those buying his/her books are more likely to be previous readers and fans who already like that writer’s work. I’m noticing that now, where the list of people who agreed to read the ARC of the third volume of my series tend to like my work already, and gave it a higher rating than a more randomly-selected previous groups which included more science-phobic readers. But this is actually not a problem; it is the expected result of consumer preference and a writer’s reputation. That writer clearly has readers who love that style, and if the rating attracts a new reader who hasn’t read the reviews and noted how science-heavy those books are, then that reader may be disappointed. But you’ve protected them as well as you reasonably can.

So, recommendations:

— As you are already doing, weight product ratings according to likely reliability of reviewers,
— Additionally rate reviewers themselves, to weed out the bad ones and give the best more weight in ratings.
— Stop erasing reviews because of de minimis connections, which abuses your access to customer data. At most put a disclaimer tag on suspect reviews and underweight them in overall rating.
— Implement a “review copy” code system, which authors can use to send out review copies. These reviews can be quality-controlled statistically and help good authors gain enough sales to eventually have a larger number of organic customer reviews.

These recommendations are based on the special nature of the ebook market, but some of them can also work for physical product reviews, like keeping ratings of reviewers as a weighting factor.