Abolish negative votes on reviews?

---

Amazon's voting question

See Was this review helpful to you? for an analysis of this question and what it might mean, if anything. For a more detailed look at those NO votes, see Spiteful votes, where I describe many of the reasons for customers voting NO. I do not discuss the issue of abolishing such votes altogether in either of those posts, though this idea has been around ever since Amazon allowed voting on reviews. A lot of reviewers would like only YES votes to be allowed and for all NO votes to be removed. Nobody likes receiving those NO votes, but opinion is split as to their worth. Arguments about the issue work on several different levels.

The abolition question

This generates a lot of heated debates. Here, I present the case for, followed by an analysis, then followed by the case against. I then link to two other blog posts that are relevant to this topic.

The case for abolition

The system as it is allows trolls to flourish, which upsets a lot of reviewers. Some trolls stalk their targets by spraying NO votes around, though the anti-campaign software that came in with the current ranking system has made life more difficult for them. Prior to the introduction of the current ranking there in March 2010, Amazon UK didn't bother with anti-campaign software of any kind although Amazon USA had software that curbed the worst excesses. In early 2005, I was subjected to one of the most vicious campaigns against my reviews ever seen on any Amazon site, albeit it was mitigated by a counter-campaign, but I learned a lot from that. If customers were only allowed to vote YES, it could never have happened. Here’s how it unfolded.

The UK voting farce

When reviewer rankings (in the form of the second ranking system, as the original system never made it across the Atlantic) were first introduced in the UK in the fall (autumn) of 2002, I was way down in the triple digits but I hadn't been reviewing for very long at the time. In November 2003, I finally made it to #2. Almost exactly a year later, I lost my internet connection from home due to my financial difficulties. At the time, I was still #2 but I'd given up hope of becoming #1 by then, because an American reviewer was up to #3 and it seemed only a matter of time before he overtook me. I still expected to overtake the #1 reviewer but not before the American had overtaken both of us. I'm glad I had that expectation as it was to stand me in good stead.

Without a direct line from home, I continued to write reviews, saving them on my computer for some unspecified future date when I could load them up. Soon afterwards. I started using the computers in my local library, where I was allowed two hours per day every day that the library was open. With such limited time, I initially decided not to use the library for loading reviews but I eventually did so, two and a half months after I lost my home internet connection. I started copying my reviews to floppies and posting them to Amazon from the library. By that time, I had a backlog of around 150 reviews. Now it wasn't always easy to post reviews in the UK, but the early months of 2005 happened to be a good period.

After about two or three weeks, I noticed a huge number of NO votes suddenly appearing on my UK reviews. Clearly, somebody was noticing the rapid increase in my total number of reviews and was panicking. To this day, I don't know and don't care, but it was obviously somebody who didn't like me and certainly didn't want me to become the UK #1 reviewer. The guilty person(s) could have included the #1 reviewer but I don’t assume that at all. I've sometimes been wrongly accused of voting against somebody else’s reviews. In such cases, it often transpired that somebody wanted to support me, but the effect was that my reviews got extra NO votes, so I knew not to jump to conclusions.

With this cyber-war declared, I made it my mission to become UK #1 reviewer as quickly as possible and teach my attacker(s) a lesson. The more British reviews I posted, the more savagely my reviews were attacked, and not just the new ones. My attacker(s) took their battle to my old reviews, too. Some of my supporters, noticing what was happening, counter-attacked by giving me plenty of YES votes. I didn't ask for such support and don't condone what they did, but I don't even know who these supporters were, just as I don't know who my attacker(s) were.

Anyway, the result of all this attacking and counter-attacking was that I ended up with hundreds of reviews showing vote totals of 3 out of 6, 4 out of 8, 5 out of 10 and so on, even including totals like 20 out of 40 and beyond. On one level, this was quite amusing but it is clear that none of the voters involved really understood the way Amazon's rankings worked, except that my supporters knew that 3 YES votes were a basic requirement for reviews to count towards ranking under the only system then in use. Until the current ranking system was installed in 2010 with the associated anti-campaign software, you couldn't find very many of my British reviews posted pre-2007 with fewer than 3 YES votes. I was by no means unique in that regard because a lot of high-ranking British reviewers could have said the same, although their reviews were never attacked on a big scale. What I didn't know then was that the British rankings were based on a very old version of the American algorithm, so every vote counted.

Even as I continued to catch up with my backlog of reviews, I continued to write more, stepping up the pace after my attacker(s) declared their cyber-war, partly to annoy my attacker(s) but primarily to ensure that the anti-campaign software would notice if it ever crossed the Atlantic. The anti-campaign software is better now than it was in 2005, but still imperfect. I needed to help it all I could, by forcing the trolls to cast more NO than they otherwise would have, making their dirty deeds even more blatant. For several weeks, I was able to post reviews at a crazy rate, as many as forty a week. That was really tough once the backlog dried up. I was glad to be a music reviewer rather than a book reviewer, although I've since learned that certain types of books are very easy to review, especially poetry and picture books.

When my ranking changed in mid-March 2005, it was down, not up, because the American finally did what I'd always expected and overtook me. The timing was, of course, unfortunate because it appeared to many people, including some of my friends, that my attacker(s) had won. I didn't agree, thinking that I must be gaining on the #1 reviewer regardless of what else was happening. If I had to be content with being second behind the American, that would be fine.

As soon as the American overtook me, he stopped posting reviews in Britain. He eventually started again but not for a long time, and he has since given up for good. I was surprised by this and it was not what I wanted. British customers don't have many reviews to choose from except on the most popular items. Nevertheless, it was clear that the American did not want to get entangled in any voting nonsense. With the #1 reviewer only posting reviews at a modest pace, the American not posting any and myself posting relentlessly, I knew that there could be only one outcome in the end. Meanwhile, my attacker(s) eased off, presumably thinking that my drop in the rankings was a result of all their spiteful votes. I also noticed that the American's reviews were not subjected to attack as far as I could see and I was pleased about that. The attacks against my reviews had been intended to stop me and were nothing more than that, so it wasm't simply a case od defending the #1 position. For the next month or so, the top three positions remained static with myself at #3, but I was undeterred.

On the 18th day of April in 2005, the American finally became the new UK #1 reviewer while I stayed at #3, but within a few hours the rankings changed again. I became #1 with the American at #2 and the original #1 reviewer at #3, so I was finally proved right in my assessment of what was happening. At this point, some of the American's supporters on the customer reviews discussion board decided that he should be the UK #1 reviewer and started voting for his reviews. The next ranking adjustment put us both as joint #1 reviewers. The American didn't like all the bogus votes he was now getting and told his supporters to stop, which they did reluctantly. I was finally able to ease off on the reviewing and things settled down with myself at #1 and the American at #2, where we stayed until June 2006. As for the original #1 reviewer, he stayed at #3 during that period except for a brief period when he dropped to #4.

During the time that I was getting all my bogus votes, I exchanged e-mails with Amazon in Britain and America. They said that they couldn't use the anti-campaign software until a lot of other software was upgraded. I suggested that maybe they could at least look at who was casting the votes and send them some kind of e-mail warning them about their voting. Amazon didn't respond to that e-mail, but several weeks later, much of the campaign voting stopped abruptly, and not only on my reviews. Other reviewers noticed that they were getting far fewer votes on their British reviews. I don't know if Amazon actually took up my suggestion but the circumstantial evidence is interesting. The main thing was that the silly voting on both sides stopped for a while, though it eventually resumed on a smaller scale.

In a period of three months, I acquired something like 13,000 YES votes in Europe. In those days all European votes were added together and displayed as a combined total, which clouded the issue because I've also posted reviews in France and Germany. I acquired many thousands of NO votes in Britain, but I didn't know how many until 2009, when Amazon UK finally provided proper totals that had already been available for USA reviews years earlier. When I finally saw how many NO votes I had, it was bad but nothing like my worst fears. Really, I wouldn't have been surprised if the total, which had continued growing since 2005, had been over 20,000, but it was actually around 10,000.

In June 2006, Amazon changed the British ranking algorithm, bringing it into line with America (thus showing me that they hadn't been the same earlier). Despite my not posting any British reviews in the first half of 2006, I retained my #1 position. The other two reviewers I've mentioned dropped a few places but later recovered some of the lost ground.

In 2009, I was subjected to another nasty campaign against my UK reviews, though not on the scale of the 2005 campaign. This time, the counter-campaigners came in for a while but soon stopped, leaving the attacker(s) with a clear run. Amazon UK still didn't use any software to clear out campaign votes, but the 2009 campaign was so blatant that it offered the best chance yet for me to convince them that they should. Without the helpful votes from counter-campaigners, I hoped to get Amazon's sympathy this time. I know enough about Amazon's software to understand that the avalanche of NO votes would be easily spotted and eradicated by the software. I didn’t get the response I hoped for, but I didn't really expect to. However, by then I knew about the current American ranking system and the associated anti-campaign software, so I pinned my hopes on that.

While the 2005 and 2009 campaigns were particularly nasty, I was subjected to all sorts of campaign voting of both kinds in the UK, going all the way back to 2002. I therefore couldn't take my UK votes or ranking seriously until the votes were cleaned up on March 9th, 2010.

In March 2010, the current ranking system was installed simultaneously in Britain, France and Germany. The anti-campaign software came with it and wiped out well over half of my YES votes and over 90% of my NO votes. As a consequence of so many votes being wiped out, my % of YES votes rose from 79 to 94 and I was installed #1 on the current ranking system. I hadn't expected that, thinking that maybe I'd recover to somewhere around 85% or a little higher at best.

Analysis of the UK voting farce

I probably posted some reviews in 2005 that I might not otherwise have posted, and some of them were very successful, as can be seen after the bogus votes were wiped out five years later. A lot of my reviews lost spotlight positions because of the NO votes, but many regained them after the clean-up. It is impossible to know whether I gained or lost overall but it doesn't really matter. I won. Amazon won. Customers won. Trolls lost.

All of the above goes to show how nasty things can get, but they also show that trolls don’t necessarily get what they want. If reviewers are easily discouraged by trolls (and some are), the trolls win. I was made of sterner stuff and defeated them. Posting reviews at a crazy rate not only showed the attacker(s) what I was made of but also left markers for the anti-campaign software when it eventually arrived five years later. Because I forced my attacker(s) to be more vicious with their spite votes, the anti-campaign software couldn’t help but notice. However, that saga and other experiences have proved to me that votes and rankings don’t indicate nearly as much as people might think. Indeed, Amazon’s ability to come up with a markedly different ranking system that ran alongside the second system for more than three years prior to the latter's abolition confirms it. A reviewer’s ranking can depend more on the method of calculation than on what that reviewer actually does. In particular, the curreny ranking system focuses heavily on NO votes while the second ranking system focused heavily on YES votes, although there are other significant differences too. See Amazon's three reviewer ranking systems explained.

Having experienced such farcical voting on my reviews, people might think that I’d be a strong supporter of abolition of NO votes, but I’m not because I also saw was that effective anti-campaign software can deal with trolls. That software is not perfect, but the answer to any of its imperfections is for Amazon to improve the software.

Nevertheless, the whole business of those NO buttons often leads to spite and bad feeling, and can be very discouraging to people who write reviews. You certainly need a thick skin to review on Amazon, but I doubt that abolition would have the effect that those who call for it would expect.

The case against abolition

On one forum, somebody suggested that if a review attracts 10 YES votes and another for the same product attracts 2, it is obvious which review is useful to the majority, but it‘s not that simple. The 10-vote review may be older, so maybe some or all of the 10 customers who voted for it didn’t have the choice of voting for the other review. Also, especially where books by obscure authors are concerned, the 10 votes (or even 20 or 30) may all be from friends, family and multiple accounts. You'll notice that in my description of the UK voting farce, I referred to my attacker(s), thus alluding to the possibility that one person with multiple accounts was responsible. Leaving that aside, in the case of one review with 10 helpful votes and another with 2, there are any number of reasons why the 2-vote review might be more helpful than the 10-vote review and it not be reflected in the votes. However, the NO option makes it easier to draw a distinction between a review that is genuinely helpful and one that is not. Spite votes can distort things; one consequence of the UK voting farce was to drive a lot of my reviews out of the spotlight, though many returned after the bogus votes were wiped out. However, mine was an extreme case. I remain unconvinced that they distort things as badly as some reviewers think. If Amazon UK had used the anti-campaign software in 2005 (which they said they weren't able to), the whole UK voting farce would have been nipped in the bud very quickly.

The likely consequences of abolition would in any case include more nasty comments and more reviews being driven off product pages altogether by frustrated customers unable to vent their feelings by voting NO. Also, a review that currently shows 10 out of 500 customers found this review helpful would show only 10 customers found this review helpful after abolition. I expect that a lot of bad reviews would resurface on the main product page if their NO votes were removed. In America, Harriet Klausner would be among the beneficiaries. We'll come back to her later.

Supporters of abolition of NO votes point out that there are like options on Facebook that can be cancelled (but not negated) by use of the unlike option. There is no dislike option on Facebook. I suppose this is intended to keep everything warm and fuzzy, but Facebook is a social networking site, not a commercial business. Even on Facebook, things can become quite nasty. Indeed, Facebook makes mainstream news for nasty stuff more ogten than Amazon.

We must remember that the primary purpose of reviews is to help customers make decisions about what to buy. Customers don't normally take any notice of votes, but the spotlighting algorithm does. We can argue about whether the algorithm gets the balance right (I think not) but it takes both YES and NO votes into consideration when it decides which reviews should be spotlighted. Bad reviews don't usually stay in the spotlight positions if there are enough good reviews to drive them out. No system is perfect, but if NO votes were abolished, there would be a lot more bad reviews in the spotlights. That wouldn't serve the interests of customers at all. Abolition would also put newly posted reviews near the bottom of the order, above older voteless reviews but nothing else.

In any case, Amazon will never abolish NO votes. The second ranking system was built around YES votes, with NO votes being largely disregarded. It's actually quite difficult to be penalised for NO votes under the second system. For any given review, you need at least 5 more NO votes than YES votes to incur a penalty at all. It still didn't stop people complaining bitterly about them.

YES votes aren't always genuine. In my time, I know that I've received YES votes for the opposite reasons that I've received NO votes. For example, I get NO votes from people who don't like me, but I get YES votes from people who do like me. In both cases, these votes could be related to the voter's own interests but aren't always. I've also given out helpful votes that have absolutely nothing to do with my own interests, but simply to endorse the reviewer. I do so in moderation and I normally try to find a product that interests me, but it doesn't always work out that way. In the same way, I'm sure many reviewers have done the same for reviewers they like. I don't see anything wrong in that providing that it is done in moderation, except that it nullifies the argument for abolition of NO votes.

When Amazon USA decided to design the current ranking system, they clearly realized that the old assumption, often repeated even now, that NO votes are usually spite votes while YES votes are usually genuine had long since been shown to be a myth. They also realized that any ranking system based largely on YES votes would keep Harriet Klausner as the American #1 reviewer in perpetuity. They wanted a system that kept her well away from #1 and that appears to be the primary reason that NO votes weigh so heavily in the current ranking system. Even with all the other differences between the second and current systems, you can be sure that simply cancelling all NO votes would put Harriet Klausner very close to being the #1 reviewer in America on the current system, if not actually make her #1. Even if it doesn't quite make her #1, her dogged persistence would be likely to promote her to #1 as others give up reviewing. That alone is enough to kill any chance of abolition of NO votes.

In my early days as a reviewer before I became high profile, I was able to learn from those NO votes. That was in the days when any votes I received were genuine. I am a better reviewer for having learned those lessons. I can still occasionally learn from those NO votes, usually if I get several of them on a review that I didn't expect to get that many. Most reviewers never achieve the high profile that I now have, so they retain the ability to learn from NO votes, should they choose to do so. If I hadn't been able to learn, would I have ever achieved what I later did as a reviewer? Maybe, but perhaps it would have taken longer. Votes aren't the only (or even the best) way of learning, but they certainly helped in the first few years, when there was no comments feature, and in the very early days before I discovered forums.

Placeholder reviews

See Placeholder reviews, which explains why this method of cheating would flourish if people couldn't vote NO.

Washing reviews of dirty NO votes

Reviewers sometimes decide to get rid of votes in the only way they can, by deleting reviews and (maybe) re-posting them. See Deleting and re-posting reviews for my thoughts on this issue.