Like Me? Follow Me.
Although Search Engine Land had already pointed out that the reason the site in question was ranking well had little to do with the reviews in question and Google has stated that they are not using sentiment analysis because it is too difficult to determine, today, the SEO world is buzzing with speculation about these changes - could they involve the ratings from review sites? Could they involve social signals more actively?
As per usual, until we see how different SERPs change, it's fairly impossible to tell (although we've been saying social signals are important for a while).
There are a few things, however, about the use of consumer reviews, that make them an extremely poor method of determining quality.
What I know about the authenticity of consumer reviews
It's no secret that I learned most of what I know about SEO working for a consumer review website. While there, I was responsible for moderating reviews submitted to the site and for encouraging content submission as well as for generating content myself.
What does this mean?
It means I wrote a lot of reviews. Hundreds. Possibly thousands. Under lots of different names. See, people do not add content to an empty page/category/site - they only write where others have written. So, as moderators we spent a lot of time submitting reviews on anything and everything we could think of - holidays we'd been on, hotels, restaurants, books, CD's, beauty products, electronics, etc. If we had been there, bought something, read something, or smeared it on our faces we wrote about it. But we had a series of sock puppet accounts and we randomly used those while writing - so it looked like more people were using the site than there were at the time.
This still wasn't enough, however. Why? Well, in order to rank for the product names that people searched for we had to add those products as items to the site so people could find them and review them (or click on the ads to buy them which was the real end goal). But we discovered that search engines don't like empty pages with repetitive text full of ads (this issue began with Florida and got very bad with Allegra) - so we needed to get content on empty items, particularly popular products and brands and items where we had ads with high CPC values.
So we scoured the internet for professional reviews and we paraphrased them on the site as consumer reviews. In some instances, where we couldn't find reviews - on finance products in particular - we were told to just make something up and make sure it was positive. I protested and refused to write finance reviews or reviews where the general online feedback wasn't positive, but not everybody agreed with me that this was ethically and morally wrong, so reviews of loan companies, credit cards and mortgages went onto the site - encouraging visitors who were researching finance products to click on the ads to apply for loans, credit cards, bank accounts, etc.
I won't even go into our deletion of negative reviews from some major brands when we were trying to get these authoritative brands to link to us....
What this means is that, considering the huge amount of fake content out there produced by the review sites themselves (and if we were doing it, I bet that others do it as well), using reviews as a signal not only gives big brands another huge boost, it also means that Google's using faked content - despite having spent the last 12 years trying to find ways to prevent "faked" links from having an impact on the SERPs.
It wasn't just us writing fake consumer reviews
I cannot begin to relate the number of faked reviews coming from marketers, businesses and their competitors. In one instance we spent two years moderating a dispute between two distributors of the same hair straightener.
Each business had taken the same raw product and created its own brand, and each was aware of the other. The moment business #1 started writing good reviews of its product (we knew it was them because they had to give an email address which they didn't disguise AND we could see the IP address of each reviewer), their competitor started writing negative reviews of business #1 and positive reviews of themselves. Then business #1 started writing negative reviews of its competitor, etc., This continued ad nauseum, punctuated by each business occasionally phoning up and complaining.
That was an instance where we caught it. There were others where marketing companies stupidly used their work emails, or where someone submitted 10 reviews of the same product from the same IP under different new accounts so we pulled them.
There were probably countless instances where we didn't catch it.
There were also the angry customers who would set up multiple accounts just to write multiple negative reviews - usually of a small business, the ones who get hurt the most by this type of review.
On top of this there's the issue that most people won't write a review unless they've had a bad experience. People expect a good experience so when they have one, they don't feel the need to go online and talk about it. People who have a bad experience do. So online reviews are stilted - except where they've been produced by marketers or review sites.
Consumer reviews will help big brands and hurt small businesses
So, what you end up with is a situation where reviews as a ranking factor in Google means big brands benefit when they don't really need the help, and small companies and brands will be hurt because they're going to have fewer reviews and are very likely to have negative reviews, even when they don't necessarily deserve them.
Sentiment analysis is impossible
Finally, I give you an example of why Google cannot possibly use sentiment analysis to any real degree...what if a review read like this:
"I ordered a widget from WidgetShop.co.uk and it's the shit! I mean this crap is so badass that it stomps all the others into the dirt. I mean you have to dig this shit - it makes me look so bad when I carry it around and my friends are jealous of my killer widget that they're complaining that I make theirs look stupid."
The 4 reasons why reviews as a ranking signal is a very bad idea
So to sum up:
- Review sites fake positive reviews and remove negative reviews - how can an algorithm possibly spot this?
- Marketers write fake positive reviews of themselves and their customers and negative reviews of their compeitors - how can an algorithm possibly spot this?
- Consumers tend towards negative reviews AND in some instances will write multiple ones under pseudonyms - and again, how can an algorithm possibly spot this?
- Sentiment analysis will be difficult because negative words aren't always used negatively.
Discussion
Leave a Reply