Even the experts can't say why a review is rejected or how to fix it.
Several articles in the expert community support area say that there is a robot scanning review text, and that it is incredibly picky. Rejected reviews have been submitted for the experts to scrutinize, and the experts have made many suggestions about how to reword a rejected review.
There are almost no reports of a successful edit after following an expert's repeated and varied suggestions. Experts have lots of examples of what could trigger a rejection (mostly involving two adjacent words, with spaces removed, containing an offensive; one has even observed that an explanation mark will trigger a rejection!) But they can't find the specific problem in the submitted review.
Further, people writing reviews get NO feedback from the robot -- not even a message to tell them that their review has been rejected, let alone a clue as to why.
So how about having the robot send a message to the submitter, listing each objectionable phrase and a couple of the surrounding words so it can be located by the writer? What would be ideal is the original text with each letter combination that triggered a failure rule underlined or bolded.
Remember some things are very subtle, and even experienced human experts can't find them. Only the robot knows why, and so far, it's not telling.
- If ebay is worried about some automation testing the robot to learn what it doesn't like, just limit the frequency of responses. Robot: "Here's everything I don't like and you can resubmit your updated review in an hour for my reconsideration."
-
- A human needs to be in the loop somewhere who can adjust the robot. Perhaps after the third rejection, the rejection message could be forwarded to the programming group. Even if they only looked at 10 rejections a day, they could see patterns and adjust the robot's programming.
-
- This is really an application which requires machine learning, adaptive neural nets, etc., with human "yes/no" feedback. A script kiddy level programming of rules (delete all blanks, scan for list of words) totally ignores context and can't allow any evolution in the rule set. It would be nice if ebay made an effort here, or looked for a provider that did it right.
-
- There are several "rate and review" service providers that do a much better job handling reviews for other sites. If ebay is programming the robot themselves, they are re-inventing the wheel and could easily get the experience of a specialist company with a turn-key solution.
-
- Look at ebay's largest competitor. The strength of their review system is a major offset to the poor quality of their search engine. In other words, buyers are entering their site after Google or Bing finds a review or production question answer. If the A3 search engine doesn't give an excellent find on the first page of results, potential buyers leave the site and use other tools/sites to locate the items they want. Good reviews can bring them back.
- Said company also tried operating a "human-less" system several years ago where all customer interaction was machine based. E-mail scanning didn't work. They found that humans had to be there to handle problems that couldn't be categorized, or required a representative to have language skills (reading for comprehension) to successfully resolve the customer issue.
WAKE UP EBAY!
Stop leaving money on the table.
Learn from the experience of others.
Replace the review robot!
