Thursday, July 27, 2017

Don't Surf Amazon in your Underwear. The Algorithm is Always Watching. (by L. A. Kelley)


Don’t worry. There’s no math. We won’t talk numbers because at the most basic level an algorithm is merely a set of rules used by a search engine to solve a problem. At a more advanced level, it’s code to tell how to sort and load product lists in response to a query. They’re not scary. You may even have used one today.  Every time you type a question or search an item on sites such as Google or Amazon it is the algorithm that supplies the answer, not some gnome living in the laptop. Google uses algorithms to scour websites for information. The Amazon algorithm does internal searches on its pages, but whether you’re looking for a vacation rental from one or a romance book from another, the basic process is the same. Using a programmed set of steps, a search engine retrieves results that the algorithm’s code says most closely matches the query.

Where did the Amazon algorithm come from?


Not some alternate dimension, but Palo Alto, California from an Amazon company called A9. The process to develop an algorithm starts by analyzing data in a catalog of information; in the case of Amazon, it’s the product description and keywords on the website. Descriptive text for every item is indexed, meaning scanned for certain words and phrases. The algorithm is also written to check past traffic patterns for you and others who searched for similar information to the query. It then ranks them and returns the results. As you type in keystroke after keystroke, the algorithm is already trying to determine what you want. This is why when you’re hunting for a fantasy, by the time you type in f-a-n-t, the drop down box appears with preliminary results and, son of gun, one of them is fantasy.

The better the algorithm understands the meaning of a query, the more able it is to retrieve accurate results. To that extent, algorithms examine the words and even the meaning behind those words. If a customer on Amazon types in the search box Texas barbecue, the algorithm first checks to see how many categories it can locate. It sees two; Texas and barbeque, and then works to bring up results that match both by scanning product descriptions. A well-written search algorithm is so strong it can even adapt to misspellings and offer suggestions for additional words. Amazon also keeps track of your buying habits so the algorithm can anticipate desires. This is why you get those helpful emails that start “Based on your recent visit, we thought you might like…” It tracks not only purchases, but also clicked items that caught your eye. 

What does this mean for the Amazon sales rank?
The information Amazon uses to suggest products during a search is based on terms in product descriptions from the catalog. For books, that means the blurb next to the book cover and the key words, but those aren’t the only factors. The algorithm also checks sales rank and that’s more complicated because rankings are always done in relation to competitors in the same category. If a book’s sales rise, that knocks down others in the same category.  One rank goes up, books close to it in rank go down. Ranking is also related to purchase time. Recent sales boost a book’s rank, so does a temporary price drop but the effect quickly fades because the algorithm also favors steady sales over dramatic surges.

What about those dreaded reviews?
Reviews on Amazon only count in rank if from a verified sale. Gifting a book is nice, but only helps sales rank if the gift is redeemed within 24 hours. The algorithm also gives more weight to newer reviews, reviews from verified Amazon purchasers, and reviews voted as helpful by other customers. Price has no effect on sales rank. Enrollment in KDP Select or Kindle Unlimited won’t confer any additional advantage. While books don’t receive an added boost in rank, downloads through Kindle Unlimited and Kindle Online Lending Library are treated as sales. Pre-orders also count as immediate sales which is why publishers love them.

What’s the myth of high sales and rank?

High sales rank doesn’t guarantee high placement in search results. Surprisingly, sales rank isn’t the only determinant. Other factors include relevance, keywords, sales history, product description, and available inventory, so a book with high sales rank may appear later in search results than lower-ranked books. How can you help your book along? Techie experts offer the following advice.

  • Have a well-written description.

  • Use keywords in the book description, but not more than once because this can look like keyword stuffing. Amazon frowns upon the practice and you can get dropped in the ranks.
  • If a reader with a verified sale lets you know they liked the book, ask them to leave a review.

  • If you gift a book ask the recipient to retrieve it immediately or it won’t count in the sales rank.
     


About the Author
L. A. Kelley writes sci-fi and fantasy adventures with humor, romance, and a touch of sass. The A in her name does not stand for Algorithm although she wishes it did because that would be cool. Check out her Amazon Author page.



7 comments:

Maureen said...

Wow- what an informative post. Thank you so much!

CJ Burright said...

That's great info!

Jane Kindred said...

How dare you cast aspersions against the gnome living in my laptop! (But thanks for the fascinating info...and for the story idea!) ;) Also, a gnome wrote this comment.

Diane Burton said...

What super info. Thanks for posting. I think Jane's got it right about the gnome, though.

Barbara Edwards said...

Sounds more like a game of caards where you need to keep shuffling. Sigh.

Aanchal Kaura said...

You really describe better for Amazon algorithm. Yes, Amazon track us and we need to do all SEO stuff in under of A9 algorithm to increase the visibility of our product in search result. Go for full article for Amazon SEO: https://bit.ly/2CLOjYJ

Anonymous said...

go to this web-site replica bags vancouver hop over to this website replica bags koh samui her latest blog replica bags thailand