👙 Naked Girl on Balcony

NOTE: if you don’t see images below, be sure your email reader loads remote images

🪆 The Naked Barbie Chronicles

I work from home on the 32nd floor of a high-rise on Rainey St in Austin.

A week ago Wednesday, I was crafting the next issue of this newsletter.

I felt the urge to stretch and grab a Coke Zero refill.

As I was moseying to the kitchen, I made a pit stop at my balcony to play clean up crew for Zoe’s little “gifts” (shoutout to Fresh Patch real grass - it’s the Rolls Royce for high-rise canine living).

As I'm on poop patrol, my spidey-senses tingled. I glanced over the railing and to my surprise saw Balcony Barbie!

One floor down and one building over, there she was, in all her naked glory, casually chatting with a guy who apparently missed the 'clothes-optional' memo.

I kid you not, this is a 100% organic, non-GMO, gluten-free true story.

Now, was she feeling the 108 degree Austin heat, or maybe she just thinks “Hey, it’s a free world, let’s feel the breeze?” Who knows?

It got me pondering about us, the mighty Amazon sellers. Sometimes, we too feel exposed, trying to decode the enigmatic dance of the A9 algorithm.

We do want to stand out, in all our glory, for the world to see.

We’ll cover that today (no pun intended).

P.S.

If you're doubting my Balcony Barbie chronicles, scroll to the end of this newsletter for proof!

ARE YOUR SOCKS ON TIGHT? - ‘CUZ I’M ABOUT TO KNOCK THEM OFF

• 💵 BANK SOME CASH this Q4 with this incredible affiliate tool

• 🅰️ One sentence in a listing could soon match 100s of search phrases

• 🚑 How Temu is crushing Amazon on price for low ticket items

• 🧮 How to use AI to turn your Amazon spreadsheets into info GOLD

• 👺 Surprise! Not all Amazon sales screen shots are true

• 💡 25,000 awesome ideas for A+, storefronts & brand stories

💲 BILLION DOLLAR STRATEGIES → what works to make moolah

Keyword stuffing in your Amazon listing may soon go the way of the dinosaurs and be irrelevant for ranking on Amazon. 

A single keyword, phrase or proper flat file set up with browse nodes and optimized category attributes, along with some sales and conversions, may soon be all you need to rank on anything related to your product.

CURRENT A9 ALGORITHM

A recent deep dive published by the guys at Seller Sessions talks about how the mysterious A9 ranking algorithm at Amazon works. While generally accurate, its conclusions are based mostly on the “2016 Sorokina Paper” and testing.

The article goes in depth about some key components of the A9:  Your ol’ buddy Kevin read all 10,000+ words, and here are the highlights:

Key Components of the A9:

• Product Search Scores: How well products match search terms.

• Query Category Score: Gauges the relevance of a search query to a product category based on clicks, purchases, and word combinations.

• Hunger Score: Eagerness of a category to be selected after a search.

• In-Category Relevance Score: Assesses how closely a product matches a search query within its category.

Notable Insights from the Seller Sessions deep dive:

• The "Honeymoon Period" post-product launch is random.

• Cold Start Override allows for manual adjustments to rankings.

• Amazon uses data purchases, add-to-carts, and clicks for training.

• Major challenge: Showing results from multiple categories.

• A9's solution: Combine search results based on historic data and predict searcher's intent using a language model.

• Power Law Distribution affects how products are ranked based on the popularity of search terms.

• Query Ranking Module structures and understands user queries, breaking them down for the algorithm.

• Behavioral features, like sales and user context, play a significant role in product rankings.

• Brands should prioritize keyword relevance, especially when launching.

COMING SOON TO THE A9 ALGORITHM

Seller Sessions’ breakdown is good, but a very recent paper gives sellers better insight into what’s probably about to happen with Amazon search results.

In a nutshell, Amazon is working on both physical GPU based technology and its own unique e-commerce LLM. It can process massive AI-related search analysis to serve up results in the blink of an eye.

This could take away the need to be indexed for everything and pretty much eliminate keyword stuffing and how we try to optimize listings to rank today.

Let me introduce you to my friend ‘BERT’ (buy him a beer so he talks more).

BERT was explained by Amazon this past May in Osaka Japan at the 27th Pacific-Asia Conference on Knowledge Discovery and Data Mining.

You need to have a PhD in data science to deeply understand what they are talking about.  But let me break it down for you in simple language.

Traditional methods of matching products with search queries just look for exact words or phrases in the product data & descriptions. This is fast and simple but has some problems. For example, if a customer makes a small spelling mistake or uses a synonym, they might not get the results they want.

To fix this, newer methods use something called "semantic matching." This is like understanding the meaning behind words rather than just the words themselves. It's like if someone searched for "sneakers," the system would also know to show "running shoes."

Now Amazon will soon be launching a new method to improve search results using a language model called BERT.

There is a four-step method to train an e-commerce LLM model that can efficiently match queries to products to buy:

• Domain-Specific Pretraining: Train it on e-commerce data. E-commerce language is different, so a general model won't work well.

  • They created a special vocabulary from around 1 billion product titles and descriptions from 14 languages, fine-tuning with 330 million query-product pairs using AWS data, taken from multiple countries in at least 4 languages.

  • Tools like Deepspeed and PyTorch were used, and training was done on AWS P3DN instances.

• Query-Product Interaction Pre-finetuning: They improve the model by making it understand how search queries relate to products. They use a dataset where the queries and products have a strong relation, and they mask parts to train the model to predict the missing parts.

• Finetuning for Matching: This stage trains the model to match queries to products. They employ a bi-encoder system which is efficient for large-scale data. The training uses a scoring system that labels pairs as positive matches, hard negatives, or random negatives.

• Knowledge Distillation to a Smaller Model: The final step is to transfer the knowledge from the large, well-trained model to a smaller one that can work quickly in real-time for product search.

Bottomline:

A single sentence in a product listing could soon match hundreds of search phrases without keywords needed explicitly mentioning them all.

Tools like Helium 10, Jungle Scout and Data Dive will need to pivot with their keyword tools, or they could become irrelevant.


As Sam Cooke sang in 1963 …

Then, I go to my brother

And I say, "Brother, help me, please"

But he winds up knockin' me

Back down on my knees, oh

It's been a long

A long time coming, but I know

A change gon' come

Oh yes, it will

BILLION $ TRIVIA: There is no such thing as an A10 algorithm - that’s marketing click bait by Youtube Gurus. It’s the A9. It evolves over time like you did during puberty. But its name doesn’t change. 

The legend goes that a few data nerds were sitting at a bar wondering what to call it. One guy says, “Hey fellas, there are 9 letters in ‘algorithm’ and it works for Amazon. Let’s call it the A9.”

Great tutorial and cheat sheet from the guys at Superhuman on how to use AI to analyze your spreadsheets. Click photo to enlarge it for easier reading.

 🔥 PRIME PICKS → Amazon’s sandbox

• AMAZON GETTING STUNG — Facebook, Instagram, and TikTok do not want to drive traffic to Amazon. They want to be Amazon. The big boys realize they are crucial for new product discovery.
  
  Starting August 31, urls in YouTube Shorts descriptions, comments and live feeds will no longer be clickable.
  
  Though Tik Tok won’t officially confirm it, they, along with Meta are making forceful moves to stop users from directing people to external websites to complete a purchase. Tik Tok is shutting down its Shopify-powered storefronts in September.
  
  In April, Facebook announced that Shops on Facebook and Instagram would no longer direct people to an e-commerce website to complete a purchase. The change has not yet affected ads.
  
  

  

• TEMU CRUSHING AMAZON on PRICE — Amazon is ignoring, rather than competing with, low prices of the same or similar items on Temu. Amazon is excluding Temu from its price searching algorithm that checks if products sold on its platform are competitive with rivals, saying the site doesn’t meet its standards.
  
  Temu is quickly taking marketshare from Amazon when it comes to sub $30 items (sales on Temu in the US are up 1000% since January to $1 billion+). For example a microfiber hair towel is $15.99 on Amazon, the same one on Temu is $2.38.
  
  Amazon's rising costs for sellers is Temu's opportunity. Amazon simply can't compete on low price items with Temu. Experts believe within a few years Temu will become THE place for cheap products and low prices in the US, not Amazon. In China alone, 850 million already use Temu and its associated platforms.

  • Temu negotiates volume discounts with Chinese manufacturers

  • $5+ subsidy per shipment from Chinese delivery companies

  • Avoids tariffs & duties importing to US since < $800 exemption

  • Fulfills from Guangzhou, China 3PL with cheap labor

  • Temu plans to spend $4.3 billion in advertising in USA in 2024

  • Average order value is $25, customers place 30 order per year

  • Guarantees delivery in < 10 days, gives $5 credit if late

  • Losing $, but building massive data vault they can leverage

🛠️ FEATURED TOOLS and RESOURCES → all the kool kids use these

• Having a brain fart for inspiration when creating A+ content, insert cards or infographics? Two knights are here to rescue you.

   
  1/ George’s Blog → 25,000+ ideas for A+, storefront & brand stories, filtered by product and category, updated weekly.
  
  2/ Behance → search by “Amazon Insert Card”, “Amazon A+”, “Amazon Infographics” and more for a gaggle of ideas for your graphic designer.

• This is an awesome AI tool for branded product photography. Just drag your product photo into the canvas, then visually describe the scene you want surrounding your product. In seconds you’ll have a high resolution file ready to download.

🛒  MONEY NEVER SLEEPS → real products … real sales … real reviews

Did you notice the line through a couple of the words directly above? That’s because you can’t always believe what “gurus” tell you.

While you don’t see it as much nowadays (because most people teaching are no longer selling), it used to be common to boast with sales screen shots.

Like this one:

But is it real or Memorex (the old folks will get it)? Take a closer look. Seems he forgot to add a “comma” after the “9” in today’s sales. Also the average selling price is all over the place. 9540/173 = $55, while 65,844/368 is $179.

Hmmm. Maybe that’s why this happened a few years later (worth the click).

Amazon FBA “gurus” on social media platforms often exaggerate initial investments, product revenues, profit margins and the idea of "passive income."

Honest FBA talks about the problem with Amazon FBA Gurus and gives examples in this Youtube video.

THERE SHE IS → BALCONY BARBIE …

✌🏼 That’s all folks.

If you enjoyed this issue, please reply and let me know (it helps with deliverability).

Be a pal. Tell a friend below. That way I can send you the 21 hacks pdf!

Back at you on Thursday …