This Issue's TLDR...

• The truth about the Honeymoon Period
• Is traditional keyword research and listing optimization dead?
• Planning a rollup strategy? Read this first

👉 Did someone forward you this newsletter? First of all, give them a crisp high five when you see them. Second, head over here to subscribe and read past issues.

​

SPONSOR TrueOps Why are you paying more than a 10% commission to your FBA reimbursements provider? Seriously, I want to know. Are you sticking with your provider because they claim to recover more $? Are you sticking with your provider because some well-known "Amazon expert" (that might not even sell on Amazon anymore) is always recommending that provider? Look, man, you do you. But, for me, I'm going to pocket the extra cash from only paying 10% vs 25%. ​ Pay Just 10% with TrueOps Today

At the end of last week, my good friend Danny McMillan, and his co-author, Oana Padurariu, published a banger of an article on the "Honeymoon Period" called The Cold Reality of the Honeymoon Period and External Traffic.

I've been geeking out about it for the past week.

I've read it, went and watched Danny's interview with Anthony Lee (see below), and then re-read it.

You might not have the time (or patience) to go deep on the article.

And that's OK.

But, if you're a serious Amazon seller, you should have a foundational understanding of everything that Danny and Oana uncovered in their research.

To help you build that foundation, I've written up Frequently Asked Questions (FAQs) on the article.

This is a little self-learning practice that I picked up at Amazon that always helped me to distill complex topics into their atomic elements.

On that note...

FAQs: The Cold Start Problem and Honeymoon Period on Amazon

Q: What is the "Honeymoon Period" on Amazon, and does it still exist?

The concept of the Honeymoon Period, which suggests that new products on Amazon receive a guaranteed boost in visibility for their first 30 days, is a myth and is not supported by scientific evidence. Anthony Lee, who initially coined the term in 2015, has since dismissed the idea (again, see video below).

The belief stemmed from observations of a pattern in the early days of Amazon, where a correlation between sales rank and sales history seemed to provide a grace period for new products. However, as Amazon evolved its ranking algorithm to prioritize keyword relevance, the Honeymoon Period became obsolete.

Q: What is the "Cold Start Problem," and how does it affect new products on Amazon?

The Cold Start Problem describes the challenge new products face when they are added to an e-commerce catalog and indexed for searches. Because they lack historical interaction data (such as clicks, purchases, and reviews), the ranking algorithm struggles to accurately assess their relevance, leading to lower rankings than they might deserve.

Q: How does Amazon address the Cold Start Problem?

Amazon tackles the Cold Start Problem using a system detailed in U.S. Patent Nos. 11,269,898 (2022) and 20230367818A1 (2023). These patents describe a process that leverages machine learning and Bayesian methods to predict the potential success of new products and provide them with initial visibility in search results.

The system works by:

1. Generating "prior prediction values": When a new product is listed, Amazon identifies similar existing items in the catalog and uses their historical performance data (clicks, purchases, engagement) to predict how well the new product might perform. This initial estimate is the "prior prediction value."
2. Using query-item pairs: The 2023 patent introduces the concept of dynamically incorporating query-item pairs into the ranking process. The system learns how specific queries relate to items, improving the accuracy of matching user searches to relevant products. This allows for real-time adjustments based on user interactions with specific query-item combinations.
3. Transitioning to "posterior prediction values": As the new product accumulates actual user interaction data, the system gradually shifts from relying on prior prediction values to using the real data. This process, called Bayesian updating, continuously refines the product’s ranking based on its performance.
4. Incorporating Bayesian context: The 2023 patent integrates a Bayesian context into the ranking process, enabling the system to continuously update its predictions as new data becomes available. This allows for faster and more accurate ranking adjustments, ensuring a smoother integration of new items into Amazon’s search.

Q: Is there a fixed time frame for the Cold Start period on Amazon?

There is no fixed time frame for the Cold Start period. The system dynamically adjusts based on real-time data. The 2022 patent states that the transition from prior to posterior prediction values can happen as quickly as three days after a product is listed, depending on the amount of interaction data gathered. The 2023 patent further emphasizes this dynamic adjustment with more sophisticated real-time data processing and machine learning models.

The system recognizes a 90-day claim window during which a product’s performance is continually assessed. However, if the system determines that an item has sufficient user interaction data before this period ends, it will no longer be eligible for the cold start mechanism and will be handled by a behavioral machine learning model.

Q: Does Amazon offer any tools or mechanisms that can manually boost a product's visibility, especially during the Cold Start phase?

While technically possible, tools like "Sales Velocity Seeding" and "Manual Override" are primarily reserved for Amazon's own brands or major vendor brands and are unlikely to benefit the average seller.

The 2022 patent mentions "manual curation," where an administrator can manually assign a prediction value to a new product to boost its visibility. However, this approach is considered inefficient and prone to errors, especially for marketplaces with a large volume of new items. Modern machine learning models have largely replaced this manual process, using predictive algorithms based on historical data to generate “prior prediction values.”

Q: What is the role of "spearfishing" in tackling the Cold Start Problem?

"Spearfishing" is a technique used to help new products gain initial traction by targeting specific, niche queries where they are highly likely to rank well and receive clicks. This method involves meticulous keyword research and optimization, focusing on long-tail keywords with clear user intent.

While spear fishing can be effective, its success depends on the seller's ability to accurately identify high-intent keywords with sufficient search volume.

Q: How has the approach to solving the Cold Start Problem evolved over time?

The evolution of solutions to the Cold Start Problem can be traced through scientific papers and patents. Early attempts primarily involved manual curation and waiting for user data to accumulate naturally.

As machine learning techniques advanced, the focus shifted to predictive priors, where algorithms used product attributes like brand and author to estimate initial engagement potential.

The development of large-scale language models improved semantic search, allowing for more contextually relevant results even with limited historical data.

The use of graph convolutional networks (GCNs) emerged as a cutting-edge solution, exemplified by the ColdGuess model, which analyzes relational graphs to identify hidden patterns and make accurate recommendations from the outset.

Amazon’s continuous refinement of its systems, detailed in the 2022 and 2023 patents, demonstrates the ongoing effort to improve the accuracy, efficiency, and real-time adaptability of its ranking algorithms.

Q: What are some key takeaways from the 2023 patent (US20230367818A1) that sellers should be aware of?

The 2023 patent emphasizes that there is no guaranteed boost in visibility simply because a product is new. The system relies on data-driven predictions and real-time user interactions to determine a product’s ranking, favoring those that align with successful attributes and demonstrate strong performance.

The ranking system is dynamic and continuously evolving. There is no fixed 30-day window of guaranteed visibility. As a product gathers user interaction data, the system updates its ranking in real time, transitioning from prior to posterior prediction values. Products that attract user interest may see increased visibility, while those that fail to engage users will see their ranking decline, regardless of how long they have been listed.

Relevance is key. When selecting keywords, prioritize those that are highly relevant to your product and align with customer search intent, rather than focusing solely on search volume. Choose competitors strategically for product targeting campaigns, focusing on those who offer products with similar features, benefits, and price points.

Q: How do external traffic sources factor into the cold start problem and Amazon’s ranking algorithm?

Driving external traffic to your Amazon product listing can impact its launch. However, the quality and relevance of this traffic are crucial. Ensure that external traffic mirrors the target audience and intent of your on-Amazon efforts to avoid negatively impacting your product’s ranking.

External traffic without relevant search queries and conversions can harm your product’s performance. For products with low search volume, focusing on platform advertising, such as Sponsored Product Ads, is generally more effective than driving external traffic.

Q: Beyond the technical aspects of the Cold Start Problem and ranking algorithms, what are some fundamental principles that contribute to a successful product launch on Amazon?

• Focus on developing high-quality products that meet customer needs and address their objections. No ranking strategy can compensate for a subpar product.
• Invest in creating compelling product listings with high-quality content, including professional images, detailed descriptions, and persuasive text that highlights key benefits.
• Choose the right product attributes and categories to ensure your product is properly indexed and positioned for visibility. Download the category report and meticulously complete all applicable product attributes.
• Prioritize early advertising strategies to generate initial user interactions and accelerate the transition from prior to posterior prediction values. Have your advertising campaigns ready to launch as soon as your product listing is active.
• Continuously monitor performance metrics and adapt your strategy based on real-time data. Track key indicators like conversion rates and sales data, and make timely adjustments to optimize your listing and advertising campaigns.

***

Epilogue: Part of the reason that I'm so excited about this article is because it brings a previously internal Amazon term -- "cold start" -- into the public discourse, meaning I can now be more open in sharing some of the things that I know about the Cold Start Problem. Things like:

• "Cold Start" is a phrase that was widely used internally at Amazon, going back as far as 2016, and was something that was actively tracked in marketplace metrics.
• There is no single class of Cold Start Problems, and there is no single solution that addresses all classes of problems. Cold Starts exist for branded search (e.g., Dyson vacuum), non-branded/generic search, new categories, and new marketplaces.
• Back in the day, one of the technical factors that gave rise to the Cold Start Problem was latency in data pipelines (i.e., feeding back behavioral data to search logs and processing engines).

FRIENDS OF B@A AMZ Innovate Want to know where the best and brightest minds in the eCommerce space will be gathering to network and learn on October 29-30? I'll give you a hint... 👉 If you want to join me, use promo code "JON100" for $100 off.

BEST from my Inbox

This question hit my inbox over the weekend, and it ​
​
Q: "Hey Jon, should I change my Amazon keyword research and listing optimization strategies because of AI advancements like Rufus, COSMO, Amelia, etc?"

My Answer: Not necessarily.

I'm of the view that the fundamental best practices for keyword research and listing optimization remain unchanged. In a world of Rufus and all the other AI goodies, you should still:

1. Conduct thorough keyword research;
2. Optimize listings with relevant keywords;
3. Focus on customer pain points and benefits; and
4. Utilize high-quality images and videos.

I just can't see a world in which those things cease to matter...

But, understand, that we're not going live exclusively in a world of Lexical Search anymore; we're moving more toward a hybrid world of Lexical and Semantic Search (Lexical Search relies on matching keywords, while Semantic Search focuses on understanding the meaning and intent behind the search).

In this world, listing optimization and SEO becomes an exercise in researching/understanding the right keywords and then weaving them together in ways that cater to both customer needs and Amazon's AI systems.

BEST from YouTube

Anthony Lee is one of the OGs in the Amazon space and, more importantly, just a wonderful human being.

He's also the BIG BRAIN behind the concept of the "Honeymoon Period" for new product launches on Amazon.

(In case you've been living under a rock, the term "Honeymoon Period" refers to the grace period after launching a product where the Amazon algorithm supposedly gave new products a ranking boost.)

But, a lot has changed since Anthony first conceived the idea back in 2015 and, if you ask him directly, he'll tell you that the Honeymoon Period doesn't exist any more.

In his own words:

• "It's about product relevance—Amazon is using AI and knowledge graphs to fill in the gaps and understand what your product is."
• "The algorithm now is all about behavior and data—what happens when they buy, not just when they search."
• "AI is playing a huge role in how Amazon understands relevance now; you need to be data-driven, not strategy-driven."

Anyway, go give this great interview with Anthony a watch.

​

​

Franchises often get 💩ed on as a entry points into entrepreneurship but success rates with franchise ownership is quite high (I've researched this extensively).

Also, it's relatively common to see franchise owners with multiple franchise units/locations.

That's a fact that I never really thought about very deeply.

Until...I read this tweet from Neel.

As he explains, franchises are extremely well-suited toward a rollup strategy because very few things need to change upon acquisition.

Worth a read if you've entertained the idea of franchise ownership.

Neel Parekh @NeelBParekh The other day I had a call with a new search fund entrepreneur. He asked me: “What’s better for a rollup strategy? A franchise or an SMB?” It’s a franchise. Here’s why… Everything I’ve heard from the successful search fund guys is that every acquisition after the first one… https://x.com/i/web/status/1846197281910890731 9:31 AM • Oct 15, 2024 2 Retweets 49 Likes Read 5 replies

​

​​​