First up, the Query Category Score measures the significance of a search query to a product category.
In simpler terms, it answers the question: “For this category, how important is this search query?”
Of course, “important” is a vague term. And Amazon doesn’t do vague.
So to determine the importance of the search query, the A9 examines the clicks and purchases for products within the category. More clicks and more purchases = a more important query.
However, not all queries are created equal. Some are less common, and these elusive queries don’t provide enough data for the A9 algorithm to accurately calculate the score.
In fact, most search queries on Amazon fall within this bucket, as we’ll see later.
So what does the A9 do when it doesn’t have the data? In these cases, it has to roll up its sleeves and get a bit more creative.
Suppose the query in question is “dog grooming brush.”
For these low-volume queries, Amazon looks at the number of clicks for the:
If this sounds a bit confusing, don’t worry. Let’s break it down with an example:
Imagine you own an online clothing store. Naturally, you want to understand the importance of different search queries related to clothing.
By doing so, you can prioritise your efforts to enhance your website’s search functionality. Or in more concrete terms – you can go where the buyers are.
You turn to the Query Category Score to gauge the importance of each search query for your clothing category. Remember, this score is calculated by analysing the number of clicks and purchases for clothing products when customers search for specific queries.
For instance, you discover that the search query “men’s casual shirts” is a rockstar, boasting a high Query Category Score.
This is because there were many clicks and purchases when this query was used.
However, a less common query like “green men’s dress shirt size medium” is more of a wallflower, with a lower Query Category Score due to fewer clicks and purchases. And the lower volume can make it difficult to gauge exactly what’s going on with this search term.
To delve deeper into these less common queries, you examine the number of clicks for the individual words, bigrams, and trigrams in the query. Crucially, this helps you understand why it has a lower score – and what you can do about it.
Let’s illustrate this with an example of analysing query data using 2-word and 3-word phrases:
You start looking at the searches made by users on your site. Some sample queries might include:
“red sneakers”
“black leather boots”
“women’s clothing sale”
To crack the code of why certain queries have lower scores, you examine the number of clicks for the individual words, bigrams, and trigrams in the query.
For the query “red sneakers”:
Individual words: “red”, “sneakers”
Bigrams: “red sneakers”
Trigrams: None
For the query “black leather boots”:
Individual words: “black”, “leather”, “boots”
Bigrams: “black leather”, “leather boots”
Trigrams: “black leather boots”
For the query “women’s clothing sale”:
Individual words: “women’s”, “clothing”, “sale”
Bigrams: “women’s clothing”, “clothing sale”
Trigrams: “women’s clothing sale”
By counting the number of clicks for these individual words, bigrams, and trigrams, you can better understand the context of these queries and identify patterns that may be contributing to lower scores.
For example, you might discover that queries containing the bigram “women’s clothing” tend to have lower scores. This could indicate an issue with the results being returned for these types of searches.
Looking for a Better Agency?
Are you a 7 or 8-figure Amazon seller who is…
Next up, we’ve got the hunger score.
This is a bit of a counter-intuitive one. But stay with it, because it shines a bright light on some of the more poorly understood aspects of the A9 Algorithm and how Amazon Ranking Models work.
The Hunger Score on Amazon is a unique metric that gauges the “eagerness” of a product category to be selected or clicked on when a specific search query is made.
A “hungry” product category is like the kid at the back of the class, waving their hand frantically to get the teacher’s attention. The longer a category goes without being selected, the hungrier it becomes. And if a category has a high Query Category Score, it gets hungry faster.
The overarching purpose is to show a balanced range of categories for a given Amazon product search query, over time. Sounds simple, but it’s a surprisingly difficult task.
To make this concept more digestible (wording chosen carefully), let’s look at some real-world examples of how the Amazon Hunger Score operates:
Before we move on, one more note on the Hunger Score vs Honeymoon Period. Because as it turns out, it’s one of the keys to unfurling a great long-standing Amazon myth.
The idea of the Honeymoon Period—the theory that an ASIN gets more limelight in the 3 months or so following launch—is often not even questioned, such is its popularity.
But as the Hunger Score shows, the truth is that the limelight is granted much more randomly. And the result is that the Honeymoon Period is actually random and not fixed.
Your ASIN might appear to be enjoying a boost, but it can most likely be explained by the ebbs and flows of Amazon’s own semi-random shuffling of the SERPs.
Now onto our third and final Product Search Score.
Thankfully, this might be the easiest one to get your head around.
The In-Category Relevance Score gauges the relevance of each product to the user’s search query, within a specific category.
It’s like a matchmaker, assessing how well each product aligns with the shopper’s needs by considering its features, attributes, and other pertinent information.
In layman’s terms, it simply answers the question, “How well does this product match the shopper’s search query?”
It’s not about popularity or sales, per se. It’s about how well the product fits the search term and the underlying buyer intent. Of course, sales volume—among many other factors—often ends up being a reliable indicator of relevance.
Anti-newsflash: The higher the In-Category Relevance Score, the more relevant the product is considered. It’s a straightforward correlation – the better the fit, the higher the score.
You can think about the In-Category Relevance Score this way: If products weren’t sorted by relevance, the shopper would have to manually sift through each option, weigh up their features and benefits, and laboriously decide which one to opt for.
So again, the point of the exercise comes back to making the shopper’s life easier.
So, how does this all come together?
Please see; The Behavioural Feautures of A9 for in-depth analysis.
Now, let’s consider a practical example – a user searching for “red sneakers”:
Query Category Score: The team will analyze the number of clicks for the individual words in the query – “red” and “sneakers” – and the bigram “red sneakers.” This data helps understand the context of the search and identify any patterns that may be contributing to lower scores.
Hunger Score: The algo will also assess the Hunger Score for the category of sneakers. If the category hasn’t been selected for a while, its Hunger Score will be high, giving the category prominence in the search results.
In-Category Relevance Score: Lastly, the team will consider the In-Category Relevance Score for each product within the category of sneakers. Products with a high In-Category Relevance Score will be deemed the most relevant and useful within the category, factoring in aspects such as popularity, brand, and customer ratings.
By integrating all three components, the search results will prioritize products that are both relevant to the user’s search query and eager to be selected.
This approach simplifies the user’s search, making it easier for them to find what they’re looking for.
So far, so familiar. But here’s where it gets interesting.
The A9 algorithm also includes a built-in override process called the cold start, as explained in Dr. Daria Sorokina’s presentation ‘The Joy of Ranking Products’ – a prime example (pun fully intended) of Amazon’s own scientific literature.
To cut a long story short: We all know that products take time to rank. But if a popular product (based on the data) is still not ranking sufficiently by the 7th day, the A9 team can and will step in to adjust the ranking of a product manually.
Shrewd observers will note that this confirms an oft-suspected fact: As the A9 incorporates both automated ranking and manual adjustments, some degree of favouritism is possible in search.
But sadly, it’s not likely to turn into a workable “back door” any time soon. Making manual adjustments for every product is orders of magnitude away from being feasible – so stick to getting on the good side of the algorithm.
The action points
So, what does all of this mean for your strategy?
It means that your primary focus should be on the extreme relevancy of your keywords and your target category placement.
Seems logical enough. And as for how to achieve that?
Well don’t go anywhere just yet, because I’ll shortly be covering a relevance pointing system to give you a better opportunity to ensure that all your keywords—if relevant—index as expected.
This approach will help prevent Amazon from moving your products to different categories on the back end without your consent.
It’s a valuable weapon in making sure that you get it right the first time – rather than resorting to short-term hacks that can cause long-standing issues.
To lay the groundwork for this strategy, let’s now take a deeper look at the inner workings of those Amazon ranking algorithm.
Just what’s really going on behind the scenes?
