When buyers sort by “Relevance” (our default sort), we use our in-house algorithm to determine the ordering of their category pages and search results. This algorithm scores all resources on our site daily, to determine how they should be ranked. This score is also personalized to individual users based on their activity, to show content that may be more relevant to them specifically.
This article outlines some of the factors we use in scoring. Some details are subject to change and some factors may be excluded to prevent anyone from attempting to artificially boost their scores or put other creators at a disadvantage. Anything not documented here, exact score weights, exact timeframes, etc, are not public knowledge and will not be disclosed.
These factors influence a resource’s score for all users before personalization is applied, if any. This means that when viewing our site as a guest, without personalization, resources will be ordered solely according to these global score factors.
The most important and heavily weighted score factor for determining a resource’s ordering is:
Average revenue generated per impression
We look at how much in revenue has been generated by a resource recently, divided by how many impressions it took to generate that much revenue. Purchases generated without an associated discovery impression (for instance, a referred purchase) are not considered towards this revenue total.
Impressions are limited, and our job is to maximize revenue for our creators. Thus it is important that we are directing impressions towards the resources that will generate the most amount of revenue from them.
Resources that waste a lot of impressions, not converting them to purchases, will score much lower than resources that convert well, even if the resource that converts well doesn’t generate as much revenue or as many purchases, although it is likely that the one that converts better will also do well in those metrics.
Because interest in resources fluctuates, we do not look at “lifelong” generated revenue or impressions, only how the resource has performed recently. This also ensures that “seasonal” resources, for instance, Christmas/Halloween products, are able to return to the front pages as soon as interest in them resurfaces, and then allows them to make room for other resources once interest in them dies down again.
Most important is to have a good resource that buyers actually want to buy. If we show your resource to our users and they aren’t buying it, you won’t rank well.
You can improve your odds of securing successful purchases by focusing on improving your conversions from impressions to purchases. Your product analytics pages have a full breakdown on the conversion funnel from: Impression → page view → add to cart → purchase.
Some things you might consider:
Other global score factors include: