When a customer searches for an item on Lyst, Lyst endeavours to return the most relevant matches for our customers. This is achieved in two stages. First, Lyst scans keywords in each product listing related to the search terms and returns an initial set of results that match the intent of the search. This phase is known as query understanding and retrieval. The second phase is ranking.
Once Lyst retrieves all product listings related to a customer’s query, we use the information we have about each product to rank the product so our customers are shown items that they are most likely to appeal to them.
Lyst uses several factors when ranking products in search results. Each factor contributes differently. The most important factors are discussed below:
Relevancy: The relevance score is a measure of how well an item matches a customer search query. Searchable attributes include the brand name, title, description, and tags.
Popularity: For each product, our system will check all impressions and interactions. The more often a product is viewed/tracked/ordered, the more popular it is considered.
Conversion: Uses data - takes displays (product detail page, search & feeds) and tracks, and builds a machine learning model that predicts for a given product its likelihood of a customer converting if it’s displayed.
Value: This gives a score to each product based on the profit Lyst expects to make from the sale. This includes the rate of commission Lyst is paid for the sale of that particular product.
Personalisation: This factor takes into account the brands the user follows (within Lyst), customer behaviour signals; products they’ve interacted with, searches, orders etc.
From time to time we run tests which may see a change in the main criteria used to rank search results and will update this page in the event that there are any material changes to those criteria.