How Findify Machine Learning works

Findify is not just a search engine, we have built several algorithms that are based on the data we collect to help your customers find and buy exactly what they need thus increasing your store conversion.

We have three layers of algorithms that are described below (please note, Personalization might not be available on your plan). Each layer is applied on top of the previous one to tailor the results to users and increase your sales.

Basic Search

This is the baseline of the results that our platform returns. By analyzing product data that we retrieve from your store (title, description, tags, other custom fields) the system extracts results that are relevant to the customer's query. Non-relevant results are removed completely (e.g. if you search for Shorts, Jackets will be removed from results) and more relevant results are ranked higher (e.g. usually title field is much more relevant than description). The weights for each field can be adjusted via the Merchant Dashboard.

Trend Scoring Model

Trend Scoring Model analyzes performance of each product throughout the whole store by taking into account clicks, purchases, page views, searches and other data that we collect about the product. We also collect product popularity for popular search queries and collections which can further influence the end result. As a result of this calculation we generate the popularity score which is merged with Basic Search score to re-rank the products that the customer sees.

Personalization Model

Personalization Model takes the product ranking to a completely new level. The system analyzes and stores actions of each and every customer and optimizes the results to match the customer's preferences (e.g. if the customer is interested in red clothes, the red color will be promoted in different searches and collections). By default the algorithm is optimizing towards maximizing user conversion, but work is being conducted to including additional optimization goals.

Due to the way personalization works, it will override the Basic Search ranking at times to maximize relevance (e.g. a product with a higher Basic Search ranking might appear lower in the results due to personalization effect), however product that are not relevant to the search query will still not appear in the results.