Ecommerce ops · Production

How Shopify uses recommender systems to personalize app, theme, and expert recommendations for merchants

The problem

As Shopify's feature set grew to serve hundreds of thousands of merchants across many channels, it became difficult for individual merchants to filter what was relevant to their specific business needs.

Workflow diagram · grounded in source
1
Collect implicit interaction signals
trigger
“user-item interactions, derived from implicit signals like the user's past purchases, installations, clicks, views, and so on”
2
Build user preference training matrix
ai_action
“We first create a training matrix of all user-item interactions by stacking users' preference vectors. Each row in this matrix serves as an individual training example.”
3
Train item-item similarity model
ai_action
“Linear methods like LRec and its variations solve this optimization problem by directly learning an item-item similarity matrix”
4
Score and rank candidate items
ai_action
“we take the items that the user has not yet interacted with and sort their predicted scores (in red). The top scored items are then the most relevant items for the user”
5
Surface personalized recommendations
output
“We applied this approach to provide recommendations to our merchants in a variety of contexts across Shopify”
Reported outcome

Merchants receiving personalized recommendations saw a 50% higher app install rate, were up to 12% more likely to find their home feed useful, and were over 10% more likely to launch their online store, with increased collaboration in the Expert Marketplace.

Reported metrics
App install rate50% higher
Home feed usefulness ratingup to 12% more likely
Online store launch rateover 10% more likely
Expert marketplace response ratehigher response rate
Show all 5 reported metrics
app install rate50% higher
home feed usefulness ratingup to 12% more likely
online store launch rateover 10% more likely
expert marketplace response ratehigher response rate
merchant-expert collaborationincreased collaboration
Reported stack
Collaborative FilteringLRec
Source
https://shopify.engineering/how-shopify-uses-recommender-systems-to-empower-entrepreneurs
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Merchants receiving personalized recommendations saw a 50% higher app install rate, were up to 12% more likely to find their home feed useful, and were over 10% more likely to launch their online store, with increased…

What tools did this team use?

Collaborative Filtering, LRec.

What results were reported?

App install rate: 50% higher; Home feed usefulness rating: up to 12% more likely; Online store launch rate: over 10% more likely; Expert marketplace response rate: higher response rate (source-reported, not independently verified).

How is this ecommerce ops AI workflow structured?

Collect implicit interaction signals → Build user preference training matrix → Train item-item similarity model → Score and rank candidate items → Surface personalized recommendations.

Related ecommerce ops cases