Freely selectable
Provider, API key and model for scoring are freely configurable - your existing account is enough.
AI Cross-Selling Manual.
How to set up market-basket analysis with AI scoring, review suggestions and write only sensible assignments to native Shopware cross-selling.
Installation
Up and running in minutes.
Installation and activation happen entirely in the Shopware backend - no code changes required.
1 · Requirements
Requires Shopware 6.7 / PHP 8.2+. As always, create a database backup before larger changes.
2 · From the Store
In the backend, go to Extensions › Store, search for the plugin and install it. Updates then appear automatically in the backend.
3 · Activate
Activate the plugin under Extensions › My extensions, then run bin/console cache:clear or clear the cache from the backend.
4 · Open
The module sits in the backend under Extensions › Stone & Water › AI Cross-Selling.
Configuration
Fine-tune the analysis.
Provider, thresholds and the weighting between statistics and AI are all set in the admin.
Minimum support & confidence
Lower bounds filter statistical noise before the AI scores anything.
Adjustable weighting
Balance the statistical signal against the AI score. Per source product the strongest N suggestions are kept.
Workflow
From suggestion to assignment.
The analysis fills a review pool. You decide what is applied - accepted and rejected pairs are kept permanently.
1 · Analyse
Market-basket analysis plus AI scoring produce suggestions with a rationale (complement vs. substitute, cold-start).
2 · Review
In the pool you accept or reject suggestions. Decisions are never overwritten on future runs.
3 · Write
Accepted pairs go into native Shopware cross-selling - no extra storefront element, full theme compatibility.
CLI & automation
Analysis on demand.
The analysis runs as a scheduled task in the background - or manually via the console, e.g. for initial population.
Run the analysis
bin/console staw:cross-selling:analyze rebuilds the suggestion pool.
Prepare embeddings
bin/console staw:cross-selling:embeddings generates catalog embeddings for cold-start cases.
Scheduled task
In the background the analysis runs via the normal scheduled-task worker and keeps the pool current.
FAQ
Common questions.
Answered briefly.
What about new products without orders?
The AI derives sensible cross-sells from catalog data (name, description, category) - cold-start is covered.
Are existing cross-sells overwritten?
Only accepted suggestions are written. Manual curation and your decisions are preserved.
Complement or substitute?
The AI separates genuine accessories from competing products, so suggestions stay relevant rather than redundant.
Better cross-selling?
Back to Cross-Selling.
Find all details on the product page. We are happy to help with AI setup and strategy.