Smart Cross-Sell — Recommendations from Real Order History

About the Extension

Recommendations grounded in your real orders

Kommora Smart Cross-Sell computes cross-selling suggestions from your actual order history — instead of static rule sets or an external recommendation service. The plugin analyses nightly which products are actually bought together, weighs the pairs by confidence score, and renders the most relevant suggestions at four strategic positions in the shop.

Why this plugin?

• Revenue through relevant suggestions. Cross-selling widgets grounded in real buying patterns convert measurably better than static rule sets. You show what your customers actually buy together — not what you manually paired.
• No external service, no monthly SaaS cost. The calculation runs nightly inside your own Shopware instance. No third-party token, no data transfer, no dependency on an external API.
• Four positions, one concept. Product page, cart, wishlist and thank-you page — everywhere your customers are already thinking about further products.
• GDPR-compliant & data-minimising. All data stays in your Shopware database. Impression and click tracking without external analytics.

Features at a glance

Four widget positions

• Product page — "Customers also bought …". Renders below the buy widget on every detail page, robust even on 6.7 CMS layouts.
• Cart — "Goes well with …". Takes the current cart products, suggests complements.
• Thank-you page — "You might also like …". Uses post-purchase attention for the next order.
• Wishlist — top-up recommendations. Particularly valuable on an empty wishlist ("popular in the shop") and reduces bounce.

Data foundation — not static, but learning

• Co-occurrence aggregator runs daily and recomputes product pairs from real order volume
• Confidence score — product B is only recommended when a significant share of A-buyers also bought B (threshold configurable, default 5 %)
• Minimum co-occurrence — weak signals with only one shared order are not shown (default 2)
• Look-back window — default 365 days, shorter values sensible for seasonal assortments
• Eligible order states as a comma list — default `paid,paid_partially,completed,in_progress,open` (the last is important for B2B shops with long payment cycles)

Manual curations

• Pairs set by the shop operator (`isManualCuration = true`) override the automatic calculation
• Ideal for launch products (no history yet), seasonal sets, cross-selling campaigns with marketing priority
• Manual pairs survive the nightly recompute

Best-seller fallback

• When too few history-based suggestions exist for a product (typical for new items), the slider fills up with the top-selling products to reach the configured limit
• Dedicated window for the best-seller cache (default 90 days)
• Sales-channel-specific — B2C and B2B best-sellers are computed separately

Impression and click tracking

• Dedicated table `kommora_cross_sell_impression` with session hash, widget position, source product, target product
• On click on a recommended product: click counter + attribution marker in the cookie
• On order completion: `CheckoutOrderPlacedEvent` checks the last impressions/clicks and writes the revenue attribution back
• Reporting via the DAL repository for your own dashboards or BI systems

Configurable per sales channel

• Look-back window (default 365 days)
• Minimum co-occurrence (default 2) and minimum confidence (default 0.05 = 5 %)
• Eligible order states (comma list)
• Per-position widget limit (default 4)
• Widget on/off per position (product page, cart, wishlist, thank-you)
• Best-seller fallback on/off with its own window
• Confidence threshold overridable via sales channel configuration

Loyalty integration (prepared)

• Interface for optional bundle bonus in combination with Kommora Loyalty
• If a customer actually buys a recommended pair together, they can earn extra loyalty points
• Cart processor implementation lands in V1.1

Ideal for

• Shops with at least a few hundred orders in the look-back window — more history means more precise suggestions
• Assortments with natural product combinations (accessories, consumables, collections, sets)
• Merchants who want to replace static cross-selling rules with data-driven suggestions
• B2C and B2B shops with recurring customers and clear buying patterns
• Shop operators who want cross-selling to be measurable (impression → click → revenue attribution)