Enterprise AI product discovery for large Shopify catalogues
Large catalogues create choice faster than they create discovery. Shoppers still encounter a fraction of the range, while merchandising teams face an impossible classification problem. RecoMelon adds visual product intelligence, natural-language discovery and merchandising controls across search, collections, PDPs and baskets - helping more of the catalogue become relevant without manually classifying every product.
Search that understands what shoppers mean. Conversational and contextual search interprets style, colour, attributes and intent. Dynamic merchandising balances commercial and catalogue priorities, while inline Shop Similar and integrated LiveGrid™ UGC keep shoppers discovering without returning to a static results grid.
Shop Similar, directly from the product image. When a product is "almost right", Reveal™ offers visually aligned alternatives from within the primary product image, above the fold. Shoppers stay on the PDP and keep their intent while finding an alternative without the attribute stopping the purchase.
Turn “almost right” into another relevant option. Shop Similar keeps high-intent shoppers exploring with visually matched alternatives based on the product they are already considering - helping them refine their choice without starting their search again.
Turn one product into a complete look. Complementary recommendations create multiple looks around a chosen product, with configurable discounts and multi-buy offers to increase AOV. Bundle Rescue swaps unavailable or unsuitable items for visually similar alternatives without breaking the set. Integrated LiveGrid™ content helps shoppers visualise the combination.
Abandoned Cart Email Rescue. Sending shoppers back to rejected products repeats the problem. SecondLook™ restores the abandoned basket and offers visually matched alternatives for each item - helping shoppers find a better fit while preserving basket intent and cross-sell opportunities.
Proven at catalogue scale
DecoratorsBest gives the large-catalogue proposition a real-world benchmark. Across a 230,000-SKU home décor catalogue, the published comparison period reported approximately 5× more recommendation clicks and 4× more recommendation-influenced orders.
The case study covers 16 March–15 April 2026 and is customer-specific evidence rather than a forecast for another catalogue.
Catalogue scale
Large catalogues create a different discovery problem at each stage of scale.
Growing catalogues
10,000
SKUs
Make the long tail visible. Connect overlooked products to shopper intent without relying on complete manual tagging or purchase history.
Large catalogues
100,000
SKUs
Turn depth into navigable choice. Combine visual similarity, natural-language search, collection merchandising and in-session intent across brands and ranges.
Enterprise catalogue planning
500,000+
SKUs
Treat discovery as infrastructure. Scope catalogue ingestion, variants, refreshes, regions, storefront performance and governance around the way your organisation actually operates. Final architecture and capacity are validated against your catalogue and deployment requirements.
Why RecoMelon
Behaviour and purchase data become more useful once products have already been seen. That creates a structural disadvantage for new arrivals and long-tail inventory.
RecoMelon can begin with the product itself. Product imagery and catalogue context establish visual intelligence and relationships; in-session behaviour and commercial merchandising signals then add context as shoppers interact.
Product imagery
Understand visible product characteristics and visual relationships.
Catalogue context
Add price, stock, variants, categories and market availability.
Shopper intent
Use the current session to refine what is relevant now.
Merchandising signals
Apply commercial priorities, exclusions and controls.
Discovery
Put that intelligence to work across search, collections, PDPs and baskets.
This means new arrivals, long-tail inventory and visually related products can participate in discovery without first needing a history of popularity.
Product intelligence
IntentWeave™ analyses product imagery and turns visible characteristics into structured product intelligence - including silhouette, neckline, sleeve, colour family, pattern and texture. RecoMelon uses those signals across search, recommendations, collections and shopper refinement.
This reduces dependence on perfectly maintained tags and metafields across very large ranges. Your catalogue remains the source of truth for commercial facts such as price, stock, variants and market availability; RecoMelon adds a visual understanding layer that would be difficult to create and maintain manually at enterprise scale.
New and long-tail products
Behaviour-led recommendation systems are strongest where products already have data. That can leave new arrivals, seasonal launches and long-tail inventory waiting for enough clicks or purchases before they become discoverable.
ColdStart™ establishes product relationships from imagery and catalogue context, allowing new and low-history SKUs to enter relevant discovery as soon as their product intelligence is available. Behaviour can then enrich those relationships rather than being a prerequisite for them.
For enterprise catalogues with frequent launches, this creates a practical route to giving more of the range a chance to be found from day one.
Across the shopping journey
Enterprise product discovery is rarely a single search box. RecoMelon can be introduced at the surfaces where catalogue depth creates the most friction, while keeping product understanding consistent across the journey.
The Index™ interprets natural-language requests and catalogue attributes. Smart Collection Pages combine visual discovery with merchandising logic for category and campaign browsing.
Shop Similar and Reveal™ give almost-right shoppers relevant alternatives without forcing them back to search. Dealbreaker™ Filters refine the attributes causing hesitation while preserving the shopper’s original direction.
Bundles™ creates complementary sets and configurable multi-buy offers. Merchandising controls let teams shape discovery around commercial priorities rather than surrendering every decision to an algorithm.
SecondLook™ restores abandoned basket context with visually matched alternatives. RescueGrid™ continues discovery when a product or URL becomes unavailable, preserving intent instead of creating a dead end.
Merchandising control
AI should reduce manual merchandising — not remove merchandiser control.
RecoMelon can generate product intelligence and relationships that would be impractical to maintain manually across 100,000+ SKUs, while merchandising teams retain control over what should be promoted, excluded, prioritised or constrained.
Build product understanding from imagery, catalogue context and in-session behaviour. Put those relationships to work through Smart Collections and The Index™, without manually pairing and classifying every product.
Product attributes · Visual relationships · Shopper intent
Use MerchAlchemy™, boostbias and inclusion/exclusion controls to shape discovery around your commercial priorities and define the boundaries within which recommendations operate.
Promote · Exclude · Prioritise · Constrain
Availability · Margin · Campaign priority · Newness · Market relevance
Bring commercial signals alongside visual and behavioural intelligence. Confirm the source and application of each signal as part of your catalogue and deployment setup.
You don’t have to choose between AI automation and merchandising control.
Regional relevance
Enterprise storefronts differ by assortment, availability, currency, URLs, language and merchandising priorities. RecoMelon Markets brings that regional context into discovery while keeping a consistent product-intelligence layer underneath.
For multi-market Shopify and Shopify Plus estates, validate the products that can be sold in each market, local product URLs, currency, translated discovery labels and market-specific merchandising before rollout. LinguaLoom™ supports language and local terminology alongside Markets.
The aim is not simply to translate one global experience - it is to keep discovery commercially correct for the storefront the shopper is actually using.
Read about regional catalogue handling
Catalogue operations
At enterprise scale, catalogue freshness is part of relevance. New products arrive, imagery changes, stock moves, prices update and ranges are retired. Discovery needs an operating model for those changes, not a one-off import.
During implementation, review catalogue feeds, update frequency, variants, regional feeds and the events that should trigger discovery updates. Confirm how new products, changed imagery, stock and commercial availability remain aligned with the storefront.
Test new arrivals, unavailable products, discontinued URLs and changed market availability. For major catalogue updates, validate representative categories and discovery surfaces before wider rollout. Refresh cadence and integration requirements are agreed around your architecture and operational needs.
Deployment
Deploy through Shopify 2.0 blocks, sections or tailored theme integrations. Start on representative templates, validate presentation and analytics, then expand placements without making a full search migration the prerequisite for launch.
RecoMelon can be scoped around the catalogue systems already feeding your storefront - including Shopify, PIM or other product-data sources - rather than requiring RecoMelon to become the commercial source of truth. Confirm feeds, ownership and update requirements during implementation.
API-led delivery can bring RecoMelon intelligence into your existing presentation layer. Scope discovery surfaces, catalogue and market structure, analytics, performance expectations, caching and release ownership with the implementation team.
Start with a category, market or high-value discovery surface, establish measurement and operational ownership, then extend. Enterprise deployment should define who controls merchandising, who owns catalogue quality, how changes are tested and how performance is measured - not just how a widget is installed.
Measurement
Start with a defined category, market or discovery surface and measure RecoMelon against the commercial behaviour that matters. Establish the baseline first, then expand only where the data supports it.
Are shoppers exploring more relevant products and going deeper into the catalogue?
Does interacting with RecoMelon increase the likelihood of purchase?
Are discovery and complementary recommendations increasing basket value?
Is more of the long tail entering meaningful shopper journeys?
Use RecoMelon reporting alongside your existing analytics and agreed attribution approach. The objective is not to replace your measurement stack - it is to make the commercial effect of discovery visible enough to decide whether to expand.
Evidence at scale
Follow the DecoratorsBest story: from a vast home décor catalogue to visually guided choices and measurable shopper engagement.
01 / The catalogue
230,000 SKUs
Wallpaper, fabric and upholstery give shoppers extraordinary choice. The challenge is finding relevant alternatives within that depth.
02 / Visual discovery
Pattern. Motif.
Colour. Texture.
RecoMelon mapped product imagery across the catalogue, connecting visually coherent alternatives without manually tagging every SKU.
03 / Shopper engagement
≈5× recommendation clicks
DecoratorsBest reported approximately five times as many recommendation clicks versus the prior period.
04 / Influenced orders
≈4× recommendation-influenced orders
Recommendation-influenced orders grew approximately fourfold versus the prior period.
The published case study reports the period 16 March–15 April 2026. These are customer-specific results, rather than a forecast for another catalogue.
No. RecoMelon can be introduced alongside your existing search and merchandising stack. Start with the gaps that are hardest to solve today - visual similarity, new and long-tail product discovery, PDP alternatives, smart collections or complementary bundles - while your current search remains in place.
For enterprise teams already using Algolia, Constructor, Nosto, Klevu or a custom search layer, this creates a lower-risk path to evaluation. RecoMelon can complement the existing retrieval layer first; The Index™ can then take a larger role only where testing shows a better shopper or commercial outcome.
Implementation is scoped around catalogue feeds, storefront architecture, analytics, markets and ownership of each discovery surface, rather than assuming a rip-and-replace migration.
Yes. RecoMelon can be introduced alongside an Algolia-powered Shopify storefront. Keep your existing search in place while evaluating visual recommendations, PDP alternatives, Smart Collections or Bundles on selected surfaces. Scope catalogue feeds, storefront integration, analytics and ownership of each surface before deciding whether The Index™ should take on a larger role. This does not assume a native Algolia connector.
Yes. RecoMelon can be evaluated alongside an existing Nosto setup on a Shopify storefront. Agree which platform owns each search, recommendation or merchandising placement, then test RecoMelon on the surfaces where it adds value. Coordinate catalogue data, market context and measurement so overlapping widgets or rules do not create conflicting experiences. A native Nosto connector is not assumed.
RecoMelon can complement Shopify Search & Discovery, and The Index™ can take on search and collection discovery where the deployment supports it. Replacement should follow a feature and relevance review: compare search behaviour, filters, synonyms, merchandising rules, regional handling and analytics with your current setup. Validate the required journeys before retiring existing functionality; this is not a blanket promise of feature-for-feature parity.
RecoMelon uses product imagery and catalogue context to create attributes and relationships across large ranges, with behavioural signals adding context as shoppers interact. New and long-tail products can become discoverable without waiting for extensive purchase history. DecoratorsBest’s published case study covers 230,000 SKUs. For your catalogue, scope ingestion, variants, refresh frequency, regional availability and storefront performance before rollout.
Yes. Shopify and Shopify Plus deployments can use Shopify 2.0 blocks, sections or tailored theme integrations. Headless storefronts can use API-led delivery within their existing presentation layer. Confirm the discovery surfaces, catalogue and market structure, caching, analytics, performance expectations and release ownership with the implementation team; headless deployment is scoped to the storefront architecture.
Yes. MerchAlchemy™, boostbias and inclusion/exclusion controls let teams shape what is promoted, excluded, prioritised or constrained while RecoMelon develops the underlying product intelligence. Smart Collections and The Index™ bring that intelligence into discovery. Define commercial priorities such as availability, margin, campaign priority, newness and market relevance, then confirm which catalogue signals and controls support the intended outcome.
No manual classification programme is required to get started. IntentWeave™ can derive visual product intelligence from product imagery and catalogue context, reducing dependence on perfectly maintained tags and metafields across large ranges. Existing catalogue data still supplies commercial facts such as price, stock, variants and market availability, and can be used alongside RecoMelon’s visual intelligence.
Yes. A phased rollout is often the lowest-risk way to evaluate RecoMelon. Start with a representative category, market or high-value discovery surface, establish the baseline and measurement approach, then expand only where the shopper and commercial data support it.
Share your SKU count, priority categories, markets and existing search or merchandising stack. We’ll map where RecoMelon can add value, what can stay in place and the lowest-risk surface to test first.
See RecoMelon on your catalogue