How Liva 7 chooses products for each shopper
A clear, merchant‑friendly explanation of how product relevance is calculated across every page of your store
Overview
Liva 7 selects products using a relevance engine that evaluates shopper behaviour, product relationships, store‑wide rules and real‑time context. Every recommendation is generated dynamically, adapting as the shopper moves through your storefront.
This article explains how product relevance is calculated, how different signals influence the outcome and how the engine balances intent, diversity and merchandising control to choose the best possible items for each visitor.
How relevance is calculated
The engine blends behaviour, product data and rules to rank products for each shopper.
Relevance is determined by combining several inputs at once. The engine evaluates:
- Behavioural signals such as product views, cart actions and search terms.
- Product relationships including similarity, complementarity and frequently bought together patterns.
- Store‑wide rules like pinned or excluded products.
- Real‑time context such as the page the shopper is on and how long they’ve been browsing.
These inputs are weighted differently depending on the situation, producing a ranked list of products that best match the shopper’s intent.
Relevance is a multi‑factor calculation, not a single rule.
How page context shapes recommendations
Each page type signals a different level of intent, and the engine adapts accordingly.
The engine interprets intent differently depending on where the shopper is:
- Homepage: broad interest, discovery, trending items.
- Collection pages: category‑level interest and browsing patterns.
- Product pages: strong intent, so complementary or similar items are prioritised.
- Cart: high‑intent stage where add‑on or value‑boosting items perform best.
This ensures recommendations always feel natural and aligned with the shopper’s journey.
Page type tells the engine what the shopper is trying to do.
How search pages influence product selection
Search pages give the engine the strongest possible intent signal because the shopper has explicitly told you what they want.
When a shopper uses your store’s search bar, the engine treats the query as a high‑confidence indicator of intent. This changes how relevance is calculated and how products are ranked.
The engine evaluates several layers of intent on search pages:
- exact keyword match in titles, tags or descriptions
- semantic relevance, understanding related terms and synonyms
- behavioural reinforcement from previous browsing
- popularity and performance of items related to the query
- fallback logic when the search term is broad or ambiguous
This creates a search experience that feels intelligent, forgiving and helpful, even when the shopper’s query is imperfect.
Search pages use the shopper’s query as the strongest signal of intent.
How behavioural signals influence product selection
Signals help the engine understand what the shopper cares about most.
Different actions carry different levels of intent:
- quick product views show light interest
- longer dwell time signals stronger curiosity
- adding to cart is a high‑intent action
- search terms reveal explicit interest
The engine uses these signals to refine recommendations in real time, adjusting as the shopper interacts with your store.
Stronger signals have a bigger impact on recommendations.
How product relationships are identified
The engine analyses how products relate to each other to build meaningful suggestions.
Relationships are based on:
- similarity in style, category or attributes
- complementarity between items that pair well
- frequently bought together patterns
- shared metadata such as tags or collections
These relationships help the engine recommend items that feel logical and helpful.
Product relationships help the engine understand what belongs together.
How diversity and freshness are maintained
The engine avoids repetitive recommendations by introducing variety.
To keep the experience engaging, the engine:
- rotates products to avoid repetition
- introduces fresh items when appropriate
- balances relevance with novelty
- ensures shoppers don’t see the same items too often
This increases discovery and prevents recommendation fatigue.
Diversity keeps recommendations interesting and effective.
How store‑wide rules shape product choices
Rules act as guardrails that influence what the engine is allowed to show.
Rules can:
- pin products to ensure they appear
- exclude products from all recommendations
- restrict collections to specific widgets or pages
- define fallback behaviour when no strong signals exist
These rules ensure the engine’s decisions always align with your merchandising strategy.
Rules guide the engine without removing its intelligence.
How stock levels and availability affect recommendations
The engine automatically avoids unavailable or low‑priority items.
To protect the shopper experience, the engine:
- removes out‑of‑stock items
- deprioritises low‑stock items
- avoids products marked as hidden or unavailable
- prioritises items that are ready to sell
This ensures recommendations always lead to purchasable products.
Only available products are recommended.
How shopper segments influence product selection
Different segments may see different recommendations based on their behaviour and history.
The engine adapts to segments such as:
- new visitors who benefit from trending or broad‑interest items
- returning customers who may see items related to past behaviour
- high‑intent shoppers who receive more targeted suggestions
- loyal customers who may see premium or personalised picks
This ensures each audience receives the most relevant experience.
Segments help tailor recommendations to different types of shoppers.
How product selection adapts across the shopper journey
Recommendations evolve as the shopper moves from discovery to decision‑making.
The engine adjusts its strategy as the shopper progresses:
- discovery stage: broad, engaging recommendations
- consideration stage: similar or complementary items
- decision stage: high‑value add‑ons or bundles
- checkout stage: subtle, non‑intrusive suggestions
This journey‑based approach ensures recommendations always match the shopper’s intent.
The engine adapts recommendations as the shopper moves through your store.
