+15.5%

Increase in Search Revenue per Visitor

+20.35%

Increase in Search Click-Through Rate

Industry:Electronics Marketplace

Business Unit:E-Commerce

Use Case:Marketplace Search & Product Discovery

Channel:Website

Key Objectives:Improve Search Relevance, Reduce Engineering Overhead

Products:

Consumer ElectronicsHome & LivingFurniture

How Kogan Unlocked $10.1M Incremental Revenue with Marqo

Company Overview

Kogan.com is one of Australia's largest online retailers, offering electronics, appliances, furniture, and more. Founded in 2006 and headquartered in Melbourne, the company operates a multi-brand e-commerce platform serving customers across Australia and New Zealand.

As online retail competition intensified from global giants and emerging players, Kogan saw search as a key differentiator — using relevance and product discovery to win shoppers in a crowded marketplace. With increasingly complex shopper intent across its multi-brand catalog, the team needed a platform that could better interpret context, understand marketplace-specific language, and adapt relevance dynamically in real time.

The objective was to move beyond incremental tuning and establish a scalable, performance-driven foundation for search and product discovery across over 16 million products.

Why Kogan Evaluated Marqo vs. Constructor

Kogan evaluated Marqo after reaching the limits of Constructor's ability to handle catalog complexity at scale. With over 16 million products across first-party inventory and third-party marketplace listings, static ranking models and manual tuning could not keep pace with how Kogan's customers searched.

Marqo's catalog-specific LLM training and real-time behavioral learning offered a path to relevance that improves continuously — without requiring a dedicated merchandising team to manually manage rules across multiple brands and categories.

Challenges

Kogan's multi-brand catalog introduces unique search challenges. Shoppers search across electronics, appliances, furniture, and everyday essentials with different intent patterns — and the same keyword can mean very different things across product categories.

Catalog Scale and Complexity — Managing over 16 million products across first-party inventory and third-party marketplace listings created fundamental search relevance challenges at scale.

Inconsistent Product Data — Inconsistent product data quality and keyword-stuffed listings buried relevant results, frustrating shoppers and reducing conversion.

Behavioral Signal Adaptation — Slow adaptation to real shopper behavior meant rankings did not reflect current demand signals across a rapidly changing catalog.

Freshness and Indexing — Indexing delays from constant price and stock updates degraded search freshness and led to stale results.

With 16 million products across multiple brands, delivering accurate, relevant results requires a system that understands how shoppers actually search — not just a list of keywords to match.

Solution

Kogan deployed Marqo's pixel across its multi-brand catalog to continuously capture and ingest shopper interaction signals. A custom LLM was trained on Kogan's specific product data and catalog structure, enabling relevance that reflects how Kogan's customers actually search — across electronics, appliances, furniture, and general merchandise.

Marqo's behavioral learning meant rankings improved automatically over time, without requiring the Kogan team to manually maintain rules or configurations across more than 16 million products.

Behavioral Signal Capture: Marqo's pixel installed across Kogan's websites to continuously capture shopper interaction signals including search queries, clicks, add-to-cart events, purchases, and browsing behavior.
Catalog-Specific LLM Training: A custom LLM trained on Kogan's unique product attributes and multi-brand structure — no generic models or manual tuning required.
Descriptive Query Handling: Stronger handling of descriptive, high-intent queries with results better aligned to how customers searched across a broad catalog.
Business Logic Integration: Improved ability to influence rankings using business logic and proprietary signals, without heavy engineering effort.

Results

Kogan evaluated Marqo using a controlled production A/B test on-site, with 50% of traffic assigned to Marqo and 50% to the existing search setup. Kogan reported significant improvements in search relevance and discovery quality. Marqo's real-time learning from shopper behavior delivered results that better matched customer intent across their entire catalog.

+15.5%

Increase in Search Revenue per Visitor

+20.35%

Increase in Search Click-Through Rate

What This Means

Kogan reported significant improvements in search relevance and discovery quality. Marqo's real-time learning from shopper behavior delivered results that better matched customer intent across their entire catalog — turning search into a measurable revenue driver for one of Australia's largest online retailers.

Descriptive queries across 16M products now return accurate, relevant results.
+20.35% search click-through rate — without manual tuning or rule management.
$10.1M in incremental revenue driven by smarter, continuously improving discovery.

Results reflect a controlled A/B test conducted on Kogan's production environment with 50% of traffic allocated to Marqo. Outcomes may vary depending on catalog size, traffic volume, and implementation configuration.

Moving Forward

What Comes Next for Kogan

With search now validated as a meaningful revenue driver, the Kogan team is evaluating opportunities to extend Marqo beyond search and into the broader discovery experience. Marqo is now a key part of Kogan's discovery stack, enabling search and ranking that adapts to shopper intent and onsite behavior.