How Kogan Drove $10.1M in Incremental Revenue with Marqo

How Kogan Drove $10.1M in Incremental Revenue with Marqo

Kogan.com is one of Australia's largest online retailers. With AUD $809M in annual gross sales, over 16 million products, and a catalog spanning consumer electronics, appliances, home and garden, furniture, fashion, health and beauty, toys, and sports, Kogan operates at a scale that breaks most search platforms.

This is the story of how Marqo delivered $10.1M in incremental revenue for Kogan and why large, complex catalogs are exactly where Commerce Superintelligence performs best.

The Scale Challenge

Kogan is not a single-category retailer. It operates a hybrid model: a mix of first-party inventory across over 20 private-label brands and third-party marketplace products from thousands of sellers. The catalog spans nearly every consumer category, from laptops and televisions to mattresses, outdoor furniture, skincare, and children's toys.

This kind of catalog creates a search problem that most platforms cannot solve:

• Massive vocabulary diversity. The word "case" means something completely different in electronics (phone case), furniture (display case), and beauty (makeup case). A search engine needs to understand which meaning applies based on context, not just keyword matching.
• Constant catalog churn. With thousands of marketplace sellers adding and removing products daily, any system that depends on accumulated behavioral data to rank products faces a perpetual cold-start problem. New products arrive with zero click history.
• Cross-category intent. A shopper looking for a product that spans multiple categories needs the search engine to understand intent across the full catalog, not just match within a single taxonomy.
• Long-tail queries. The majority of search queries on a catalog this large are low-frequency, specific queries that no merchandiser has time to write rules for. These long-tail queries collectively represent significant revenue that most rule-based systems miss entirely.

This is the exact problem that Commerce Superintelligence was built to solve.

Why Traditional Search Fails at This Scale

Traditional ecommerce search platforms handle scale in one of two ways, both of which break down on catalogs like Kogan's:

Keyword-based systems rely on exact matches and manual synonym lists. They work for head queries ("Samsung TV") but fail on intent-driven queries ("something to watch movies on in the bedroom"). At Kogan's scale, maintaining synonym lists across every product category is operationally impossible.

Behavioral-only systems learn from clicks and conversions to improve ranking over time. They work well for popular products with rich interaction history but fail for new products, new sellers, and low-traffic queries. On a marketplace with constant catalog churn, a significant portion of the catalog is always in cold-start territory.

Neither approach understands products. They match patterns or learn from behavior, but they do not comprehend what a product actually is, what it looks like, or how it relates to other products.

What Marqo Did Differently

Marqo trained a dedicated AI on Kogan's entire catalog. This means:

Every product understood from day one. When a new seller lists a product on Kogan's marketplace, Marqo's model understands it immediately from the product title, description, images, and attributes. No waiting for clicks. No cold-start period. The product ranks on merit from the moment it enters the catalog.

Cross-category intelligence. Because the model is trained on the full catalog, it understands relationships across categories. It knows that a shopper searching for "home office setup" might want a desk, a monitor, a chair, and a desk lamp, and it can surface products from multiple categories in a single result set.

Vocabulary calibrated to Kogan's catalog. The dedicated AI learned Kogan's specific product vocabulary, brand relationships, and category structure. It does not confuse Kogan's private-label brands with generic terms. It understands the specific meaning of attributes in the context of Kogan's catalog.

Long-tail coverage. Because the model understands products semantically, it handles long-tail queries that no merchandiser would ever write rules for. When a shopper types "waterproof bluetooth speaker for the shower," the model understands every word and surfaces relevant products, even if no product in the catalog contains that exact phrase.

The Results

The result: $10.1M in incremental revenue from search alone.

This was not a percentage improvement on a small base. This was $10.1M in new revenue for a retailer already doing AUD $809M in annual gross sales. It came from better search relevance across the full query distribution, from head queries to the long tail, from new products to established bestsellers.

After seeing the search results, Kogan expanded to the full Marqo product suite: search, recommendations, merchandising, and smart category pages. When a retailer with 16 million products and AUD $809M in revenue chooses to go all-in on your platform after seeing the search results, that is the strongest validation enterprise buyers can look for.

What This Means for Enterprise Retailers

The Kogan result demonstrates something important: Commerce Superintelligence performs best at enterprise scale. The larger and more complex the catalog, the greater the advantage of a dedicated AI over rule-based or behavioral-only approaches.

Here is why:

• More products = more cold-start opportunities. Enterprise catalogs constantly add new products. A dedicated AI understands them immediately. Behavioral systems leave them invisible until they accumulate enough clicks.
• More categories = more vocabulary complexity. A dedicated AI trained on your specific catalog understands that the same word means different things in different categories. A shared model treats every retailer's vocabulary the same.
• More queries = more long-tail revenue. Enterprise retailers see millions of unique search queries. The vast majority are low-frequency queries that no merchandiser will ever manually optimize. A dedicated AI handles them all automatically.
• More sellers = more catalog churn. Marketplaces with third-party sellers have products entering and leaving the catalog constantly. A system that depends on behavioral data to rank products will always lag behind the catalog itself.

This is why Marqo's largest revenue results come from its largest customers. The dedicated AI advantage compounds with scale.

Enterprise Results Across the Portfolio

Kogan is not an outlier. Marqo has delivered the largest published revenue uplifts in the product discovery category:

• Kogan: $10.1M incremental revenue
• Redbubble: $11M incremental revenue
• Mejuri: 19.84% increase in search revenue per user
• KICKS CREW: 17.7% uplift in conversion rate
• SwimOutlet: 10.6% increase in search add-to-cart rate, live in less than two weeks

These are named retailers with dollar-denominated results. See the full details on our Customer Stories page.

See It on Your Catalog

If you operate a large, complex catalog and your current search platform struggles with new products, long-tail queries, or cross-category intent, the Kogan result is directly relevant. See what your catalog looks like through Commerce Superintelligence. Book a demo and we will run Marqo on your products live.