Skip to content

Deploying DeepFM and Vector Retrieval for Fashion at Scale

/images/VividVisionContacts-hero.jpg?x-oss-process=image/resize,m_fixed,m_lfit,w_300

The Challenge of Complex Inventory and Data Sovereignty

Context: For VividVision Contacts, scaling to $1.8M monthly GMV brought a critical challenge: Wearing comfort issues leading to high return rates and negative reviews.

At this stage, standard rules failed. The Fashion-conscious adults aged 18-35 seeking both vision correction and aesthetic enhancement through colored contact lenses demanded relevance. Their self-developed system couldn’t handle thousands of prescription-power-color combinations across warehouses, while third-party SaaS solutions created data sovereignty risks for sensitive prescription data.

From Rule-Based Sorting to AI-Powered Personalization

To solve this, VividVision Contacts deployed WooRec. Note: Depending on their scale, they leveraged WooRec Private Deployment for the perfect balance of speed and control.

The transformation wasn’t instant. We architected a phased evolution of their recommendation engine:

Phase 1: Expanding the Candidate Pool with Vector Retrieval

Powered by WooRec Strategy Module

We needed to move beyond simple keyword matching. We implemented a Hybrid Recall strategy:

  • Foundation: Vector Retrieval (Embedding):
    • We started by mapping users and items into a high-dimensional vector space using FAISS, solving the Cold Start problem for new lens variants.
  • Advanced: Real-time Intent (U2I):
    • The Logic: We captured real-time user interactions (e.g., color zooms, prescription selections) to infer immediate purchase intent.
    • The Result: This allowed us to capture “latent interests”—finding lenses that matched aesthetic preferences beyond basic color categories.

Phase 2: Precision Ranking with DeepFM

Powered by WooRec Model Serving

This is the core of the engine.

  • The Iteration: Initially, the system relied on simple rule-based sorting. However, this failed to capture complex feature interactions like prescription-power-to-comfort mappings.
  • The Upgrade: We upgraded the ranking model to DeepFM.
    • Why this model?: By combining factorization machines with deep neural networks, WooRec could now learn both memorization (historical prescription preferences) and generalization (new comfort discoveries).
  • The Goal: A precise score for p(CTR) and p(CVR), directly targeting Repurchase Rate.

Phase 3: Business Logic and Personalization

Powered by WooRec Rule Engine

Raw scores aren’t enough. We applied final adjustments to align with business KPIs:

  • Diversity (MMR): We enabled Maximal Marginal Relevance to prevent category fatigue (e.g., over-recommending blue lenses).
  • Business Weighting: We boosted items based on Inventory Depth, ensuring the AI drives not just clicks, but sustainable revenue across prescription variants.

The Seamless Frontend Experience

Here is how these intelligent recommendations appear on the VividVision Contacts storefront:

dashboard *Figure 1: The result of WooRec's engine—hyper-relevant product recommendations displayed to the user.*

The Impact: 52% Repurchase Rate Growth

The speed of deployment meant faster results. By toggling on these strategies, VividVision Contacts achieved:

> *Interactive Chart: The rapid growth curve following WooRec configuration.*
  • Repurchase Rate: Increased by 52%.
  • Average Order Value: Improved by 29.4%.
  • Conversion Rate: Surged by 68%.
  • Return Rate: Reduced by 61.1%.
  • Customer Satisfaction Score: Improved by 22.2%.
  • Inventory Turnover: Increased by 50%.

Customer Voice

“Moving from manual rules to DeepFM was a turning point. The system now understands user intent in real-time. The 52% lift in Repurchase Rate speaks for itself.” — Sarah Chen, Head of Data Analytics at VividVision Contacts

Ready to Configure Your Growth?

You don’t need a team of data scientists to build a world-class recommendation engine. With WooRec, it’s just a matter of configuration.

Launch Your Strategy with WooRec