TL;DR

We’ve built Shaped Value Modeling, a framework that enables real-time optimization of recommendation and search ranking models for multiple objectives. It allows teams to blend different objectives (e.g. relevance, conversions, diversity) and dynamically run AB tests that drive business outcomes (with built in analytics tooling). Our community has seen awesome results, so we’re excited to bring this framework to the masses!

Watch our demo here.

Key Features:

• Multi-objective optimization: Combine multiple models and signals (e.g., click prediction, purchase likelihood, content quality) into a single, customizable scoring function.
• Dynamic ranking logic: Define and adjust ranking objectives as your business needs change using a simple python DSL.
• Real-time API: Modify ranking scores on the fly with simple API calls, enabling rapid experimentation and A/B testing.
• Easy to integrate: Connect and deploy rapidly with direct integration to your existing data sources.
• Explainability: Gain insights into why certain items rank higher through interpretable scoring logic, and a built in analytics platform out of the box.

Why This Matters:

• Rapid iteration: Experiment with new ranking strategies and adapt to changing business needs without lengthy retraining cycles.
• Improved control: Fine-tune your ranking algorithms with precision and transparency.
• Better alignment: Optimize for multiple KPIs simultaneously, ensuring your recommendations drive both engagement and business value.

How it Works:

1.Train separate models for different objectives (e.g., LightGBM for click prediction, BERT4Rec for purchase likelihood).

2. Define a "value model" that combines these models using mathematical expressions and custom logic. Let's say we have a lightgbm click-through-rate model called ctr_model and a transformerd based next purchase model called purchase_model, you can create an ensembled value model as follows:

value_model = 0.5 * ctr_model + 0.5 * purchase_model + log(item.quality_score)

3. Dynamically adjust the value_model via API calls to optimize for different goals or user segments:

{
  "user_id": "83NSLX",
  "config": {
    "value_model": "0.3 * ctr_model + 0.7 * purchase_model"
  }
}

How our customers are using value modeling:

• Marketplaces: Dynamically mix personalized scores with a geo-location penalizer to connect users with nearby items.
• E-commerce: Balance purchase intent, engagement, and product quality.
• Content platforms: Implement time-decay, boost high-quality content, and personalize recommendations.
• Social media: Dynamically switch between personalized and trending models.

Our ask

If you know decision-makers in any of these types of organisations, we’d be super grateful for an introduction at daniel@shaped.ai

1. Marketplaces
2. E-commerce
3. Content platforms
4. Social Media

For more information see our website here and developer docs here

Get a demo if you’re interested in testing value modeling!