Algorithms

Built-in Algorithms

FirstBatch provides multiple built-in algorithms serving distinct purposes.

  • Navigable UX

  • User-Intent AI Agents

  • Individually crafted recommendations

  • Unique Journeys

  • User-centric promoted content curations

Navigable UX

Transform your user experience into a navigable journey by leveraging user interactions. Every user action contributes to shaping their unique experience.

Target

Anonymous sessions to navigable experiences.

Description

Serving various content until receiving the first signal from users. Then crafting the experience depending on the signals. Gradually providing more space for exploration if the user keeps losing interest in current topics.

Potential KPIs

  • Any engagement metric

  • Time spent on the application or Bounce rate

  • Conversion rate of anonymous sessions

User-Intent AI Agents

Empower your AI agents with real-time insights into user intentions, derived from their interactions. This infusion of user intent brings intimacy to AI-driven experiences, making users feel more connected and understood.

Target

Serving personal AI assistance that reflects user interactions that are not restricted to prompts.

Description

Not giving space for false navigation and keeping the AI agent as close to user intentions as possible due to people being demotivated by hallucinated conversations with AI too fast.

Potential KPIs

  • Time spent with AI agents

  • Chat Rating

  • Conversion rate through AI agents

Individually Crafted Recommendations

Offer users not only similar but also adjacent items in a personalized manner. This approach allows users to discover new and relevant content on their own terms, enhancing their exploration and satisfaction.

Target

Increase up-sell and help you to improve average order value.

Description

Making highly focused recommendations after the first interaction; at the same time enabling users to explore more items from a wider perspective to keep them within the recommendations space until they find something to add to their cart.

Potential KPIs

  • Up-sell and cross-sell metrics

  • Average Order Value

  • Number of items per order

Unique Journeys

Enable users to access the right content from the very beginning by tailoring their experience based on their starting point.

Target

Shape the user journey from the very beginning. This might be the best for recurring visitor experiences and welcome routines that are traffic source-based or seasonal campaigns-based.

Description

Providing focused content starting from the first load by utilizing user embeddings from previous sessions or adding seasonal effects to the experience.

For example, adding summer collection as a bias during summer; and then letting users navigate themselves just as we do in the Navigable UX algorithm.

Potential KPIs

  • Up-Any engagement metric

  • Time spent before the first interaction

  • Conversion rate of recurring visitors

  • Time spent on the application or Bounce rate

Not User Targeting but User-Centric Promoted Content Curations

Shift away from conventional targeting techniques and embrace a user-centric approach to deliver promoted items or ads in a captivating format.

This approach allows users to actively influence the curation of promoted content, ensuring it aligns seamlessly with their preferences and resulting in a highly interactive and enjoyable experience.

Target

Not force your users to see irrelevant promoted content but provide an engaging campaign discovery.

People ignore ads because targeting only pollutes feeds. Therefore the aim is to improve the campaign CTR by providing true content for the right users at the right time.

Description

Promoting content in a hyper-personalized manner by keeping curation focused after the first interaction to forever.

Potential KPIs

  • CTR

Custom Algorithms

FirstBatch also enables the creation of custom algorithms through the User Embeddings Dashboard. A detailed walkthrough on how to create algorithms can be found here.

Creating a Session with an Algorithm

When creating a session, you can provide the algorithm name (as an Enum for Python, or a typed-string for TypeScript) to the function. The accepted algorithms are the following:

Algorithms are attached to sessions through AlgorithmLabel enum. You can simply use the enum fields when creating a session: 

personalized.session(algorithm=AlgorithmLabel.SIMPLE, vdbid="my_db")

PreviousFirstBatchNextVectorStores

Last updated