What is AI Hub in Synerise?
AI Hub is a set of AI-powered tools in Synerise that delivers personalized product recommendations, propensity predictions, and send-time optimization. Built on Synerise's Cleora AI engine, it analyzes each customer's behavioral and transactional data to surface relevant content across websites, emails, push notifications, and mobile apps — without manual model configuration.
How do AI recommendations work in Synerise?
Synerise AI recommendations use the Cleora engine to analyze customer visit history, purchases, and product feed data, generating real-time personalized suggestions. The model retrains weekly and after feed updates to stay current. Results are delivered across channels: website dynamic content, email, web push, mobile push, and mobile apps.
What is the difference between AI recommendations and rule-based recommendations?
Rule-based recommendations apply static logic such as "most popular products" or "recently viewed items" uniformly across customers. AI recommendations use machine learning to model each customer's individual preferences and predict which products they are most likely to engage with, producing more precise and personalized results at scale.
What types of Predictions does Synerise offer?
Synerise offers four Prediction types: Regression and Classification (predicts probability of events like churn or conversion), Lookalikes (finds customers who resemble your top performers), Propensity (calculates likelihood to purchase items with specific attributes like brand or category), and Best Fit (identifies the most suitable item or attribute value per customer).
How are Prediction results used in campaigns?
After training, Predictions save scores as events on customer profiles. The event name depends on the prediction type: Propensity saves snr.propensity.score, Best Fit saves snr.bestfit.score, Lookalikes saves snr.lookalike.score, and Custom (Regression/Classification) saves snr.prediction.score. These scores are available across Decision Hub, Experience Hub, and Automation Hub, letting you build segments of high-propensity customers or trigger personalized automations based on each customer's predicted likelihood to act. For more information, see AI Predictions.
What are response attributes in AI feed configuration?
Response attributes are the product properties returned in the recommendation payload to your website or application. You select them during AI engine configuration from two types: textual attributes such as product name, brand, or color, and range attributes such as price or size. Select only attributes you plan to use in templates to minimize payload size.
What happens if I turn off the Feed auto update feature in AI engine configuration?
Turning off Feed auto update stops the AI engine from syncing with the latest version of your product catalog. The catalog data itself stays current, but the AI engine continues operating on the snapshot taken before the feature was disabled, potentially serving recommendations based on outdated or discontinued products.
Which events are taken into consideration when using the time optimizer?
The events analyzed depend on the selected mode. The Email mode uses page.visit, newsletter.open, and newsletter.click. The Mobile mode uses screen.view and screen.interaction. The Web mode uses page.visit, product.addToCart, and form.submit. Custom modes let you select one predicted event and up to six input events.
What is the maximum time range in Predictions?
The maximum time range in a Predictions expression is 90 days. This window defines how far back the model looks when calculating propensity scores. If you need to evaluate longer behavioral patterns, consider combining Predictions with segmentation conditions that cover a wider historical range.
Can AI recommendations work for anonymous visitors?
Yes. Synerise generates recommendations for both recognized and anonymous visitors. For customers with no purchase or browsing history, recommendations are based on items clicked by other first-time visitors in the last 90 days. This ensures relevant suggestions are shown from the very first visit, without requiring login or identification.
How many products do I need in my catalog for recommendations to work?
There is no fixed product minimum for most recommendation types. For optimal performance, Synerise recommends at least 50,000 unique profiles who have visited multiple products and over 1,000,000 interaction events. The Top items type requires at least 10 unique products with 10,000 page visits or transactions. Data quality matters more than catalog size.
Can I exclude certain products or categories from recommendations?
Yes. You can apply filters using the visual builder or IQL queries to exclude products by category, brand, price range, or any custom item attribute. Filters can be applied at the slot level within a recommendation, and global item filters defined in AI engine configuration apply across all recommendation campaigns.
How do I know if my recommendations are performing well?
Each recommendation campaign has a Statistics tab showing CTR (clicks divided by generated recommendations), conversion rate, revenue, and average revenue per customer — all calculated over a rolling 30-day window. An item-level breakdown shows top-performing products by clicks, generations, and revenue. You must generate recommendation.click events for complete statistics.
Can I filter recommendations by category, brand, or price range?
Yes. Filters support category (including subcategory hierarchy levels), brand, price range, and any custom item attribute using operators such as equals, does not equal, less than, or greater than. You can also apply Distinct filters to ensure variety — for example, showing no more than two items per brand in a single recommendation.