In short: Smart Filtering uses a two-step AI system, pre-filtering then filtering, so only relevant, actionable results reach your dashboard, and it's retrained every 1-2 weeks to stay accurate.
Understanding Smart Filtering
Smart Filtering streamlines how data flows into Zeal using a continuously trained, two-step system. It makes sure only the most relevant and actionable results reach your dashboard, while optimising overall data collection.
By evaluating content at multiple stages, it widens your search net without wasting resources on irrelevant data.
1. How it works: the two-step system
Each step decides whether a result is worth keeping, so your team doesn't wade through noise.
Data is processed through a sequential filtering pipeline that evaluates both post features (title, description, price) and image features (Optical Character Recognition (OCR), logo detection) to decide how relevant it is.
Pre-filtering: the system rapidly scans the initial search page results to decide whether they need deeper scraping. This step evaluates sources quickly, keeping data usage low while maintaining a broad search net.
Filtering: results that pass the first stage get detailed, granular checks, including full-text analysis, logo detection, keyword matching and image category recognition, to confirm the content is directly relevant to your brand.
Learn more: Optical Character Recognition (OCR) · Logo Detection
2. How the model learns and adapts
Smart Filtering keeps up with new trends instead of relying on fixed rules.
It's an AI system that learns continuously: the models are retrained every 1 to 2 weeks to adapt to new trends and improve accuracy. To do this, the system monitors and labels data from three sample groups: unfiltered results, pre-filtered results and filtered results.
Both the AI models and human moderators label posts from these groups. The system then compares the model's performance against the human labels, and uses this data to retrain and refine the filtering logic.
Tip: Accurate moderation helps Smart Filtering improve, because the models are compared against human labels when they're retrained.
Learn more: Understanding Moderation · Labels
3. What it means for you
You get cleaner results with less wasted data collection.
Optimised data usage: less unnecessary scraping keeps data collection efficient and cost-effective.
Refined analysis: combined text, price, logo, keyword, OCR and image checks tailored to your brand.
Relevant data: your dashboard is loaded only with actionable results.
Continuous adaptation: filtering performance improves over time, with models retrained every 1-2 weeks.
Learn more: How to Improve Data Relevance · Keyword Intelligence
Quick reference
If you want to know… | Look at… |
How results are filtered before they reach you | 1. How it works: the two-step system |
How often the models are retrained | 2. How the model learns and adapts |
What you gain from Smart Filtering | 3. What it means for you |
How to make your results more relevant |