Boosting, Penalizing, and Excluding Items

Data Streams give you full control over the content your Agents are allowed to use. The most important part of that control lies in how you score and filter metadata using three core actions:

• Boosting – increases the visibility of content
• Penalizing – reduces the likelihood of inclusion without blocking
• Excluding – completely removes content from the stream

You can apply these actions to both Items (topics, people, entities, etc.) and Sources (websites, publishers, domains). Understanding when and how to use each action is essential for shaping the editorial focus of your Stream.

Boosting Items and Sources

Boosting is used to prioritize certain themes, keywords, or sources. When an item or source is boosted, articles that match it receive a higher score and are more likely to appear in your Stream results.

Boosting Range

• Boost scores range from +1 to +1000
• A score of +1000 means the highest possible priority
• Scores between +600 and +800 are typically strong enough for noticeable influence without dominating the stream

When to Boost

Use boosting when you want to elevate coverage of a specific topic or source, but don’t want to make it mandatory.

Examples:

• Boost "Federal Reserve" at +850 to surface articles about interest rate policy
• Boost "Olympic Games" at +750 to promote coverage during an event cycle
• Boost "Christian worldview" at +900 to favor theological alignment in cultural articles
• Boost "Reuters" at +800 to increase the weight of high-trust reporting

Boosting can also be applied proactively to balance out other penalties or exclusions.

Penalizing Items and Sources

Penalizing is a way to deprioritize certain types of content without fully excluding them. Penalized content will appear less frequently and only if it meets a very high editorial standard or is strongly boosted by other factors.

Penalizing Range

• Penalty scores range from –1 to –1000
• Scores of –400 to –700 are useful for soft filtering
• –1000 is essentially equivalent to exclusion in most use cases

When to Penalize

Use penalizing when content may be relevant occasionally, but generally doesn’t align with your goals or tone. Penalizing is ideal for handling noisy, low-quality, or off-topic themes.

Examples:

• Penalize "NFTs" at –700 in a traditional finance stream to discourage crypto hype
• Penalize "Reddit" at –600 to reduce results sourced from anonymous or user-generated content
• Penalize "Satire" at –800 to avoid comedic pieces without blocking legitimate cultural analysis

Penalizing is also useful for controlling tone. For example, you might penalize "partisan commentary" while boosting "economic policy" to guide the narrative direction of general news articles.

Excluding Items and Sources

Exclusion is the most restrictive filter. Any article that matches an excluded item or source will be removed from the stream, even if it scores highly on other criteria.

When to Exclude

Use exclusion when certain topics, themes, or publishers are inappropriate, irrelevant, or in direct conflict with your brand standards.

Examples:

• Exclude "BuzzFeed", "TMZ", or "The Onion" to maintain journalistic credibility
• Exclude "Clickbait" and "Celebrity gossip" from professional health or finance feeds
• Exclude "Op-eds" or "Sponsored content" to avoid non-editorial or promotional pieces
• Exclude sensitive or off-mission themes like "identity politics", "lifestyle activism", or "polarizing social issues" in mission-driven content

Be careful not to overuse exclusions — they override scoring logic entirely. Too many exclusions may result in an empty or underperforming Stream.

How to Apply Boosts, Penalties, and Exclusions

Each Filter Group in the Data Stream builder allows you to apply these actions through the score selector.

1. Click + Add Filter Group
2. Choose Items or Sources
3. Search and select relevant metadata from the item browser (categories, topics, people, organizations, etc.)
4. For each entry, set a score:

• +1 to +1000 for boosts
• –1 to –1000 for penalties
• Click ‘Exclude’ to fully block

You can organize filter logic by naming groups clearly — for example:

• Included Items (required)
• Boosted Items (600–1000)
• Penalized Items (–400 to –900)
• Excluded Items
• Boosted Sources
• Excluded Sources

Best Practices

• Use boosts and penalties before exclusions – try nudging before blocking
• Keep Included Items short – 2 to 4 strong tags is ideal
• Balance scoring logic – if you boost too many items, your Stream becomes unfocused
• Organize filters by purpose – separate groups for content, tone, and trust
• Always preview – test your results using the Preview tool before assigning the stream to Agents

Sample Scoring Table

Summary

Boosting, penalizing, and excluding are the core tools for controlling what your Data Stream delivers. They allow you to:

• Shape tone and focus
• Prioritize reliable, aligned sources
• Suppress off-brand or low-value content
• Maintain editorial consistency across all your outputs

Used effectively, these filters turn raw internet noise into a usable stream of high-quality content that reflects your voice and strategy.