Scoring Content Sources
Some articles hit the mark. Others just take up space. Scoring is how you tell Ongage Studio which content signals to trust — and which to quietly move aside.
Whether you're building a Datastream for deep-dive finance insights or faith-driven parenting advice, source scoring is your tuning fork. It helps you boost credibility, filter fluff, and shape your Agent’s research before a single word gets written.
Why Source Scoring Matters
Datastreams don’t just scan — they judge. But not all sources deserve the same weight. Your scoring tells Studio:
- Which publishers are worth elevating
- Which ones should be downplayed or deprioritized
- Which to block completely (because life’s too short for clickbait)
By tuning your Source Scores, you control the quality and tone of everything your Agents create.
How Scoring Works
Scoring happens inside the Datastream metadata filters. You’ll find four levels of control:
| Filter | What It Does |
|---|---|
| Included Sources | Only these sources are allowed — all others are ignored |
| Boosted Sources | These get a numerical score boost (0–1000) — higher means stronger visibility |
| Penalized Sources | These still appear, but with lower ranking (–1 to –1000) |
| Excluded Sources | These are fully blocked — they’ll never appear in results |
The more refined your filters, the more refined your output.
Example Use Cases
| Goal | Scoring Strategy |
|---|---|
| Prioritize expert journalism | Boost: Reuters (900), WSJ (850), NYT (800) |
| Downplay low-quality aggregators | Penalize: Patch (–700), Medium (–500), BuzzFeed (–800) |
| Block politically biased coverage | Exclude: Breitbart, HuffPost, The Nation |
| Highlight niche expertise | Include: Stateline, Pew Research, Local trade journals |
| Avoid commercial content | Exclude: Sponsored blogs, affiliate sites, advertorials |
Scoring in Action
Let’s say you’re building a Health & Wellness Datastream:
- Boost: Healthline (800), Cleveland Clinic (900), Harvard Health (850)
- Penalize: Medium blogs (–600), sponsored content networks (–700)
- Exclude: Goop, TikTok health creators, Daily Mail
Now, every Agent pulling from that Datastream will “see” the world through this lens. The content they generate will reflect it — more medical clarity, less pseudoscience.
Tips for Smart Scoring
- Don’t just exclude everything you don’t like — penalizing lets good content through, but with caution.
- Boost conservatively — a score of 1000 means “always show,” so reserve that for top-tier sources.
- Combine source and item logic — e.g., “Boost Climate Policy only from Reuters and Pew.”
- Preview after changes — always click “Preview Articles” to see the effect of your scoring in real time.
Pro Tips
- Exclude liberally for known low-trust or ad-heavy domains.
- Use boosts to steer niche tone (e.g., scholarly, mainstream, faith-based).
- Penalize “maybe” sources instead of blocking — Studio will learn what sticks.
- Save a template if you use the same scores across multiple Datastreams.