Scoring, and how to bend it toward your business
Every signal carries a score out of 100, built from seven components:
Each listener has a score bar: the minimum a conversation must clear to be kept, 55 by default. Below the bar, a candidate is recorded as rejected with its reason rather than stored. Competitor and content rows skip the bar, since low-scoring intel is still intel.
How much of this you can tune is a four-rung ladder rather than a switch, because the middle rungs are the interesting ones:
The presets are weight sets we have checked rather than seven free sliders, which is what stops a workspace weighting itself into a scale where nothing scores above 40:
See openpulse.cloud/pricing for which plan includes score tuning.
Quality: seeing the filter’s working
This is the part no comparable product shows you, and it is deliberately product surface rather than diagnostics.- Rejections are kept, with reasons. Every dropped candidate records why: wrong page shape, off-topic, excluded by your own keyword, blocked domain, too old, duplicate, ranked out, below the score bar, or noise. Kept for 14 days. When a listener feels quiet, this tells you whether the searches returned nothing or the filters threw everything away.
- Per-query and per-source yield. Which of your searches earn their cost and which have never produced a kept signal. A query that produces candidates but no signals for three runs in a row is disabled automatically and reported in the run log.
- Your thumbs move the bar. Rated signals are fed back into that listener’s next classification as worked examples, so the definition of relevant becomes yours rather than one a model guessed at on day one.
- The excluded list shows what was dropped, in the interface, next to what was kept.
get_quality_report. See MCP & Agents.