Signal score
The Signal score is the 0–100 number on every bug. It’s the priority read — the number you scan a row by, and what ranks the Needs you now list at the top of your dashboard. Higher means “look at this sooner.”
Your Inbox list itself is newest-first; the score doesn’t reorder it.
What goes into it
The score blends two things you can see, gated by how confident the classification is:
| Dimension | What it captures |
|---|---|
| Severity | How damaging the bug is — a Critical bug starts far higher than a Low one. |
| Frequency | How often this issue keeps coming back. A repeat report raises the score, but with diminishing returns — the tenth re-report doesn’t drown out a brand-new critical bug. |
The detail panel shows these as a breakdown so you can see why a bug scored the way it did.
Two more signals
Alongside Severity and Frequency, the breakdown shows two more dimensions that round out the picture of a report:
| Dimension | What it captures |
|---|---|
| Novelty | How unlike anything already in your tracker this report is. A brand-new kind of issue scores high; a near-copy of something you’ve already seen scores low. |
| Sentiment | How urgent the report reads — the reporter’s tone. A calm, factual note sits low; an “URGENT — customers are blocked!” message sits high. |
These two are display only — they add context for your decision, and the priority number itself is built from Severity and Frequency alone. A bar stays locked when the dimension was never measured for that report: reports from before these shipped, and the occasional one where the tone read couldn’t be made.
Novelty reads at its maximum whenever duplicate detection returned nothing at all — including on a brand-new workspace whose tracker index is still filling in. Early on, a full Novelty bar says more about the index than about the report.
The score also carries a confidence marker — a small × conf next to it.
When a report is thin and the read is less certain, that marker damps the
score down. It’s never zeroed, just held back until there’s more to go on.
Frequency, in plain terms
Frequency counts the times a report was approved as a link to the same ticket — not the times the bug was mentioned. So the number moves when you act on a repeat, not the moment it arrives. The curve flattens as the count grows, so a noisy, often-repeated low-severity issue can’t out-rank a fresh critical one on volume alone. Severity stays the dominant signal.
Provisional vs calibrated
While your workspace is young there isn’t much routed history behind a score, so scores carry a [ PROVISIONAL ] tag — they work, but they’re still settling in. Once you’ve routed enough bugs the tag drops away. See Calibration.