Reference
How scoring works
From evidence to a 0–100 score: rules, confidence, decay, and the breakdown.
Nothing here is a black box. Every account carries two auditable numbers, answering two different questions — Fit says whether; Event says when.
Two scores, two questions
| Score | What it is |
|---|---|
| Fit (0–100) | Is this company the kind of customer you sell to? Judged by research against your qualification criteria, with the reasons shown on the account page (a transparent heuristic fills in until research runs). Mostly static — fit doesn't decay. A dash means not assessed yet. |
| Event score (0–100) | Is something happening right now? The evidence math this page explains: rule points × freshness × confidence. Decays as news ages; drives the priority bands, the Queue, and alerts. |
Fit 95 / Event 20 → watch, don't call. Fit 90 / Event 92 → contact now. The Today list ranks by Event × Fit, and the Proof report measures how well the Event score predicted pipeline. The rest of this page is the Event score math — a sum you can audit. Click any score in the app to open the breakdown and see every line of it for that account.
The pieces
- 1Rules say what matters and how muchEach workspace has scoring rules like “Hiring for security engineering — 25 points” or “New engineering leader — 15 points”. See and tune them in SettingsScoring. Untick Enabled to switch a rule off without deleting it; owners and admins can also delete any rule after a confirm (evidence tagged with it stops adding points, and every account is rescored).
- 2Signals activate rulesA sourced signal linked to a rule contributes that rule's points. Each rule counts only its single strongest signal, capped at the rule's points — so extra sources for the same story strengthen confidence (and show as supporting signals) rather than stacking more points. A signal without a source contributes zero. Categories that come from your own activity — website visits, email clicks, logged engagement and a champion joining a tracked company — are filed by Diurno HQ itself; research can't file a signal under them.
- 3Freshness decaysEvidence fades as it ages — a hiring push from last week matters more than one from last winter. You choose the decay style in SettingsScoring, on the Signal decay model card: Per-signal half-lives (a smooth fade) or Step windows (stepped cliffs). The card spells out your own windows, the score breakdown says how much each signal still counts (“full weight”, “N% weight (aging)”), and the age chip on a signal gives “Age weight N%” from the settings of the workspace the account belongs to. Evidence whose source doesn't state a date isn't treated as new: it is scored as ageing (about six weeks old on the standard freshness windows; your own windows apply), and its detail says no publication date was found.
- 4Confidence scalesFirst-party sources (the company's own site, filings) count fully; weaker or second-hand evidence counts partially; “needs verification” counts nothing.
From score to band
Three thresholds cut the 0–100 range into the four priority bands, shown everywhere as Call now, Call this week, Monitor, and No trigger yet. They're per-workspace settings (owners and admins) — make “Call now” stricter or looser to fit your team's capacity; saving rescores every account.
When scores change
- New research or a new signal → instant rescore of that account.
- Changing a rule's points or switch, deleting a rule, or changing the thresholds or decay settings → every account is rescored on the spot.
- The daily agent → whole-universe rescore with decay applied.
- A score drop with no bad news usually just means time passed — that's decay doing its job.
Trust, then verify