The score is a summary, and it is the least important number on a decision record. It exists so that a hundred accounts can be ordered on a screen. It is not the reason any of them was qualified — the argument is.
What the score is#
A 0–100 figure describing how well-supported the case for contacting an account is right now. It moves as evidence arrives, ages and is verified, so it is a reading rather than a grade.
It is deliberately not comparable across workspaces. Two companies with different ICPs and different markets will produce different distributions, and a 72 in one is not a 72 in the other.
The four inputs#
| Component | Question | Drawn from |
|---|---|---|
| Fit | Is this the kind of company we sell to? | The firmographic half of the ICP, matched against the account record |
| Timing | Is there evidence the window is open? | Stacked signal evidence, weighted by strength and decayed by age |
| Reachability | Can we actually get to the person who owns this? | Whether an owner has been identified, and how the address was obtained |
| Capacity | Can they afford the thing we would be proposing? | Size, funding, expansion and any published spend signals |
Reachability is the component people expect least and argue with most. It is included because an account nobody can be reached at is not a good account, however perfect the other three are — and because it makes contact-data quality visible instead of letting it silently degrade results.
How the components combine#
Not as a weighted average. An average lets a strong component paper over a fatal weak one, which is exactly the failure mode that produces confident nonsense.
- Fit is a gate, not a term. Below the fit floor the account is D tier and no other component is computed.
- Timing is the primary term. With fit passed, timing carries the majority of the remaining weight, because timing is the thing the product claims to know.
- Reachability and capacity are dampers. They can pull a score down substantially; neither can push it up past what timing supports.
- The challenge applies last. If the adversarial pass succeeded, its effect is applied to the composite after the components are combined, and it is shown separately so you can see what it cost.
Shape matches the ICP on all four must-haves
Two families stacked; one record downgraded on review
Owner identified, address pattern inferred not verified
Size and expansion imply budget above the floor
Score is not priority#
A high score means the case is well supported. It does not mean you should contact that account first. Priority is set by the contact-by date, which comes from the window estimate.
| Account | Score | Contact by | Work it |
|---|---|---|---|
| A leadership hire, 3 weeks into tenure | 71 | In 6 days | First — the window is short and closing |
| A funding event, month 2 of 8 | 88 | In 5 weeks | Later — the window is long and the score is not going anywhere |
Calibration#
A score is only useful if it means the same thing over time. Calibration compares what the scores predicted against what actually happened, per workspace, and adjusts the mapping from evidence to number — never the evidence itself.
- Positive outcomes. A reply classified as interested, a meeting booked, a deal opened. Weighted by how far down the funnel it got.
- Negative outcomes. Wrong-person replies, hard bounces, explicit not-interested. A wrong-person reply is charged to reachability specifically, not to the composite.
- Neutral outcomes. No reply. Treated as almost no information, because it usually is.
Calibration needs volume before it means anything. Until a workspace has accumulated a meaningful number of classified outcomes, the mapping stays at its defaults and the console says so rather than showing a curve fitted to a handful of events.
Reading a score breakdown#
The breakdown is available on every decision. The two fields worth reading are limited_by, which names the component holding the score down, and challenge_delta, which shows what the adversarial pass cost.
{
"account_id": "ACC-19022",
"composite": 76,
"components": {
"fit": 88,
"timing": 74,
"reachability": 61,
"capacity": 80
},
"limited_by": "reachability",
"challenge_delta": -8,
"calibration": { "state": "defaults", "outcomes_recorded": 41 },
"computed_at": "2026-07-24T06:02:11Z"
}limited_by: "reachability" across many accounts is not a scoring problem. It is a contact-data problem, and it is fixed upstream rather than by adjusting weights.
Next: campaigns — what happens to an account once it is qualified.