What it is
fit_score is a 0-100 measure of how closely an account matches the kind of customer your business does best with, where a higher number means a stronger match. It covers the whole relationship, from first contact through renewal — it is not a sales qualification score that stops at the close.
It is modelled, not measured. A measured value is reproduced by running the same arithmetic over your data. A modelled value is composed: it weighs many things against each other, it is calibrated against what actually happened to past customers, and it improves with operating tenure. Everything on this page follows from that difference.
The shape you get depends on where you read it. On the read interface and MCP tools you get the exact score, with its as-of and the version of the model that produced it. In your CRM it lands as a band — high, medium or low — with its own as-of. A CRM field strips the definition, confidence, version and freshness that make a modelled number safe to read, and a bare score in that setting reads as a measurement. A band stays useful without that context and cannot be quoted as a precise fact.
How it is calculated
Modelled. Five dimensions are scored and combined at fixed weights. Each dimension is itself built from many components — 108 across the five.
The five weights always sum to 1.00. Four weight sets ship — balanced, growth-aggressive, retention-first and margin-optimised — and which one is in force follows your growth plan. Changing it is a joint CFO and CEO decision, never an automatic one. Two segment adjustments apply on top: enterprise accounts carry three points more weight on risk and three fewer on growth, SMB accounts three points more on product match and three fewer on profitability. The tier thresholds at 45 and 75 are defaults and can be set for your company.
Completeness is published with the score. It states how much of the 108-component model is currently active for that account. Components not yet available count neutral rather than dragging the score down, so an early score is incomplete rather than pessimistic.
Where an account’s identity is resolved below 80 confidence, the whole score is scaled down in proportion and the flag says data quality is the constraint rather than fit.
Where it comes from
Fit score is composed from values Beacon already holds rather than read from one source system, so how complete it is depends on how much of Beacon is switched on.
A score exists from first contact — the creation of a prospect record, or the first intent signal — and is recomputed from then on. These are design targets; actual completeness depends on your data and how deeply each tool is connected.
How fresh it is
Recalculated in full every month for every active account, and partially in between whenever a behavioural signal category moves by ten points or more, another part of Beacon is switched on, or a contract event lands — a renewal, an expansion, a contraction or a churn. Each recalculation snapshots the weight set and tier thresholds in force, and the score is sealed on that cycle. The as-of and the model version travel with the exact score wherever it is published, so a score you quote can always be traced to the model that produced it.Currency and rounding
Neither applies. The score has no unit and holds no money value.What changes it
Movement in any of the five dimensions. A change of segment changes the adjustments applied. A change of growth plan can change the weight set, which moves the score without anything changing at the account. Switching on another part of Beacon raises completeness and can move the score, as components that were counting neutral start contributing. The model is retuned once a quarter, against what actually happened. Churns, expansions and retention outcomes from the prior quarter are compared with the dimension scores those accounts carried at the time. Where the relationship between a dimension and real outcomes has materially shifted, a weight change is proposed. Proposed, not applied — a named owner approves it, the change is logged, and it takes effect at the next monthly recalculation. Nothing in the model changes itself, and everything downstream is told when it does. Between retunings the model is watched for drift — whether it returns a materially different answer on unchanged inputs, whether it moves while its inputs have not, whether it starts contradicting other parts of the system, and whether it recalculates off its own schedule. A review can clear the value, hold it, or escalate it. A hold pauses consequential use; it does not change the score. Separately, a named owner — never an agent — can override the value used for one decision, with the reason recorded. An override changes what that decision consumes. It never changes the model.What it is not
- Not a score for companies you have never dealt with. Fit runs on accounts in your own connected data. Beacon will not score a company it has never seen; instead it exports the model itself for your own tools to apply. See Boundaries.
- Not
customer_value. Fit is how closely an account matches who you sell best to; value is what the account is worth to you. A high-fit account can be low value. The outcomes behind customer value are part of what recalibrates this model each quarter, not the other way round. - Not
health_score. That one is measured, reproducible from your data, and about how strongly an account is progressing. This one is modelled and about match quality. They can disagree without either being wrong. - Not one of the five dimension scores. Those are the parts; fit score is the whole. Only the composite is published on the account object.
- Not a deal probability. Fit describes the account. A deal’s likelihood of closing is a separate value at deal grain, and fit is one of the things that weights it.
- Not a reason on its own. Fit is a summary of many things, and the completeness figure beside it tells you how much of the picture it is drawn from.