Audit the number, then audit the data
A score you cannot check is a rumour with a colour. So the method is published, the track record is frozen, and every rate arrives with the number of cases behind it. The same rule governs the data underneath: one path in, per-user identity, a confirm step on every write.
You can read the method
Weights, thresholds and caps are published rather than buried in a rules engine, and each one is a pure function with unit tests behind it. You can disagree with a score on the merits and point at the line you disagree with.
The track record cannot be edited
A prediction is written down the day it is made and graded once the outcome settles. Nothing recomputes it on read, so tidying up a close date in the CRM months later cannot quietly improve how the system looks in hindsight.
Isolation by deployment, not by query
Every tenant, including VERA itself, runs its own database, instance, and URL. Your data never shares a table with another customer. Isolation is enforced by the deployment boundary, which is the strongest kind.
A single governed door
No application layer ever calls a vendor SDK directly. CRM reads, writes, and model inference all flow through your own gateway, where authentication, per-user identity, the write-gate and the audit log live. That is what makes the provenance of any input to a score a query rather than an archaeology project.
The full picture
Why you can trust the number
Two questions decide whether a score is worth anything. Can you audit how it was reached, and can you trust where its inputs came from? Both answers are structural here, not policy.
The method is readable
Weights, thresholds and caps are published, not buried in a rules engine. You can disagree with a number on the merits, which is the only kind of disagreement worth having about a score.
The track record is frozen
Graded predictions are append-only and immutable. Nothing recomputes them on read, so a correction upstream can never quietly improve how the system looks in hindsight.
Every rate ships its denominator
Accuracy figures travel with the number of cases behind them, and a sample too thin to support a claim reports insufficient evidence rather than a percentage.
One governed data path
No layer touches a vendor SDK directly. CRM data and model inference both route through a single governed gateway, which is what makes the provenance of any input to a score a query rather than an archaeology project.
Per-user identity, not a shared key
Reads and writes run as the real person through OAuth, enforced at the database with row-level security, so people see only their own portfolio.
Propose then confirm on every write
Nothing reaches your CRM without an explicit confirmation step. Every write is attributed to a real user and audited, and read-only viewers can never write.
Could you defend your health score in a board meeting?
Bring us the score you already have and the calls it got wrong. We will tell you what we would measure differently and how we would grade it. If that turns into software, good. If it turns into a better spreadsheet, also good.