Evidence-based account scoring ranks sales attention using signals that retain a source, observation date and reason to act. The seller should be able to inspect a material trigger before outreach and distinguish verified evidence from a weak or stale proxy.
Composite reference architecture. A reference RevOps design; no pipeline or revenue uplift is asserted.
The operating problem
A high score is useless if sales cannot tell whether it came from a stale firmographic field, a weak intent proxy or a verified operational trigger.
How the workflow works
Create an account-signal ledger that joins CRM history, relevant company data and observed operational events. Resolve the account identity before scoring, and apply freshness rules appropriate to each signal. A scoring service can prioritise the next research task or candidate account, while the analyst records the evidence that confirms or contradicts the trigger.
Where review and controls matter
A high score should not automatically create outreach. Sellers validate identity and relevance, respect consent and suppression rules, and preserve the reason an account was prioritised or held. Outcome feedback can improve the model only if the team also reviews neglected accounts; training solely on contacted winners can reinforce the existing selection bias.
Risk reference: NIST AI RMF.
What to measure
Use a fixed cohort and measurement window. Count numerator events only within the denominator's eligible cohort; sampled measures use the same audit sample for both terms. Record the source, exclusions and owner, and keep missing data visible. These definitions are proposed measurement contracts, not observed results.
| Measure | Definition and unit |
|---|---|
| Validated-signal rate | Reviewed signals confirmed against a suitable source divided by reviewed signals; percent. |
| Accepted accounts | Accounts accepted by the seller for a stated next action divided by accounts proposed for review; percent. |
| Meetings | Count of attended, qualified meetings linked to the scored cohort under a declared attribution rule. |
| Paid pilots | Count and contracted value of paid pilots linked to the cohort; exclude verbal interest. |
| Win rate | Closed-won opportunities divided by closed-won plus closed-lost opportunities for the cohort; percent. |
| Sourced pipeline | Value of qualifying opportunities first sourced by the workflow under a fixed attribution rule; currency, with stage and as-of date. |
A useful first pilot
Use a dated account cohort and a small set of inspectable signals. Compare scoring with seller acceptance and subsequent qualified outcomes, keeping rejected accounts and missing evidence in the evaluation.
Bring the workflow, its source records and the decision that needs an owner to a BergLabs pilot discussion.
