A score is a routing decision, not a verdict
The purpose of lead scoring is to decide who gets called first, not to decide who is worth talking to. Treated as a verdict it does predictable damage: unusual but valuable enquiries get filtered out, and because the rejected ones are never contacted, nobody finds out. We set an explicit review threshold so that anything the policy is unsure about goes to a person rather than to a bin, and we treat missing data as missing rather than as negative — a company that did not fill in a revenue field is not a small company, it is an unknown one.
Speed matters more than sophistication
For inbound B2B enquiries the single largest measurable improvement is almost always response time, not scoring accuracy. A simple, transparent rule set that responds in two minutes will beat a sophisticated model that responds in two hours, and it has the additional advantage that a sales manager can read it, argue with it and change it. We start there, instrument the outcome, and add sophistication only where the data shows the simple rule is actually costing conversions. Consent rules constrain what the automated response may contain, so opt-in status is checked before any outbound step, not after.
Walk through a lead qualification pilot
For example, a form submission might contain a company domain, role, requested service and contact preference. Keep fit, buying intent and eligibility as separate fields with sales-approved rules; if the domain is missing, signals conflict or a duplicate is possible, create a review task instead of rejecting the lead. Write only the agreed qualification and owner fields to the CRM, preserve the source record, and make retries check the event or record ID before creating anything again. Before launch, the sales owner records current time to first human response, reviewed routing errors and exception volume; accept the pilot only if those same measures improve without increasing missed or misrouted enquiries. Reconcile failed writes from the source queue and alert the named owner rather than silently dropping the lead.
How to measure the pilot
A pilot is accepted against the current process, not an industry promise. Agree the sample, baseline and owner before launch, then review the same measures after the workflow has handled real cases.