Quick answer: define acceptance before assigning points

Start by agreeing what qualifies a lead for sales review. Specify service fit, geographic eligibility, buyer intent, contactability and any operational constraints. Only then design a score. A model that gives points for every download and page view may rank enthusiastic researchers above ready-to-buy customers.

Separate fit from engagement. A person can show strong intent but be outside the service area; another can be an ideal customer who has barely interacted with content. Two dimensions are often easier to diagnose than one opaque number. Make the score explainable to the team that must act on it.

Map the lifecycle and its owners

Define inquiry, marketing-qualified lead, sales-accepted lead, sales-qualified opportunity, customer and disqualified lead in a shared glossary. Each stage needs an entry rule, owner, timestamp and exit condition. A lead should not become accepted merely because a workflow changed its status automatically.

Agree on response targets, reassignment rules and a feedback window. If sales rejects a lead, require a structured reason such as no fit, duplicate, unreachable, student/researcher, not ready or capacity unavailable. Free-text rejection notes alone are difficult to analyze and often become a proxy for inconsistent follow-up.

Build a simple starting score

Use a transparent scorecard with a few fit attributes and a few meaningful intent signals. Example: eligible service location, relevant business type, explicit request for a consultation and a confirmed contact method. Give little or no weight to superficial actions such as an email open, which can be distorted by privacy features.

Set negative rules for disqualifying conditions, repeated spam submissions and existing-customer support requests. Avoid sensitive or discriminatory attributes and seek appropriate legal review where relevant. Do not build a complex predictive model until the CRM has enough clean accepted and rejected outcomes to evaluate one.

Calibrate against actual sales feedback

Take a recent cohort and compare scores with sales acceptance, qualified opportunities and eventual customers. Build a confusion matrix: high-score accepted, high-score rejected, low-score accepted and low-score rejected. Investigate the disagreements. A model that creates a large volume of false positives increases sales workload even if marketing's dashboard looks stronger.

Segment by source and service type. A webinar lead and an emergency-service enquiry may have very different timelines, so one global threshold can be misleading. Adjust the rules in documented increments and recheck the next cohort rather than continuously moving the threshold to make the reported acceptance rate look better.

Example: fixing an inflated MQL pipeline

Suppose a campaign generates 200 inquiries, marketing labels 140 as qualified, sales accepts 50 and only 20 become genuine opportunities. The first problem is not necessarily insufficient advertising volume. It may be that the qualification definition overweights form completion and underweights service eligibility or actual buying intent.

Review a sample of the ninety rejected leads, classify reasons and check whether the sales team followed up consistently. If most were outside the target geography, improve targeting and the form. If most were not contacted promptly, repair the response process before declaring the acquisition channel low quality.

Create a closed-loop operating cadence

Hold a short weekly review of rejected and high-value accepted leads with marketing and sales. Update the reason taxonomy, note recurring objections and choose one change to test. Each month, review acceptance rate, speed to first contact, opportunity creation, disqualification reasons and source-specific conversion quality.

Connect accepted or qualified outcomes back to ad platforms only when identifiers, consent and data quality permit. A high-quality offline conversion import is more useful than a large but inconsistent set of CRM statuses. Keep the scoring and attribution definitions aligned with the primary conversion actions used in advertising.

Govern the model and its limitations

Document the score owner, version, point rules, exclusions, training date, last review and known biases. Retain a manual override with a reason so sales can flag unusual but promising leads. Audit whether model changes disproportionately remove a legitimate customer segment or create a backlog of ignored enquiries.

Judge success by qualified opportunity rate, customer outcomes, sales workload and response quality rather than the number of MQLs created. Pair this work with the marketing lead handoff and attribution-versus-incrementality guides to keep the entire funnel accountable.