What you’re working toward.
A tested fit-and-engagement model connected to a capacity-aware follow-up rule.
- Bring these people
- Marketing operations, sales leadership, and the queue owner
Have these inputs ready
- An agreed ICP and a definition of an accepted sales handoff.
- A sample of accepted, rejected, and inactive leads with outcomes.
- An inventory of reliable properties and trackable engagement events.
Start with a shared definition
Ask sales and marketing which account attributes indicate a good fit and which actions indicate timely interest. Exclude customers, partners, and other records that should not enter the prospect queue.
- Ask sales to label actual examples as good fit, poor fit, or unknown, and explain each decision. Keep unknown separate from poor fit when enrichment is missing.
- Document exclusions before assigning points: customers, active opportunities, competitors, internal contacts, and other non-prospect populations as appropriate.
A scoring brief with the target population, exclusions, and the action the score should trigger.
Separate fit from activity
Keep company or role fit distinct from engagement signals. Check field fill rates before relying on a property. Choose a small set of signals with a clear business rationale rather than rewarding every possible interaction.
- Create a rule register with signal, rationale, source, freshness, and proposed weighting. Check how often the required field is populated in the target population.
- Treat repeated low-intent events differently from a distinct high-intent action. Cap repetition and define how old activity should lose relevance in the model you implement.
Separate fit and engagement definitions, with a reason for every rule.
Test the distribution
Evaluate known good-fit and poor-fit examples, including active non-buyers. Check whether the model concentrates nearly everyone in one band. Tune thresholds against your sales capacity and outcomes, not a round number.
- Run the draft model against the labeled examples. Review the highest-ranked poor-fit records and the lowest-ranked good-fit records before adjusting weights.
- Estimate how many records enter the action band in a normal week. Compare that queue to the team’s follow-up capacity and response commitment.
A calibration sheet showing misclassifications, queue volume, and proposed changes.
Connect score to ownership
Define what happens when a threshold is met, who reviews the record, and when it leaves the queue. Revisit the model against accepted opportunities. Confirm scoring capabilities and limits in your current HubSpot subscription before implementation.
- Give each band an action: immediate review, nurture, research, or exclude. Define the owner, due time, and how an existing task prevents a duplicate handoff.
- Review accepted opportunities and rejection reasons after a full operating cycle. Record the model version so changes do not silently alter historical interpretation.
A routing rule and review cadence tied to observed sales outcomes.
Use a decision matrix before a points model
Illustrative routing logic. These are working bands, not universal point thresholds.
Know when the work is ready.
Use these checks in the handover. Keep evidence of the result and assign an owner to anything that still fails.
- The test set includes active non-buyers and incomplete records.
- A repeated event cannot flood the queue with duplicate tasks.
- The action band produces a workload the team can handle.
- Rejection reasons feed the next calibration review.
Catch the problems early.
A high score is not an opportunity
Scoring prioritizes review. Keep qualification evidence and sales acceptance separate from the numeric result.
Rewarding every click creates noise
Start with a few interpretable signals. A large rule set is hard to diagnose when the wrong records rise to the top.
Product references
Check the supported record types, permissions, and scoring options for your account before translating the model into configuration.