Feature · Scoring Models

Effective lead scoring models that reps actually use.

Lead scoring works when the model is simple enough to defend, honest enough to argue with, and tuned to the sales team's real motion. The four model shapes below (rule-based fit, intent-weighted, engagement decay, Strkr AI model-called) are the ones that actually move pipeline. The common mistakes that collapse scoring programs are explicit so new teams avoid them from the start.

Start free trial See pricing
The four model shapes

What an effective lead scoring model actually looks like.

Lead scoring programs fail more often than they succeed. The common failure modes are over-complexity (a 60-signal model nobody can defend), rule drift (nobody updated the model in a year, the business changed), and model-worship (the score is treated as truth and the rep stops using judgment). The four model shapes below are the ones that avoid those traps and produce scores reps actually act on.

Rule-based fit

Firmographic match to the ICP.

The rule-based fit model scores leads on firmographic match: industry, revenue tier, employee count, country, department of the contact. The rules are explicit and defendable: a industry match adds 20 points, a revenue tier above threshold adds 15, a target department adds 10. The score is reproducible and the sales leader can argue with each rule.

Intent-weighted

Behavior that signals buying intent.

The intent-weighted model scores on behavior that correlates with buying intent: pricing page view, pricing page view twice in a week, demo-request form submission, competitor comparison page view, specific whitepaper download. The weights come from historical close-rate correlation rather than guesswork. The model is honest about which behaviors actually predict purchase.

Engagement decay

Recent activity worth more than old.

The engagement decay model weights recent activity higher than old activity. A pricing page view today is worth more than one three months ago. The decay curve is exponential rather than linear, which reflects the real-world pattern: a prospect who went silent six months ago is a different lead than one who was active yesterday.

Strkr AI model-called

A pattern-matched score, with reasoning.

The Strkr AI model-called score reads the full lead context (fit, intent, engagement, historical pattern) and renders a 0-to-100 score with one-line reasoning. The score is pattern-matched against the tenant's own closed-won and closed-lost history, so it calibrates per tenant. The reasoning is explicit so the rep can argue with it.

Negative scoring

Signals that remove points.

Effective models include negative signals: free-email domain, student or academic domain, competitor-employee domain, country outside the sales territory. Negative signals prevent high-engagement bad-fit leads from scoring at the top of the queue. The pattern is counterintuitive to marketers who only think in positive weights but essential for sales productivity.

Composite

The four models, blended.

Most mature programs run a composite: fit (30 percent weight) plus intent (30 percent) plus engagement decay (20 percent) plus Strkr AI model-called (20 percent). The composite captures the strengths of each lens and reduces the brittleness of any single model. The blend is tenant-configurable rather than vendor-fixed.

Threshold-based routing

Score above threshold becomes an MQL.

The score drives the handoff. A composite above 70 becomes a marketing-qualified lead and routes to sales. A score below 40 stays in nurture. A score between 40 and 70 enters a review queue for an SDR to manually qualify. The threshold is tenant-configurable per segment so enterprise and SMB can have different cutoffs.

Transparency

The score shows its math.

Every score shows its components: fit contribution, intent contribution, engagement contribution, model-called contribution, negative contribution. The rep sees why the score is what it is and argues with specific components rather than the number as a whole. The pattern turns the score into a working artifact rather than a mystery.

Audit the model

A quarterly backtest, honest about miss.

Every quarter, the model gets backtested: did the leads scored above 70 convert at a higher rate than those below? Did the top decile close at a better rate than the second decile? The backtest is the discipline that keeps the model honest. Models that drift from the backtest get recalibrated rather than defended.

The common mistakes

What kills lead scoring programs in month six.

Lead scoring is more frequently abandoned than it is tuned. The failure modes are predictable and the counter-patterns are specific. The design points below are the ones that keep a scoring program alive past the first quarter.

Over-complexity

A 60-signal model nobody defends.

Teams build models with every signal they can think of: 60 rules, 20 negative signals, three decay curves, five composite weightings. The resulting score is a black box that the sales leader cannot argue with and the rep stops trusting after the first bad referral. The counter-pattern is radical simplicity: six to ten signals, maximum.

Static rules

Nobody updates the model.

The model is set at launch and never updated. The business changes (new ICP segment, new product line, new motion) and the model lags. The counter-pattern is a quarterly review cycle where sales ops, marketing, and the senior sales leader walk the model, audit the backtest, and update the weights.

Model worship

The score is treated as truth.

Teams treat the score as the final word rather than an input to rep judgment. The rep stops asking questions and the model's drift goes unnoticed. The counter-pattern is explicit framing in rep training: the score is an input, not a verdict; the rep uses it to prioritize, not to decide.

Missing negatives

High engagement from bad fits.

Models that only score positive signals surface high-engagement bad-fit leads at the top: a student doing a research project, a competitor employee doing competitive intelligence, a prospect in an out-of-territory country. The counter-pattern is explicit negative signals that remove points for known-bad patterns.

No decay

Old activity worth the same as new.

Models without decay treat a pricing-page view from six months ago as equivalent to one from today. The resulting score is wrong for the way sales motions actually move. The counter-pattern is explicit exponential decay on engagement signals.

No threshold

Every scored lead goes to sales.

Without an explicit threshold, every scored lead flows to sales, which drowns the sales team in low-quality leads and burns the trust in the scoring system entirely. The counter-pattern is a composite threshold (typical: 70) above which leads route to sales, below which leads stay in nurture.

No feedback loop

The sales result does not update the model.

Models that do not learn from the sales result stay stale. The counter-pattern is to capture the sales outcome (closed-won, closed-lost, closed-no-decision) and feed it back into the next quarter's model tuning. The Strkr AI model-called score automates this feedback loop.

Vendor lock-in

The score lives in a black box.

Vendor-provided scoring models that live in proprietary black boxes are impossible to tune, audit, or export. The counter-pattern is explicit scoring rules that live in the CRM, auditable in the admin UI, exportable through the open API, and portable if the team ever changes platforms.

One model for everyone

The same score for every segment.

Models that score the same across segments miss the segment-specific patterns. A high-engagement signal in SMB is different from one in enterprise. The counter-pattern is segment-aware scoring, where the model weights adjust based on the segment the lead resolves to.

Three model templates

Starting points for three common motions.

A blank model is a research project. The three starting templates below are the ones Strkr customers typically start with and tune over the first two quarters.

Mid-market B2B SaaS

Fit plus intent, decay, AI overlay.

The mid-market B2B SaaS template weights fit (industry, revenue tier, employee count) at 30 percent, intent (pricing page, demo request, specific whitepaper) at 35 percent, engagement decay at 15 percent, and the Strkr AI model-called score at 20 percent. Negative signals include free-email domain, country outside territory, and competitor-employee domain. The threshold for MQL routing is 70. The template fits most B2B SaaS motions out of the box.

Enterprise outbound

Fit-heavy with account-level signals.

The enterprise outbound template weights fit at 45 percent (industry, revenue tier, named-account match), account-level engagement (multiple contacts active, executive engagement) at 25 percent, intent at 15 percent, and Strkr AI model-called at 15 percent. The threshold is higher (80) because enterprise cycles cannot absorb low-quality routing. The pattern suits teams running a targeted-account motion.

Product-led growth

Product usage as the primary signal.

The product-led growth template weights product usage (feature touches, user count, workspace activity) at 50 percent, fit at 20 percent, intent at 15 percent, and Strkr AI model-called at 15 percent. The threshold is 60 because PLG motions often qualify leads earlier than enterprise. The pattern suits teams where product adoption predicts purchase intent better than marketing engagement.

SMB inbound

Engagement-weighted, fast decay.

The SMB inbound template weights intent at 40 percent, fit at 25 percent, engagement decay at 25 percent (steeper decay curve because SMB cycles are shorter), and Strkr AI model-called at 10 percent. The threshold is 60 because SMB motions can absorb more volume. The pattern suits teams running high-volume inbound where cycle length is weeks rather than months.

Industry-specialized

Vertical-specific signals.

For teams selling into a vertical (healthcare, financial services, legal), the template adds vertical-specific signals: compliance certifications, EHR integrations, regulatory keywords. The Strkr AI model-called score helps surface the vertical patterns the rule-based model misses. The threshold adjusts to the vertical's typical cycle length.

Account-based marketing

Named-account scoring takes precedence.

For teams running account-based marketing, the template elevates named-account match to a dominant weight (50 percent) with the account-level signals (multiple engaged contacts, executive engagement) at 30 percent and everything else at the remainder. The pattern inverts the usual scoring: a low-engagement contact at a named account outscores a high-engagement contact at a non-target account.

Implementation and governance

How to roll out a scoring program that sticks.

A scoring program sticks when it has an owner, a review cadence, and a visible audit trail. The design points below are the ones that keep the program alive past the first quarter.

Named owner

One person accountable for the model.

The scoring model has a named owner, typically the senior sales ops lead or the demand gen leader. The owner answers for the model's performance, drives the quarterly review, and signs off on weight changes. The pattern prevents the “nobody owns the scoring model” diffusion that kills programs in month six.

Quarterly review

A disciplined tuning cadence.

Every quarter, the owner, the sales leader, and the marketing leader walk the model, review the backtest, and update the weights. The review is a two-hour meeting with a agenda: top deciles conversion rate, missed opportunities below threshold, model drift. The cadence keeps the model honest without turning it into a weekly debate.

Backtest discipline

The honest look at the model's accuracy.

The backtest compares the model's scores against actual sales outcomes. Did leads scored above 70 close at a higher rate than those below? Did the top decile outperform the second decile? The backtest is run quarterly and the result is in the review. Models that drift from the backtest get recalibrated rather than defended.

Rep feedback loop

Reps mark the misses.

Reps have a one-click surface to flag a score as “obviously wrong” (too high or too low). The flagged leads feed into the quarterly review as evidence of model drift or specific gaps. The pattern turns reps from skeptics into model-tuning collaborators.

Model versioning

Every change tracked.

Every model change (weight adjustment, rule addition, threshold shift) logs with the owner, time, and reason. The audit trail is important for the quarterly backtest and for the “the model is suddenly surfacing different leads” investigation that occasionally happens. Model versions are reversible if a change does not perform.

Communication

Reps and marketers know what changed.

When the model changes, reps and marketers get a notification with the diff and the expected impact. The transparency keeps the “the model is doing something weird” complaints down and the collaboration up. The pattern turns scoring into a shared discipline rather than a sales-ops secret.

Compliance posture

Scoring under privacy regulations.

Lead scoring touches personal data and automated-decision regulations. The posture below is explicit.

GDPR

EU residency, SCCs, transparency.

Strkr is GDPR compliant with Standard Contractual Clauses on file and an EU-hosted path for EU tenants. For EU recipients, scoring is subject to transparency obligations: the recipient has a right to know the logic of significant automated decisions. Strkr's scoring is explicit (fit, intent, engagement, model-called, with components surfaced), which meets the transparency bar.

CCPA

Consumer rights honored.

Strkr honors CCPA data-subject rights across the scoring surface. A deletion request removes the subject's score history within the regulatory window. A right-to-know request returns the scoring components and the current score.

HIPAA

Compliant, BAA available.

Strkr is HIPAA compliant and executes a Business Associate Agreement on request. For healthcare tenants, scoring runs under the BAA with the field-level mask on sensitive fields respected. PHI-adjacent signals can be used in scoring under the BAA scope.

SOC 2 Type II

In progress through 2026-Q4.

Strkr is actively pursuing SOC 2 Type II with report completion targeted through 2026-Q4. The scoring engine, model configuration, backtest surface, and audit log are all in scope. Interim artifacts are available under NDA.

Open API

Scores in the public API.

Every score and its components are exposed through documented public API endpoints. A customer who wants to pipe scores into a parallel BI tool, a routing engine, or a data warehouse can do so with a documented call. Open API, no black box.

Explainability

The score shows its reasoning.

Every Strkr AI model-called score includes a one-line reason that explains the top three drivers. The transparency meets the EU's explainability bar for significant automated decisions and the general trust bar every scoring program needs to earn adoption.

A scoring model reps actually use, inside the CRM, on every paid tier.

Fit, intent, engagement decay, Strkr AI model-called, and composite blending ship with Strkr on every paid tier. Rule-based scoring, segment-aware weighting, named-account elevation, and the open API are included. Start a trial and configure your first lead scoring model before the end of the week.

Start free trial See pricing
Common questions

What buyers ask about this feature.

What are the most effective lead scoring models for a B2B sales team?

The four model shapes that actually move pipeline are rule-based fit (firmographic match to the ICP), intent-weighted (behavior that signals buying intent), engagement decay (recent activity weighted higher than old), and Strkr AI model-called (pattern-matched against the tenant's own historical closed-won and closed-lost). Most mature programs run a composite: fit 30 percent, intent 30 percent, engagement decay 20 percent, Strkr AI 20 percent. Negative signals (free-email domain, out-of-territory country, competitor-employee domain) prevent high-engagement bad-fit leads from topping the queue.

Why do lead scoring programs fail by month six?

The common failure modes are over-complexity (a 60-signal model nobody can defend), static rules (nobody updates the model when the business changes), model worship (the score becomes truth and the rep stops thinking), missing negatives (high engagement from bad fits tops the queue), no decay (old activity worth the same as new), no threshold (every scored lead floods sales), no feedback loop (the sales result does not update the model), and vendor lock-in (the model lives in a black box). The counter-patterns are explicit: six to ten signals maximum, a named owner, quarterly review cadence, explicit negative signals, exponential decay, threshold-based routing, outcome feedback into the next quarter, and open API portability.

Should I use rule-based scoring or AI scoring?

Both. Rule-based scoring is defendable and tunable; the sales leader can argue with each rule. AI scoring is pattern-matched and captures the subtle correlations that rules miss. The right pattern is a composite of both: rule-based for the explicit components (fit, intent, engagement decay) at 60-70 percent of the weight, Strkr AI model-called for the pattern-matched component at 20-30 percent of the weight. The composite captures the strengths of each lens and reduces the brittleness of any single model. Models that worship either the rules or the AI exclusively tend to drift.

How often should we tune the lead scoring model?

Quarterly. The quarterly review walks the backtest (did leads above threshold convert at a higher rate), the top-decile performance (does the top 10 percent outperform the second 10 percent), the model drift signals (changes in the signal distribution), and the rep feedback flags. Weight changes are explicit and tracked with the owner, time, and reason. Weekly or monthly tuning typically produces false signal; annual tuning typically produces drift. The quarterly cadence hits the right balance for most B2B motions.

How does segment-aware scoring work?

Segment-aware scoring adjusts the model weights based on the segment the lead resolves to. A high-engagement signal in SMB is different from one in enterprise. The SMB model might weight engagement at 40 percent of the composite; the enterprise model might weight engagement at 15 percent and fit at 45 percent. Strkr supports per-segment weightings in the admin UI, with the sales ops lead configuring the segments and the weights. The pattern lifts conversion rate materially over a single-model-for-everyone approach.

Is Strkr's lead scoring HIPAA, SOC 2, and GDPR compliant?

Strkr is HIPAA compliant with a BAA available on request, GDPR compliant with Standard Contractual Clauses on file and an EU-hosted path for EU tenants, and actively pursuing SOC 2 Type II with report completion targeted through 2026-Q4. The scoring engine, model configuration, backtest surface, and audit log are all in scope. Every Strkr AI model-called score includes a one-line reason, which meets the EU's explainability bar for significant automated decisions and the general trust bar every scoring program needs to earn adoption.

Related features

Pair this with...

Define sales forecasting Drip email automation Dialer for cold calling

Try it free. Bring your team next week.

No sales call, no migration consultant, no four-month implementation. Enter your card, get 14 days of the full Pro tier, cancel any time before day 14 with zero charge. Spin up a workspace, import your CSV, and have something useful before lunch.

Start free trial See pricing