Feature · AI Lead Scoring

AI lead scoring that cites the signals it scored on.

Most AI lead scoring products return a number and ask the sales team to trust it. The useful version returns a number with the top five drivers next to it, the records those drivers came from, and the counterfactual breakdown that shows what would move the score up. Strkr AI runs lead scoring nightly on every inbound lead and every open opportunity, cites the drivers on the record page, and lets the sales ops lead tune the model against ICP without a data science ticket. $50 per seat flat with no per-score meter.

Start free trial See pricing
What AI lead scoring has to do

Seven signals real scoring reads.

The lead scoring products that get shelved are the ones that return a black-box number. The lead scoring products that run are the ones that return a number plus the top five signals that drove the number, cited on specific records. A scoring model that cannot explain itself is a model that nobody trusts at a sales ops standup. The seven signals below cover the common drivers; the model combines them with per-tenant weights that the ops lead tunes without a data science engagement.

Firmographic match

ICP alignment, cited.

The firmographic signal reads industry, company size, revenue tier, headquarters region, and tech stack signals. The score cites the fields that drove the match (for example "fits ICP: SaaS, 500-1000 FTE, North America"). Teams tune the ICP definition on an admin page; the model recomputes nightly without re-training.

Behavioral intent

Product usage, campaign touches, document reads.

The behavioral signal reads product usage events, campaign email reads and clicks, documentation reads, pricing page visits, and demo request forms. The score cites the events that drove the signal (for example "visited pricing page 3x in last 7 days, read documentation on the enterprise tier"). The common failure where a lead score ignores recent behavior does not happen.

Engagement cadence

Reply rate, meeting rate, response latency.

The engagement signal reads reply rate on outbound touches, meeting booking rate, and response latency to rep touches. A lead that replies within four hours scores higher than a lead that replies after a week. The signal is cited on the record so the rep knows why the lead is scored where it is.

Champion signal

Who is engaging, in what role.

The champion signal reads which contacts on the account are engaging and in what role. Engagement from a VP of Engineering on a developer-tool lead scores differently from engagement from a marketing manager. The signal cites the engaging contacts and their roles so the rep knows the champion profile.

Timing signal

Trigger events, budget cycles.

The timing signal reads trigger events (funding announcement, exec hire, product launch, buyout) and known budget cycle timing (fiscal year end, Q4 budget flush). A lead that fits the ICP during an active trigger event scores higher than the same lead at a quiet time.

Custom field weights

The fields that make your business your business.

The custom field signal reads the per-tenant custom fields that the standard model does not know about (industry subcategory, regulatory tier, strategic account flag). Weights are configured per tenant without a data science engagement. The common failure where an off-the-shelf scoring model ignores the fields that drive the business does not happen.

Historical conversion

What your won deals looked like at this stage.

The historical signal reads the pattern of your won deals at the same stage: firmographic distribution, engagement velocity, champion role mix. The model weights signals that correlate with your won deals rather than a generic benchmark. The scoring tunes to your business rather than to the vendor average.

How the model works

The parts of the model the ops lead actually sees.

Lead scoring that nobody can tune becomes shelfware. Strkr AI lead scoring surfaces the model internals to the ops lead through configuration pages rather than Python notebooks. The ops lead sees the signal weights, adjusts the ICP definition, tunes the custom field weights, and the model recomputes nightly. The common failure where lead scoring requires a data science engagement for every tune gets closed off.

Signal weights

Adjustable per tenant.

Signal weights (firmographic, behavioral, engagement, champion, timing, custom, historical) are adjustable through the ops lead admin page. Default weights come from the Strkr benchmark across customers. The ops lead tunes based on business reality. The common failure where the vendor default never matches the actual business shape gets avoided.

ICP definition

Industry, size, geography, stack.

The ICP definition is an admin page with the industry list, company size tiers, geography scope, and tech stack signals. The ops lead edits in UI, not in SQL. The model picks up the change on the next nightly run. Multiple ICPs supported per tenant for teams with distinct segments.

Nightly recompute

Every lead, every opportunity, every night.

The model runs nightly across every inbound lead and every open opportunity. The score lands on the record by Monday morning for the rep to work from. Intra-day re-scoring runs on explicit triggers (new behavioral event, new campaign touch) so recently-moved records show the updated score within the hour.

Explainability panel

Top five drivers on the record page.

The record page shows the top five drivers behind the current score with the cited records. The rep sees why the lead is scored where it is without clicking into a separate tool. The common failure where a scoring tool returns a number the sales team ignores because the number has no explanation gets avoided.

Counterfactual

What would move the score up.

The counterfactual panel shows what would move the score up (for example "a reply from the VP of Engineering would move this to score 85"). The SDR works with context. The rep learns which signals matter in the current segment and the model gets more useful over time rather than more opaque.

Threshold rules

Route by score, with the rep opener drafted.

The ops lead configures threshold rules: score above 70 routes to the enterprise AE queue with the opener drafted, score 50-70 routes to the SDR cadence, score below 50 goes to nurture. The common failure where a hot lead sits in a shared inbox for three days because the threshold routing was manual gets closed off.

What ships with the scoring

Six capabilities that make the score actually useful.

A lead score that lands on the record and does nothing is a wasted calculation. Strkr AI pairs the score with routing, opener drafting, cadence sequencing, pipeline analytics, and model performance tracking so the score drives action rather than sitting on a dashboard. The score is the start of the workflow, not the end.

Score-driven routing

High-score leads to the right queue.

High-score leads route to the enterprise AE queue. Mid-score leads route to the SDR cadence. Low-score leads route to nurture. The routing rules live on the admin page. The common failure where a hot inbound lead sits in a shared inbox for three days because the routing was manual gets closed off at the score boundary.

Opener drafting

First-touch opener in the SDR voice.

For high-score leads, the AI drafts a first-touch opener in the SDR voice that references the signals that drove the score (the pricing page visit, the campaign click, the champion contact). The SDR edits in one minute and sends. First-touch time drops from hours to under a minute.

Cadence sequencing

Score-aware touch pattern.

The cadence sequence adapts to the score. High-score leads get an accelerated multi-channel cadence (email, phone, Strkr Messaging SMS). Mid-score leads get the standard cadence. Low-score leads get the nurture track. The SDR works a sorted list with the cadence already configured.

Pipeline analytics

Score vs win rate, by segment.

The analytics panel shows win rate by score band, by segment, by rep, over the last 90 days. The ops lead sees whether the scoring is predictive. If the top band wins at the same rate as the middle band, the model needs tuning and the panel surfaces that signal rather than hiding it.

Model performance

MAPE, calibration, drift.

The model performance panel reports MAPE on the scoring, calibration between predicted and actual conversion, and drift over the trailing 30 days. The ops lead sees whether the model is holding up or whether a retune is needed. The common failure where "the scores are getting worse" has no answer gets avoided.

A/B shadow

Candidate weights vs live.

Shadow mode runs a candidate weight configuration alongside the live configuration on the same inputs without writing the candidate output. The ops lead reviews a 100-lead sample against the actual conversion outcome and promotes the candidate when it wins. Model improvements ship with evidence.

Pricing and compliance posture

Flat per seat, no per-score meter.

Lead scoring vendors often charge per lead scored or per API call. The teams that score the most pay the most. Strkr prices lead scoring inside the flat $50 per seat. A team scoring 50,000 leads a month costs the same as a team scoring 500. The compliance posture holds across all tiers: HIPAA BAA available, SOC 2 Type II in progress, audit log on every score decision.

No per-score meter

Flat seat covers unlimited scoring.

The lead scoring runs inside the flat seat price. A nightly scan across 50,000 leads costs the same as a nightly scan across 500. The common failure where a vendor meters per-score and the team ends up scoring less than they should because the invoice would land in a conversation with finance gets avoided.

HIPAA BAA

Signed on request for paid tenants.

Strkr signs HIPAA Business Associate Agreements for paid tenants in healthcare. The scoring model runs on the compliant model path for tenants with PHI exposure. The audit log carries the full chain of custody for every PHI interaction. The BAA is available at the paid tier, not behind an enterprise gate.

Audit on every score

Compliance answers in a filter view.

Every score decision lands in the audit log with the signal inputs, the weight configuration, the model version, and the resulting score. A compliance query like "every lead score above 80 in Q3 that did not convert" is a filter on the audit page. The audit log is append-only with cryptographic integrity.

Permission inheritance

Scores respect record permissions.

The scoring model runs with the acting user permissions. A rep can see scores on records the rep can access. A rep cannot see scores on records the rep is denied. The common failure where an AI tier exposes data a human would be denied in the UI does not happen because the authorization layer is shared with the CRM.

Data residency

Model runs in configurable region.

Scoring model inference runs in the configured region (US, EU, APAC) matching the data residency requirement of the tenant. The common failure where "the AI processes data in a region that violates our contract" gets closed off at the model routing boundary. Residency is a tenant setting, not an enterprise upsell.

Export portability

Scores and drivers export via API.

Every score, every driver breakdown, and every signal input exports via REST or GraphQL API. If the team eventually migrates the scoring model to an internal engine, the historical data moves. The vendor relationship stays about value because the exit cost is measurable in script-writing hours.

AI lead scoring that explains every score.

Strkr AI runs lead scoring nightly on every inbound lead and open opportunity. Top five drivers cited on the record page, counterfactual breakdown, score-driven routing, opener drafting. Flat $50 per seat with no per-score meter.

Start free See pricing
Common questions

What buyers ask about this feature.

How does AI lead scoring differ from rule-based lead scoring?

Rule-based scoring assigns points for discrete events (10 points for a demo request, 5 points for a pricing page visit). AI lead scoring combines firmographic match, behavioral intent, engagement cadence, champion signal, timing, custom fields, and historical conversion pattern into a model that tunes to the business shape. The AI version is more accurate across segments because the weights are learned from the actual won deals rather than guessed. The explainability panel shows the top five drivers so the SDR can trust the number rather than ignoring it.

Can the ops lead tune the model without a data science engagement?

Yes. The ops lead admin page covers signal weights, ICP definition, custom field weights, and threshold routing rules. Changes picked up by the nightly run. For teams that want to validate a tune before promoting it, A/B shadow mode runs a candidate configuration alongside the live configuration on the same inputs without writing the candidate output. The ops lead reviews a 100-lead sample against actual conversion outcomes and promotes when the candidate wins. The common failure where lead scoring requires a data science engagement for every tune gets closed off.

How is the score explained on the record?

The record page shows the numeric score plus the top five drivers that contributed to it, with the cited records (for example, the VP of Engineering replied within 4 hours, the lead visited the pricing page three times in the last 7 days, firmographic match on SaaS at 500 to 1000 FTE in North America). The counterfactual panel shows what would move the score up (for example, a reply from the CFO would move this to a score of 85). The SDR works with context rather than ignoring an opaque number.

How often does the model recompute scores?

The model runs nightly across every inbound lead and every open opportunity. Scores land on records by Monday morning. Intra-day re-scoring fires on explicit triggers like new behavioral events (pricing page visit, demo request form submission) and new campaign touches, so recently-moved records reflect the updated score within the hour. The nightly cadence covers the slow signals; the intra-day triggers cover the fast ones.

How does Strkr handle AI lead scoring for regulated industries?

Strkr signs HIPAA Business Associate Agreements for paid tenants in healthcare. For tenants with PHI exposure, the scoring model runs on the compliant model path with sealed logs. The audit log carries the full chain of custody for every score decision: signal inputs, weight configuration, model version, and resulting score. Data residency is configurable per tenant (US, EU, APAC). SOC 2 Type II audit is in progress with Type I available on request. The compliance posture is shared with the CRM, not invented separately for the scoring tier.

Does AI lead scoring ship with a per-score meter?

No. Strkr is $50 per seat flat with no per-score meter, no per-lead meter, and no per-API-call meter. A team scoring 50,000 leads a month costs the same as a team scoring 500. The common failure where a scoring vendor meters per-score and the team ends up scoring less than they should because the invoice would land in a conversation with finance does not happen. The seat price on the pricing page is the full price for the lead scoring capability.

Related features

Pair this with...

Lead Scoring Lead Scoring Models Lead Scoring Criteria Examples Lead Scoring Engine Criteria B2B Lead Scoring Criteria Predictive Sales Analytics Lead Routing Rules AI Sales Assistant

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