Feature · Lead Scoring

Lead scoring that tells reps who to call, and why.

Fit score plus engagement score plus Strkr AI predictive score, running on every lead in the system. Score decay, threshold routing, and model explainability built in. No data-volume floor, no tier gate on the predictive side, no middleware stack to glue it to the CRM.

Start free trial See pricing
What lead scoring actually gets right vs wrong

The honest view most buyers never hear in a demo.

Lead scoring is one of the most oversold features in the CRM category. Every tool claims it. Most teams who adopt it either turn it off within a quarter or run it as decoration while reps ignore the score entirely and work the queue by instinct, account name, or whichever lead the manager yelled about most recently. The reason is almost never the math. The reason is that nobody set the model up to answer the question reps actually ask: should I call this person right now, and if yes, what do I say when they pick up. The honest buyer view starts by naming what a score is for, what the real failure modes look like, and what a score should never try to do. Everything else in this page assumes that foundation.

What scoring is for

Rank the inbound queue, not predict destiny.

A lead score is a ranking signal, not a verdict. The job is to tell a rep which five of today's thirty inbound leads to call first and in what order. It is not the job of the score to decide whether a lead will ever close, who should own it long term, or whether the deal will be big or small. Teams that treat the score as a prophecy burn out on it inside a quarter. Teams that treat it as a sort order keep using it for years without a single complaint.

What scoring is not for

It cannot replace discovery.

A high score does not mean a rep should skip qualification. It means the lead probably deserves the next sixty seconds of attention. Discovery still has to answer whether there is a budget, a buyer, a timeline, and a pain. The score opens the door. The rep walks through it. Teams that confuse the two end up losing deals the model said were slam dunks.

What usually breaks

Rules written once, never revisited.

The most common failure mode is a scoring model built on day one of CRM adoption, then left alone for three years. Industry shifted. Buying motion changed. New product launched. The rules that scored last year's pipeline well now score noise and the top of the queue slowly fills with leads that will never close. Scoring needs the same quarterly review as a comp plan or a territory map, and no model survives without it on the calendar.

What rules miss

Combinations beat any single signal.

A lead who visited the pricing page three times is a weak signal on its own. A lead who visited the pricing page three times AND is a VP of Sales at a company with 200 employees AND opened the last two emails within 48 hours is a very strong signal. Rule-based scoring handles the first case easily. Predictive scoring finds the combinations rules were never written for.

What AI gets wrong

Black-box scores reps will not trust.

Reps ignore any score they cannot explain. A black-box predictive model that says a lead scored 87 but will not say why gets routed around by the AE within a week. The score becomes decoration. Explainability is not a nice-to-have. It is the difference between a model reps use and a model reps avoid. Strkr AI returns the top three contributing signals for every score, every time, with the signal values exposed in the UI the rep already looks at.

What good looks like

Rules plus AI plus explainability.

A good scoring system runs rule-based fit and engagement scores for transparency, runs predictive scoring in parallel for the combinations rules miss, and shows the rep why the score is what it is. The rep sees a number, the signals behind it, the trend over the last 30 days, and the recommended next action. That is a model reps keep using through five quarters, three comp-plan rewrites, and two territory reshuffles.

How Strkr handles lead scoring

Three scores, one record, one honest number.

Every lead in Strkr carries three scores: fit, engagement, and Strkr AI predictive. The three combine into one overall score that drives the queue, routing, and the rep view. Each component is explainable on its own, so the rep can see fit was 42 of 60 and engagement was 18 of 30 and the AI add-on nudged it another 7 because of a specific identified pattern. The components ship on every paid tier, with no data-volume floor and no gate on the predictive piece. The sections below walk the nine building blocks that make the full scoring surface work the way reps and ops teams actually want it to.

Fit score

Demographic and firmographic signals.

Score demographic fields on the lead (title, seniority, department) and firmographic fields on the account (industry, headcount, revenue band, geography, tech stack). Each criterion gets a weight. The fit score adds up and caps at a configurable max. The result is the "would this lead ever buy from us" number, independent of what they have done so far.

Engagement score

What the lead has actually done.

Email opens, link clicks, page views, pricing-page visits, demo-video views, form fills, chat sessions, webinar attendance, inbound replies. Each event earns points. Events can carry different weights depending on how high-intent they are. The engagement score answers "is this lead actually paying attention right now" without needing a human to look.

Strkr AI predictive

The combinations rules miss.

Strkr AI reads the full lead and account graph, compares it against historical closed-won versus closed-lost patterns, and returns a probability score. Combinations of signals that no rule writer would ever think to encode show up here. The predictive score ships on every paid tier and does not require a minimum data volume to turn on.

Model explainability

Every score shows its work.

Click the score on any lead to see the top signals that drove it. Pricing-page visit on Tuesday worth 12 points. Title match worth 8 points. Industry match worth 10 points. Strkr AI contribution worth 7 points, driven by "similar closed-won accounts in the last 90 days." The rep sees the number and the reason in one click.

Score decay

A lead that goes cold stops being hot.

Engagement points decay on a configurable half-life. A pricing-page visit from 60 days ago is worth less than one from yesterday. The decay runs nightly. A lead that was hot in Q2 but went quiet is not still ranked at the top of the queue in Q4 just because they clicked a link once. The score reflects reality now, not reality then.

Threshold routing

Score crosses a line, a flow fires.

Set a score threshold. When any lead crosses it, fire a flow. Common patterns: score above 80 routes to an AE in a Slack notification, score above 60 drops to a nurture track, score below 20 flips to disqualified. Routing on score threshold is wired through the same Flows engine every other automation uses, so there is one audit trail.

Score history

The timeline, not just the current number.

Every lead carries a score history line chart over the last 90 days. Rep opens the lead, sees the score moved from 32 to 71 in the last 30 days with the steepest climb in the last week. That trend is more useful than the current number alone. Reps call the lead whose score is climbing fast, not the lead whose score has been high and flat for six months without any new engagement to justify it.

Negative scoring

Points off for the right reasons.

A free-email domain, a role title of student, a company size of one, a bounce on the last email, a competitor-domain submission. All subtract points. Negative scoring is not a punishment track. It is the model being honest about the fact that some signals make a lead less interesting to the revenue motion, and ignoring them inflates every score in the system and devalues the top of the queue.

Multi-model

Different scores for different motions.

A self-serve SaaS lead and a strategic enterprise lead do not share the same scoring model. Define a separate scoring config per segment, per product line, or per territory. The right lead gets the right rubric. One model does not have to be the compromise between two very different buyer profiles, and the top of one segment's queue does not have to compete with the top of another segment's queue on an incompatible scale.

Rule-based vs AI vs hybrid

Three approaches, and when each one wins.

Every scoring vendor positions their approach as the right one. Rule-based vendors claim transparency. AI vendors claim lift. The truth is each approach wins in a different scenario, and the honest answer is almost always a hybrid. The question is not whether to use rules or AI. The question is which signals are reliable enough to encode as rules, which signals only emerge as combinations the model finds, and how to combine the two without losing the explainability reps need. The next six cards walk the trade-offs so a buyer can hold a scoring vendor accountable to the right standard.

Rule-based wins

Transparent, slow-moving signals.

Fit criteria that rarely change (industry, title, headcount, geography, tech stack, revenue band) belong in rules. Reps and marketers can read a rule, understand it, and tune it quarterly. The transparency is worth more than any marginal lift a model could add on signals this stable. Fit scoring should almost always be rule-based for exactly this reason, and any vendor that insists on treating fit as a black-box prediction is overcomplicating a problem that does not need it.

AI wins

Combinations and non-linear patterns.

Combinations of signals that no human would think to encode (industry X plus title Y plus visited page Z within 48 hours of event W, three touchpoints in a 72-hour window across two different channels, pricing-page bounce followed by a return visit three days later) are what AI predictive scoring is actually good at. It finds the pattern because it is reading closed-won versus closed-lost outcomes, not because someone wrote a rule saying "this combination matters." Humans do not write rules this specific. Models find them routinely.

Hybrid wins

Transparency plus pattern discovery.

The best scoring systems run both. Rules handle fit, so the model is explainable. AI handles the behavioral combinations rules would never catch. The two components ladder into one overall score, with each component visible to the rep. The rep sees fit was 42 of 60, engagement was 18 of 30, and the AI piece added 7 on top because of a specific pattern. That is the shape Strkr ships: rules you can read, AI that finds what rules miss, and an overall number everyone agrees on.

HubSpot comparison

Rules on all tiers, predictive on Enterprise.

HubSpot ships manual lead scoring on all paid tiers. Their predictive scoring (HubSpot score) is gated behind Marketing Hub Enterprise, roughly $3,600 per month billed annually. For a growing team, the practical result is rules-only until the budget supports Enterprise, which usually means never. Strkr ships both on every paid tier.

Salesforce comparison

Einstein needs data volume to work.

Salesforce Einstein Lead Scoring is powerful, but it requires a minimum of 1,000 closed-won leads and 1,000 closed-lost leads in the last 6 months to activate. Growing teams rarely have that volume. Einstein also requires Sales Cloud Enterprise or above, plus the Einstein add-on. Strkr predictive runs with no data-volume floor on every paid tier.

MadKudu or 6sense

Bolt-on, expensive, out-of-band.

Dedicated predictive scoring vendors do good work, but they sit outside the CRM. The score is pushed in via integration, which means a delay between behavior and score update. Explainability lives in a separate app the rep has to context-switch to. Pricing starts at $20,000 per year and climbs quickly past six figures as volume grows. For teams that need scoring inside the CRM the rep already lives in, a native build with explainability lands cleaner and costs an order of magnitude less.

The honest trade

Native is cheaper, bolt-on is more specialized.

A dedicated scoring vendor will outperform a native scoring feature on the margin for an enormous enterprise with billions of touchpoints and a dedicated data-science team to tune the model. For every team under that ceiling (which is almost every team), the practical gain from a specialized vendor is smaller than the gain from having the score live inside the CRM the rep already uses, update instantly on every event, and explain itself in the record view. Strkr targets the second category directly and does not pretend to compete with hyperscale ABM platforms.

Common lead scoring failures and how to avoid them

Seven failure modes that break every scoring project.

Every scoring project runs into the same seven failure modes within the first year. Teams that know the failure modes in advance survive them. Teams that do not know them turn the scoring off, forget to turn it back on, and six months later wonder why reps are still working the queue in alphabetical order by account name. The failures are model drift, bad signal, missing score decay, overfitting, rule bloat, dual sources of truth, and score hoarding. All seven are preventable with the right instrumentation, and the Strkr admin surfaces every one of them so the quarterly review takes an hour instead of a day.

Model drift

Yesterday's model, today's pipeline.

The scoring rules you wrote six months ago reflected six-month-old buyer behavior. The product changed. The ideal customer profile narrowed. The sales motion shifted upmarket. The rules now score noise. Strkr tracks score-to-conversion correlation per rule weight and surfaces drift in the admin view with a red flag on any rule where the correlation has dropped more than 15 points. A quarterly review takes 20 minutes instead of a full-day workshop.

Bad signal

Tracking a thing nobody actually does.

A rule that gives 10 points for watching the demo video is useless if nobody watches the demo video. Low-volume signals look authoritative on paper and contribute nothing in practice. Strkr's scoring admin shows event volume alongside the rule weight, so a signal that fires on 2 percent of leads gets caught before it ships into production. Rules below a volume floor are flagged automatically and surface in the quarterly review queue.

No score decay

Hot leads from 2024 that are not hot anymore.

Without decay, every event ever recorded still contributes to the score. A pricing-page visit from 18 months ago still ranks a lead above a brand-new, actively engaged lead. Reps notice this within a month, stop trusting the score, and start working the queue by instinct again. Strkr ships decay on by default with a sensible half-life per event type, so the score stays current without a batch job for someone to maintain or forget to run.

Overfitting

A model that only works on last quarter's leads.

A predictive model trained too tightly on past data starts recognizing the quirks of the past instead of patterns that generalize. Strkr AI holds back a validation set on every retrain, measures lift on the held-out set, and refuses to promote a new model that does not beat the old one by a configurable margin. A regression ships only when a human admin forces it with an explicit override. The default behavior is "keep the current model until the new one is demonstrably better."

Rule bloat

Fifty criteria, nobody can audit.

A scoring model with 50+ rules nobody can read is as bad as a black-box model nobody can audit. Strkr caps the practical signal set at the top 10 to 15 criteria and reports which rules fire most and which rules barely fire at all. Prune the bottom quartile every quarter. A smaller, tuned model always outperforms a sprawling, untouched one, because the signal-to-noise ratio stays high and every rule still earns its keep.

Dual sources of truth

Marketing scores one way, sales scores another.

If marketing automation computes an MQL score and the CRM computes a separate lead score, the two will disagree. Reps lose trust in both. Finance builds a report that uses whichever one is cheaper to query, and nobody can reproduce it a quarter later. Strkr computes one score inside the CRM, exposes it via API to the marketing side, and uses the same number everywhere. One scoring surface, one audit trail, one argument to have in the quarterly review.

Score hoarding

Threshold set so high nothing qualifies.

A team sets the MQL threshold at 90 and watches the queue stay empty. The SDRs run out of leads to call. The AEs stop trusting the handoff. Thresholds should be set so that the top 15 to 20 percent of inbound crosses the line on a normal week. Strkr's scoring admin shows the historical distribution of scores, so the threshold is calibrated against reality instead of guessed from a round number that felt rigorous in the planning meeting.

Three real lead scoring playbooks

What production scoring actually looks like.

Six scoring configurations teams run on Strkr today. Each one is the full picture: the fit rules, the engagement weights, the AI piece, the threshold, the routing flow, and the rep-side view. The point of showing them all is to anchor the pattern. Lead scoring is not one abstract thing applied uniformly to every revenue motion. It is a handful of concrete motions, each with its own shape, each tuned against a specific buyer profile. The six playbooks below cover the motions most Strkr tenants start from, and most teams combine two or three of them under the multi-model config.

SMB SaaS inbound

Pricing-page intent plus free-trial signal.

Fit scored on company size (10-200 employees), title (manager+), and tech-stack match. Engagement scored heavily on pricing-page visits and free-trial sign-ups. Strkr AI adds a lift on the "similar accounts that closed in last 90 days" signal. Threshold at 65 routes to an SDR in Slack within two minutes. 7x conversion lift reported for sub-hour contact.

Mid-market B2B outbound

Fit-weighted, engagement-watched.

Fit dominates the model (industry, revenue band, geography, headcount). Engagement is a secondary tiebreaker. Strkr AI contributes a probability score based on account-level patterns, not individual lead patterns. Threshold at 55 routes to a named AE based on territory. Score history chart drives the AE's weekly prospecting plan.

Enterprise consulting

ABM scoring across the account.

Scoring runs at the account level, not just the lead level. Every known contact at an account rolls their engagement up into the account score. Fit scoring is strict (industry, revenue above $100M, geography). Strkr AI finds buying-committee patterns (multiple titles active, multiple touchpoints within 30 days). Threshold at 70 fires a flow that notifies the AE and the account exec.

PLG freemium

Product usage as the dominant signal.

A freemium or free-trial product has one scoring signal that beats everything else: in-app behavior. Strkr pulls product events through the Flows engine, scores leads on usage milestones (second session, feature X activated, invited a teammate), and routes to a growth AE when a usage threshold crosses. Fit scoring runs as a tiebreaker on company size and title.

Partner-sourced leads

Weighted by partner source quality.

Not every partner-sourced lead carries the same quality. Strkr adds a partner-source weight to the engagement side of the model, calibrated against historical conversion rates by partner. The top partners' leads get a boost. Low-quality partners' leads get a penalty. The weight is reviewed quarterly against actual closed-won rates per partner, not vibes.

Reactivated leads

Old leads that came back warm.

A lead that disqualified six months ago but just visited pricing twice this week needs a different motion than a brand-new lead with the same score. Strkr tags re-engaged leads with a reactivation marker, boosts the engagement side of their score, and routes them to the original owner first if that owner is still active. Historical context (the last disqualification reason, the prior rep notes, the activity timeline) does not have to be lost just because a lead went cold for a quarter.

Lead scoring on every paid tier: rules, AI, explainability, no gate.

Starter ships with rule-based fit and engagement scoring. Pro adds Strkr AI predictive scoring and multi-model configs. Scale and Enterprise lift the signal cap and add account-level ABM scoring with buying-committee pattern detection. The predictive piece has no data-volume floor and ships on Pro, which means a team in year one gets the same scoring surface a legacy CRM customer pays Enterprise plus an add-on to assemble.

Start free trial See pricing
Common questions

What buyers ask about this feature.

How is Strkr lead scoring different from HubSpot or Salesforce lead scoring?

HubSpot ships manual lead scoring on all paid tiers but gates predictive scoring (HubSpot score) behind Marketing Hub Enterprise, roughly $3,600 per month billed annually. For a growing team, that gate usually means the predictive piece never gets turned on. Salesforce Einstein Lead Scoring requires Sales Cloud Enterprise or above plus the Einstein add-on, and will not activate until the tenant has at least 1,000 closed-won and 1,000 closed-lost leads from the last 6 months, which rules out most teams under $50M in revenue. Strkr ships rule-based fit scoring, rule-based engagement scoring, and Strkr AI predictive scoring on every paid tier, with no data-volume floor on the predictive side and model explainability built into every score. The combined effect is that a team in year one gets the same scoring surface a Salesforce or HubSpot customer has to pay six figures a year to assemble.

Does Strkr AI predictive scoring need a minimum data volume to turn on?

No. The predictive model activates on the first paid tier regardless of how many closed-won or closed-lost leads the tenant has. Early-stage teams see useful lift from day one because the model uses account-graph and behavioral patterns, not just raw outcome volume. The predictive scoring does not try to run a cold-start regression on 20 historical deals. Instead, Strkr AI reads enrichment signals, engagement patterns, and account-graph structure and applies patterns learned across similar segments. As your own closed-won history grows past the 100-record mark, the model begins blending tenant-specific outcome patterns into the score. Past 500 closed-won records, the tenant-specific signal dominates. There is no gate, no 1,000-record floor, no six-month warm-up period, and no manual retrain-request process to open the predictive piece.

How does score decay work and can I turn it off?

Engagement points decay on a configurable half-life. By default, a point earned today is still worth a point tomorrow, worth about 0.7 points in 30 days, and worth less than 0.3 points in 90 days. The decay runs nightly across all leads in a background job that writes atomically alongside the normal score recompute. Fit points do not decay by default, because industry and title do not go stale the same way behavior does. Admins can tune the half-life per event type (pricing-page visits can decay slower than generic email opens, demo-video views can be weighted to decay very slowly). The decay can be turned off per event type, but turning it off globally is strongly discouraged because every scoring model that abandons decay drifts into a state where every lead with a long tenure in the CRM ranks above every freshly-engaged lead, which is the opposite of what the queue should surface.

Can reps see why a lead scored what it scored?

Yes. Click any score on any lead to see the top contributing signals and their point values. Pricing-page visit on Tuesday worth 12 points. Title match worth 8 points. Industry match worth 10 points. Strkr AI contribution worth 7 points, with the AI piece naming the pattern that drove the lift (for example, "similar closed-won accounts in the last 90 days visited pricing within 72 hours of a product-demo view"). Explainability is on by default and cannot be turned off, because a score reps cannot explain is a score reps do not use. The explainability panel also shows the score history as a line chart, so the rep can see not just the current number but the shape of the trajectory. A lead at 72 trending up is a very different conversation from a lead at 72 trending down.

Can one tenant run more than one scoring model?

Yes. Define a separate scoring config per segment, per product line, or per territory. A self-serve SMB lead and a strategic enterprise lead do not share the same rubric. Each config has its own rule set, its own thresholds, and its own routing flow. Reps see the score computed under the config that matches the lead's segment, so one buyer-profile's rules do not inflate another buyer-profile's scores. Admins can set the segment-to-config mapping as a Flows condition, which means edge cases (a lead that qualifies for two segments, a lead whose segment changes mid-cycle) are handled by the same conditional logic the rest of the automation stack uses. Scoring configs can be cloned, versioned, and A/B tested the same way flows can.

How do score thresholds trigger routing?

Set a score threshold in the scoring admin. When any lead crosses it, fire a flow using the same Flows engine every other Strkr automation uses. Common patterns include "score above 80 routes to a named AE in Slack within two minutes," "score above 60 drops into a nurture track with a sequence of three emails over 10 days," and "score below 20 flips to a disqualified status and removes the lead from active queues." Threshold routing is just a flow with a score-change trigger, which means it inherits atomic writes, dry-run testing, cascade guards, and the shared audit trail. Teams can set multiple thresholds on the same scoring config and have different flows fire at different levels, so one config can drive the full lifecycle from first-touch nurture through hot-handoff.

What is the right way to review a scoring model over time?

Review the scoring config quarterly. The Strkr admin surfaces three things that make the review fast: event volume per rule (so you can see if a signal nobody fires still has weight), rule-weight-to-conversion correlation (so you can see if the signal you weighted heaviest actually predicts closed-won), and the historical distribution of scores across all leads (so you can see whether the threshold needs to move). A good quarterly review takes under an hour. Prune the bottom-quartile rules by event volume. Rebalance the top-quartile rules by correlation to closed-won. Move the threshold if the distribution has shifted. That cadence is what keeps a scoring model useful into year three instead of silently rotting into noise.

Related features

Pair this with...

Lead Routing Workflow Automation Forecasting Reports Custom Objects

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