Built for RevOps teams designing a defensible scoring model
A lead scoring model that maps revenue to points, intent to weighting, and routing to reality.
Lead scoring breaks on three decisions: which revenue ranges map to which point values, how to blend firmographic fit with behavioral signal, and how to turn the final score into routing. Strkr ships a scoring engine that handles all three with transparent math, visible weighting, and routing rules that respect ownership.
Lead scoring is one of those capabilities that every CRM claims to ship and few actually ship well. The common failure modes are specific and repeated: the revenue-range mapping is a hidden formula nobody can read, the intent weighting is a black box that the vendor says is "AI-driven" without showing the math, and the routing fires on a score that only loosely correlates with real buyer intent. The sales team stops trusting the score within a quarter, the marketing team keeps feeding leads through, and the RevOps lead ends up defending a model nobody can audit. The six failure patterns below are the ones Strkr's scoring engine was designed to fix, with a bias toward transparent math and visible overrides rather than clever opacity.
The revenue-range mapping is opaque
Nobody knows why "500K to 2M" is 15 points.
Most scoring tools let the admin set point values on revenue ranges but hide the resulting curve. The team ships a model where "500K to 2M revenue" scores 15 points and "2M to 10M" scores 25 points, with the ratios reverse-engineered three months later when sales asks why. Strkr's scoring engine shows the full curve as a visible chart with the point values on each range, so the admin and the sales team see the mapping in one glance.
Firmographic data decay makes revenue a moving target
The revenue field was accurate 11 months ago.
Revenue data on a prospect account decays like every other firmographic field. The score that was built on 2024 revenue data is scoring 2026 pipeline against stale numbers. Strkr tracks a per-field last-verified timestamp on revenue and other firmographic fields, with the score engine optionally weighting fresher data higher. The pattern avoids the common failure of a lead scoring higher than it should because its revenue field hit the right range a year ago.
Intent weighting is a black box
The vendor says "AI-driven" and will not show the math.
The common pattern on HubSpot predictive scoring, Salesforce Einstein, and MadKudu: the scoring model is "AI-driven" with limited visibility into which signals weight how heavily. The RevOps lead cannot defend the model under scrutiny because the math is opaque. Strkr's scoring engine makes the weighting visible: each signal has a point value, each range has a mapping, each override has a reason, and the final score is explainable.
Negative scoring is handled poorly
The disqualifier does not actually disqualify.
Lead scoring models often ship positive scores for fit and engagement without a clean way to express disqualifiers (competitor employees, student emails, revenue below floor, geography outside served markets). The result is that leads score high on fit and engagement despite being structurally unqualifiable, which pollutes the sales queue. Strkr supports negative point values and hard-stop disqualifiers that cap the score regardless of other inputs.
Routing fires on a score the rep does not trust
The score is 85 and the rep sees it as a 20.
When the sales team does not trust the score, they ignore the routing and work on hunches. The ops team keeps the score alive on dashboards, the sales team works around it, and the whole model becomes a reporting artifact rather than a routing driver. Strkr's scoring surface shows the component scores (fit, intent, fit decay, override) alongside the composite, so the rep sees why a lead landed at a particular score and can trust the output.
Model updates break historical reporting
A tweak to the point values rewrites January.
When the RevOps lead updates the point values on a scoring model, most tools apply the new model retroactively to historical data. The quarter-over-quarter trend becomes unreliable because the baseline moved. Strkr versions the scoring model so each lead carries the score computed at the version in effect at lead creation, with a version history visible in the audit log. Historical reporting stays reliable across model updates.
What the scoring engine ships
Transparent scoring with visible math and auditable updates.
The scoring engine is designed for the RevOps lead who has to defend the model under scrutiny: to a sales team that wants to trust the routing, to a marketing team that wants attribution, to a CFO that asks why the pipeline conversion dropped last quarter. The capability cards below are what the admin configures and what the sales team sees.
Revenue-range to point mapping
Visible curve with range breakpoints and point values.
The admin configures the revenue-range to point mapping as an interactive curve: breakpoints (under 500K, 500K to 2M, 2M to 10M, 10M to 50M, 50M to 250M, over 250M) with configurable point values per range. The chart renders in the admin UI so the RevOps lead sees the shape of the curve, with per-range tooltip detail. The pattern eliminates the "why is this range 15 points" question because the math is visible.
Firmographic signal weighting
Industry, employee count, geography, tech stack as inputs.
Beyond revenue, the firmographic layer supports point values for industry fit (per-industry mapping), employee count ranges, geography (country, state, region), tech stack (specific products that signal fit), and funding stage. Each signal has a visible point value and a visible contribution to the composite score. The admin can disable or re-weight signals without a model rebuild.
Behavioral signals (pricing page visit, demo request, email open, email click, in-product trial activity, meeting attendance) have per-signal point values with time decay. A pricing page visit in the last 7 days scores higher than one 90 days ago. The composite captures the first-party behavioral layer with the same visibility as the firmographic layer.
Third-party intent blend
Bombora, G2, and licensed intent as weighted inputs.
Third-party intent sources (Bombora buyer intent, G2 buyer intent, G2 Scoop where available) feed into the composite score with configurable weighting. The RevOps lead sets the weight based on how predictive third-party intent has been for the specific motion. The pattern blends first-party signal (which is specific but limited) with third-party signal (which is broad but noisy) into a composite that beats either alone.
Negative scoring and hard-stop disqualifiers
The competitor employee lead stays at zero.
Negative point values express disqualifiers (competitor employee, student email, revenue below floor, geography outside served markets). Hard-stop rules cap the score regardless of positive inputs: a competitor employee lead cannot score above 10 even if they engaged with every piece of content. The pattern keeps the sales queue clean of structurally unqualifiable leads.
Decay and recency weighting
Older signals contribute less.
Behavioral signals decay over time: a pricing page visit yesterday scores higher than one 60 days ago. Firmographic signals decay more slowly but still carry a recency factor on the last-verified timestamp. The decay curves are configurable per signal class. The pattern avoids the common failure of a lead scoring high today because of behavior from a year ago that is no longer predictive.
Score bands and routing
Hot / warm / cold with per-band routing rules.
The composite score maps to score bands (hot above 80, warm 50 to 80, cool 25 to 50, cold under 25) with per-band routing rules. Hot leads route to a named rep or team; warm leads route to a nurture track; cold leads route to a long-cycle re-engagement sequence. The band breakpoints are configurable and the routing writes to the lead ownership field via the standard assignment rules.
Version history and model updates
A model change does not rewrite history.
Model updates version the scoring configuration so each lead carries the score computed at the version in effect at lead creation. Historical reporting stays reliable because the past does not get rewritten when the current model changes. The version history is visible in the admin audit log with the reason for each update attached.
Score explainability on the record
The rep sees why a lead scored 85.
The lead record shows the composite score with a breakdown by component (fit from revenue, fit from industry, intent from behavior, intent from third-party, overrides). The rep clicks the score and sees the full math. The pattern replaces the "the AI says it is a 72" surface that most tools ship with a transparent breakdown the rep can trust and defend.
Where Strkr beats a HubSpot or MadKudu scoring model
Six places opaque scoring actually breaks.
HubSpot predictive scoring, Salesforce Einstein, MadKudu, and Breadcrumbs ship capable scoring engines with varying degrees of model transparency. The specific failures below are where opacity undermines the model, and where Strkr's approach makes the different choice.
Visible math
The scoring is explainable, not black-box.
The composite breaks down into component scores (fit, intent, decay, override) that the admin and the sales team can read. The RevOps lead defends the model under scrutiny because the math is visible, which is a materially different posture than "the AI says it is a 72."
Visible curve, not hidden thresholds
Revenue-range mapping renders as a chart.
The revenue-to-points mapping shows as an interactive curve rather than a buried form field. The admin sees the shape, adjusts breakpoints, and understands what the model does. The sales team can read it too, which is the first condition for trust.
Negative scoring and disqualifiers
Unqualifiable leads stay out of the queue.
Hard-stop rules cap scores on structurally unqualifiable leads. Most scoring tools treat negative inputs as weak-positive signals or require awkward workarounds in rule builders. Strkr handles disqualifiers as first-class rules.
Version-safe history
Model updates do not rewrite the past.
Each lead carries the score computed at the model version in effect at lead creation. Quarter-over-quarter trend reporting stays reliable through model updates, which eliminates a specific class of reporting failure that undermines scoring programs over 12+ months.
First-party and third-party blend
Behavioral plus Bombora in one composite.
The scoring engine blends first-party behavioral signal (pricing page, email, in-product) with third-party intent (Bombora, G2) and firmographic fit into one composite. Most scoring tools handle one or two of these cleanly, not all three.
Routing writes directly to ownership
Score band maps to assigned rep.
The score band maps to a routing rule that writes the lead ownership field, which is the only place the score matters operationally. Most scoring tools stop at the score and leave routing as a separate project.
What this changes for RevOps and sales
The operational outcomes of a defensible scoring model.
A scoring model that works operationally changes three specific conversations: the daily routing conversation between marketing and sales, the quarterly conversion-rate conversation between RevOps and finance, and the annual planning conversation around funnel efficiency. The outcomes below are what each conversation looks like once the model is live.
Marketing-to-sales handoff tightens
The sales team stops rejecting MQLs.
When sales trusts the score, the handoff from marketing to sales tightens materially: MQL rejection rates drop from a typical 25 to 40 percent to 8 to 15 percent, and the time from MQL to first touch drops because routing fires reliably. The pattern eliminates a specific class of marketing-sales conflict that defines mid-market revenue orgs.
Conversion rates become comparable
Quarter-over-quarter trend holds up.
Because the model versions and historical scores stay fixed, quarter-over-quarter conversion rates (lead to opportunity, opportunity to closed-won) stay comparable through model updates. The CFO can hold the conversion-rate conversation on real data rather than noise.
Disqualifiers clean the queue
Reps stop working leads that cannot close.
Hard-stop disqualifiers (competitor employee, student, revenue below floor) keep structurally unqualifiable leads out of the sales queue. The common pattern of a rep burning two hours on a lead that was never qualifiable drops materially, which recovers rep time for leads that can actually close.
The RevOps lead can defend the model
The scrutiny conversation has an answer.
When finance, sales, or the CRO challenges the model, the RevOps lead walks through the component scores, the revenue curve, the intent weighting, and the override log. The conversation ends with alignment rather than a defensive back-and-forth. The pattern replaces the "the AI said so" answer that undermines scoring programs over time.
Routing stays current with coverage
Score bands update as the team grows.
As the sales team grows, score band breakpoints and routing rules update to reflect new coverage. The band structure that worked at 10 reps evolves at 30 reps and 100 reps without a model rebuild. The pattern lets the scoring program scale with the business rather than require a reset every 18 months.
Attribution gets honest
Marketing-sourced revenue traces back to the scoring.
When the scoring model is transparent and version-safe, marketing attribution becomes defensible: this campaign generated leads that scored above threshold, which routed to sales, which converted at this rate. The attribution number is auditable rather than claimed. The pattern tightens the budgeting conversation between marketing and finance.
Head-to-head
Strkr scoring vs HubSpot predictive, Salesforce Einstein, MadKudu.
The scoring tool market splits between "predictive AI" tools that promise accuracy with limited visibility and rule-based engines that offer transparency with limited sophistication. Here is the head-to-head on the capabilities that drive a defensible scoring program.
Feature
Strkr
HubSpot Predictive / Einstein / MadKudu
Model transparency
Component scores visible on every lead
Predictive score, limited breakdown
Revenue-range mapping
Visible curve with configurable breakpoints
Internal model input, limited admin control
First-party behavioral blend
Native with configurable decay
Varies by platform
Third-party intent blend
Bombora, G2 integration with weight controls
Via connected tools, limited blending
Negative scoring and hard stops
First-class with reason attribution
Varies, often awkward
Model version history
Each lead carries version-at-creation score
Current model applies retroactively
Score explainability on record
Clickable breakdown with math
Single composite number typical
Routing writes to ownership
Score band to assignment rule
Score computed, routing often separate
Admin profile required
Ops lead, no certification
Varies: analyst or admin per tool
Pricing
Included on CRM per-seat line
Varies; HubSpot predictive on higher-tier plans
How teams use Strkr
How RevOps teams run scoring on Strkr today.
The playbooks below are from mid-market and growth-stage revenue teams that moved scoring off a HubSpot predictive model or a MadKudu install onto Strkr. Each is live today.
B2B SaaS, 42-rep SMB motion
Revenue-to-points curve replaced a HubSpot predictive score.
A B2B SaaS running an SMB-and-mid-market motion ran HubSpot predictive scoring. The sales team did not trust the score and routinely overrode routing. The RevOps lead rebuilt the model on Strkr with a visible revenue curve (under 500K at zero, 500K to 2M at 15 points, 2M to 10M at 25, 10M to 50M at 20 because the ICP was below enterprise, 50M+ at 10 as outside ICP). Sales adoption of the score climbed to above 85 percent trust-and-work-the-routing within one quarter.
Mid-market industrial, 68 reps
Negative scoring eliminated competitor-employee leads from the queue.
A 68-rep industrial team had been working an average of 11 competitor-employee leads per week because the scoring tool had no clean disqualifier. The team moved to Strkr with a hard-stop rule on competitor email domains, which cut the leaked leads from 11 per week to zero. The rep hours recovered per week landed at roughly 42 hours across the team, which the ops lead attributed to the single disqualifier.
Series B fintech, 24 SDRs
Third-party intent blended with first-party behavior.
A Series B fintech ran Bombora intent data on a separate contract and HubSpot predictive scoring. The two signals never joined; the SDRs worked whichever list was in front of them. The team moved scoring to Strkr with Bombora intent as a weighted input alongside first-party behavior (pricing page, demo request, email engagement). Meeting-set rate climbed 31 percent over the following quarter, with the ops lead attributing roughly 60 percent of the lift to blended scoring and 40 percent to tighter routing.
Enterprise ABM team, 110 named accounts per rep
Score bands drove the Monday prospecting queue.
A 15-rep ABM team ran 110 named accounts per rep. The Monday prospecting queue was previously built on a weekly export plus a hand-sorted priority list. The team configured Strkr score bands (hot above 80, warm 50 to 80) and let the queue order by composite score. Rep-reported prospecting-queue satisfaction (a weekly survey) climbed from a 3.1 average to a 4.4, and meeting-set rate climbed 23 percent over the first quarter.
Growth-stage PLG, 18-person sales team
In-product activity signals folded into the scoring model.
A growth-stage product-led SaaS ran trial activity signals (number of users invited, workflows created, API calls made) that the sales team used informally to prioritize trial outreach. The team folded the signals into the Strkr scoring model as behavioral inputs with per-signal point values. The composite score produced a prioritized trial-outreach queue that landed on the sales team's dashboard. Trial-to-paid conversion on the top 20 percent of scored trials climbed from 11 percent to 19 percent.
Lead scoring that maps revenue to points, blends intent, and explains the math.
Start a 14-day trial with the scoring engine, revenue-to-points curve, intent blending, and routing rules enabled on a sample segment. Build a defensible model against your actual pipeline and see the component math on every lead.
How do we decide the point values on the revenue-range mapping?
The practical starting point is to look at the historical conversion rate (lead to opportunity) by revenue range on your actual pipeline for the last 4 to 8 quarters, and assign point values proportional to conversion rate. If leads in the 2M to 10M revenue range convert at twice the rate of leads under 500K, then that range should score roughly twice the points. Strkr's scoring admin UI shows historical conversion rates per range alongside the point configuration, so the RevOps lead calibrates the values against real data rather than guessing. The first-pass model usually needs one or two tune cycles over the first quarter.
How does the model handle leads with missing revenue data?
When the revenue field on a lead is missing or stale (over the configurable recency threshold), the scoring engine applies a configurable fallback: either a null-fallback (lead scores on non-revenue inputs only, with a flag), an industry-median fallback (lead scores against the median revenue for the industry, with a confidence discount), or an infer-and-flag path where Strkr AI runs an inference step over public sources with a confidence score attached. The admin picks the policy per segment. The pattern avoids the common failure of a lead scoring low because the revenue field was never captured, which produces false negatives on real-but-undocumented ICP fit.
How do we blend firmographic fit with behavioral intent in the composite score?
The composite is a weighted sum: fit score (from firmographic signals including revenue, industry, employee count, geography) plus intent score (from behavioral signals including pricing page visits, email engagement, in-product activity) plus third-party intent (from Bombora, G2, or other licensed sources) plus or minus overrides. The weights are configurable per segment. The common starting point is to weight fit at roughly 40 percent, first-party intent at 35 percent, third-party intent at 15 percent, and reserve 10 percent for overrides. The weights tune over the first 2 to 4 quarters based on which component best predicts closed-won.
Can the scoring model be audited by finance or an external reviewer?
Yes. The scoring configuration (ranges, points, weights, decay curves, hard stops, overrides) is a readable JSON document exportable from the admin UI. The version history shows every change with who made it, when, and the reason. Historical lead scores retain the version in effect at lead creation, so an auditor can reconstruct the scoring decision for any lead at any point in time. For regulated industries (financial services, healthcare, insurance), the audit trail is a baseline rather than a premium feature.
How does the model handle score decay over time for a lead that engages and then goes quiet?
Behavioral signals decay on configurable curves: a pricing page visit scores full points in the first 7 days, half points at 30 days, and tapers to zero by 90 days (the defaults are tenant-configurable). A lead that engages hot and then goes quiet watches its behavioral score decay back toward the fit-only baseline over the decay window. The pattern avoids the common failure of a lead scoring high today because of behavior from months ago that is no longer predictive, which was a specific complaint on first-generation scoring models that treated signals as permanent additions.
What is the migration path from HubSpot predictive scoring or MadKudu to Strkr?
For teams on HubSpot predictive, the migration consists of: export the current scoring logic (where visible), rebuild the equivalent model on Strkr with the component structure (fit, first-party intent, third-party intent, overrides), run both models in parallel for 2 to 4 weeks with Strkr computing scores alongside HubSpot, compare the resulting rankings against actual conversion, tune the Strkr model until it matches or beats HubSpot on predictive accuracy, cut over. Typical migrations complete in 4 to 8 weeks including the parallel-run calibration period. For MadKudu or Breadcrumbs, the migration follows the same pattern.
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.
We use cookies.
Essential cookies keep the site working. If you accept, we also enable Google Analytics
so we can see which pages help and which don't. Reject to opt out entirely. Details in our
Privacy Policy.
