B2B lead scoring mistakes derail pipeline efficiency and waste sales cycles on low-intent prospects. This guide dissects the structural errors—over-reliance on demographics, ignoring decay, conflating implicit and explicit signals—that plague scoring models in Canadian and global B2B contexts.
Many B2B lead scoring models assign heavy points to company size, industry vertical, and job title—essentially betting that a VP at a 500-person SaaS firm is inherently more qualified than a director at a 50-person agency. This demographic bias ignores intent entirely. A VP who visited your pricing page once six months ago is not more sales-ready than a manager who attended two webinars, downloaded a comparison guide, and visited case studies this week. The error compounds in Canada's mixed-size market: a mid-market Montreal manufacturer with urgent buying signals often converts faster than an enterprise contact in Toronto with no active project. Firmographics set your addressable market and inform segmentation, but they should anchor the floor score, not dominate it. Reserve the majority of points for behavioral and engagement signals that indicate active research or problem urgency. If your model gives more weight to LinkedIn title than to recent high-intent content consumption, you are scoring fit over readiness—and sales will chase logos instead of opportunities.
Static scoring treats all actions as永久 permanent: a lead who downloaded an ebook nine months ago carries the same points today as one who did it yesterday. This produces a backlog of stale high-scorers—contacts who showed interest in a different quarter, under different business conditions, possibly in a different role. Without decay, your MQL threshold becomes a historical artifact rather than a real-time signal. Implement exponential or linear decay: reduce points for actions older than 30, 60, or 90 days depending on your sales cycle length. A three-month decay window works for many B2B cycles; shorter cycles (30-45 days) suit transactional SaaS, longer (120+ days) fit complex enterprise deals. Pair decay with recency bonuses—award extra points for clusters of activity in the past seven or fourteen days. This surfaces leads in active buying mode and prevents your pipeline from filling with names who were curious last winter but have since gone dark. Decay also exposes leads worth re-engaging: if someone scored high, decayed out, then spiked again, that re-engagement is a strong buying signal.
Not all clicks mean the same thing. Viewing a blog post about industry trends is exploratory; requesting a demo or pricing sheet is transactional. Scoring them equally—or worse, giving blog reads the same weight as bottom-funnel actions—floods your MQL list with researchers, students, and competitors. Separate implicit signals (page views, email opens, social follows) from explicit signals (form submissions, meeting bookings, direct replies, trial signups). Implicit actions should add small incremental points and build a picture of engagement over time; explicit actions should trigger larger jumps or even instant MQL status if the context is right. For example, visiting your careers page or a generic About page tells you almost nothing about buying intent. Visiting ROI calculator pages, integration documentation, or case studies in your target vertical tells you much more. Create signal tiers: high-intent pages and actions worth ten to twenty points, mid-intent worth three to five, low-intent worth one. If you are not segmenting by intent type, you are scoring activity, not opportunity.
Lead scoring is not purely additive. Negative scoring subtracts points or flags disqualification when a lead exhibits behaviors or attributes that indicate low fit or lost interest. Common negative triggers include unsubscribes, email hard bounces, spam complaints, job changes out of your ICP (a director moving to a non-target industry), and repeated visits to competitor comparison pages followed by no engagement. Many systems never subtract—once a lead hits MQL, it stays there even after the contact goes cold or moves companies. This creates phantom pipeline and wastes rep time on dead-end follow-ups. Assign negative point values: minus ten for unsubscribe, minus fifteen for bounce, minus five for extended inactivity (no engagement in 120+ days if typical cycle is 60). Set automatic disqualification rules for role changes detected via enrichment tools or for domains that match known competitors, agencies doing research, or free email providers in contexts where you sell only to businesses. Negative scoring also trains the model faster—leads that churned or were marked lost should lose points retroactively in your scoring logic review, so future similar profiles start lower.
Marketing builds the scoring model in isolation, sets thresholds based on volume targets, and hands MQLs to sales. Sales complains the leads are unqualified, stops following up promptly, and the model never improves. This is the most pervasive B2B lead scoring pitfall in Canada and elsewhere—no closed-loop validation. The fix requires process, not just technology. Run monthly or quarterly win/loss scoring audits: pull all closed-won deals and compare their score progression and actions to closed-lost and stalled opps. Which behaviors and score ranges correlate with wins? Which high-scorers never converted, and why—bad timing, wrong persona, competitor incumbent? Use CRM disposition codes (not interested, bad fit, timing, budget, competitor) to tag losses, then map those back to scoring attributes. If leads scoring 80+ on competitor comparison content rarely close, lower those point values. If leads who attend live demos convert at twice the rate of whitepaper downloaders, reflect that gap. Surface rep feedback formally—ask sales which MQLs felt premature and which should have scored higher earlier. Adjust thresholds and weights based on evidence, not opinion. Scoring is a model, and models require training data. Your CRM outcomes are that data.
A SaaS startup selling to both small agencies and enterprise clients often uses one MQL threshold—say, 100 points—for everyone. But enterprise deals involve longer cycles, more stakeholders, and committee-based research spread over months. A score of 100 might represent early-stage exploration for enterprise and ready-to-buy for SMB. Forcing both through the same gate either under-qualifies enterprise (they never hit the threshold before going dark) or over-qualifies SMB (reps get swamped with tire-kickers). Segment your scoring by persona, deal size, vertical, or region if behaviors differ meaningfully. Create separate MQL thresholds or even distinct scoring models. For example, a Vancouver tech company might score agency leads on speed-to-engage and content downloads, but score enterprise on meeting acceptances, multiple-stakeholder engagement (different contacts from the same domain), and procurement-content access. In bilingual Canadian contexts, engagement in French-language content from a Quebec lead might signal higher intent than English-only engagement and merit bonus points. If your product has both self-serve and sales-assisted motions, score them separately—self-serve might auto-convert at a lower threshold with no human handoff. Single-threshold models optimize for neither segment and waste both marketing spend and sales capacity.
Lead scoring assigns points based on fields—job title, company size, industry—but those fields are often incomplete, outdated, or wrong. A lead self-reports as 'Manager' with no company name; your scoring model has no firmographic data to evaluate, so it defaults to median points and the lead drifts into a black hole. Or a contact changed employers eight months ago but your CRM still associates them with the old domain, so you are scoring them against the wrong account. Data decay is constant—contacts change roles, companies get acquired, email addresses bounce. If you are not running regular enrichment (Clearbit, ZoomInfo, Cognism, Apollo) and deduplication, your scoring logic operates on fiction. Set up hygiene workflows: enrich new leads on intake, re-enrich existing database quarterly, merge duplicates weekly, and suppress role-based addresses (info@, sales@) that will never convert. For Canadian contacts, validate province and language preference if regional targeting matters—scoring a French-preference lead the same as English when you have no French content is a waste. Also audit form fields: if your form asks for 'Company Size' as free text instead of a dropdown, you will get 'small', '10-50', 'startup', and '20 employees' as four different values your scoring can't parse. Garbage in, garbage out. Scoring only works when the data it consumes is clean and current.
Over-reliance on demographics (job title, company size, industry) while underweighting behavioral intent. Many models give a VP at a large firm a high score by default, even if that person has shown zero recent engagement. True qualification comes from clustering recent, high-intent actions—demo requests, pricing page visits, case study downloads—not from static profile attributes. Prioritize engagement recency and explicit signals over firmographic fit alone.
Time decay reduces or removes points from actions that occurred weeks or months ago, preventing stale engagement from inflating scores indefinitely. A lead who downloaded a whitepaper nine months ago and did nothing since is not sales-ready today. Decay keeps scoring aligned with current buying intent and surfaces leads in active research mode. Typical decay windows range from thirty to ninety days depending on sales cycle length; pair decay with recency bonuses to amplify recent high-intent behavior.
No. Enterprise and SMB buyers exhibit different behaviors and timelines—enterprise involves longer cycles and multi-stakeholder engagement, while SMB often moves faster with fewer touchpoints. Using one MQL threshold either under-qualifies enterprise leads (they never reach it) or over-qualifies SMB leads (too many low-intent contacts). Segment scoring by deal size, persona, or vertical, and set thresholds that reflect each segment's typical journey and conversion patterns.
Negative scoring subtracts points or disqualifies leads when they exhibit low-fit or disengagement signals—unsubscribes, bounces, job changes out of ICP, or prolonged inactivity. Without it, leads accumulate points forever and clog the pipeline even after they go cold or leave the company. Negative scoring keeps the model realistic, flags contacts to suppress or re-segment, and prevents sales from chasing dead ends. Common negative triggers include spam complaints, competitor domain visits with no follow-up, and extended silence beyond your typical cycle length.
Quarterly at minimum, monthly if you have high lead volume or short sales cycles. Pull closed-won and closed-lost data from your CRM, map back to lead scores and behaviors, and identify which signals correlated with wins versus losses. Adjust point values and thresholds based on actual outcomes, not assumptions. Also gather qualitative feedback from sales on MQL quality—which leads felt premature, which should have scored higher sooner. Scoring is a model that improves with training data; your CRM disposition codes and win/loss records are that data.
Scoring logic relies on fields like job title, company size, and industry to assign points. If those fields are blank, wrong, or outdated—contact changed jobs, company was acquired, email bounced—the model cannot evaluate the lead accurately and defaults to generic scores. This produces both false positives (stale high-scorers) and false negatives (qualified leads stuck at low scores due to missing data). Regular enrichment, deduplication, and hygiene workflows are essential to keep scoring inputs reliable and actionable.