A Claude prompt that diagnoses where junk leads are entering pipeline through paths that bypass the deployed ICP scoring rubric — surfaces the top form sources, channel-tier combinations, partner referrals, and event registrations causing leakage, with a prioritized fix memo per leakage point. The bypass companion to Signal Quality Audit.
A B2B SaaS team ships an ICP scoring rubric. They run Signal Quality Audit monthly — the rubric is firing accurately, no calibration issues, distribution is healthy. But three months later, the sales team is still complaining about junk leads. The MQL-to-SQL rate stays at 14% instead of climbing toward the 25-35% benchmark that ICP scoring should produce. The team's instinct is to re-tune the rubric. The actual problem: 40-60% of leads are entering pipeline through funnel paths the rubric never gets applied to. Event registration forms route through a separate marketing automation system. Gated content forms have a webhook that bypasses HubSpot's lead-scoring logic. Partner referrals arrive pre-flagged as 'qualified' regardless of actual fit. The rubric is great. The bypass is killing it.
The deeper problem is that most ICP rubric implementations focus on the rubric itself, not on the funnel paths that should trigger it. A team can have a perfectly calibrated 100-point rubric with healthy weight distribution, accurate firmographic data, and validated thresholds — and still have major junk lead volume because the rubric never runs against 50% of inbound leads. This is a different operational failure mode than a poorly-tuned rubric, and it requires a different diagnostic. Signal Quality Audit answers "is the rubric working where it's deployed?" Junk-Lead Leakage Diagnosis answers "where is the rubric being bypassed?"
This workflow runs the bypass diagnostic. Claude pulls the last N days of leads from HubSpot, classifies each by source × score × form type, surfaces the top leakage points by volume + cost, and produces detailed fix memos per leakage point. Run quarterly, with trigger-based re-runs on operational changes (new content gating, new partner channel, event sponsorship changes). Form-level fixes ship in 1-2 weeks; deeper integration fixes take 4-12 weeks. Re-run after 30 days to verify leakage closure.
The gold variables — your brand, ICP rubric thresholds, lead source taxonomy, and lookback window — are the parts you edit. Run quarterly with the same lookback window so leakage trends are comparable across cycles.
hubspot via the free Growthspree MCP for source × form × score data. ICP Scoring Rubric Builder must run first; Signal Quality Audit is the within-rubric companion to this bypass diagnostic.Sample output for a hypothetical mid-market B2B SaaS brand with deployed ICP rubric, ~340 leads/month over 90-day lookback. Diagnosis surfaces 3 dominant leakage points accounting for 67% of total bypass volume.
Pulling 90-day lead dataset and classifying each lead by score-applied vs bypassed.
1,022 total leads in 90-day window. Classification: 412 SCORED+QUALIFIED, 198 SCORED+DISQUALIFIED, 412 BYPASSED. Bypass rate: 40% of total. Estimated leakage cost (90 days): $12,800. Top 3 leakage points: (1) Gated content forms via Zapier webhook — 178 leads, 71% junk rate, $5,700 cost. (2) Webinar registrations from separate marketing automation — 124 leads, 64% junk rate, $3,600 cost. (3) Partner referrals pre-flagged as qualified — 56 leads, 52% junk rate, $1,300 cost. These three account for 67% of total leakage cost. Fastest win: gated content form webhook reconfiguration — 1-2 week ship, closes 43% of leakage volume.
contact_created event which triggers lead-scoring workflow. Net change: ~3 hours Marketing Ops time, removes Zapier as middleware for this workflow.is_partner_qualified = true set by the partner program's submission form. This flag triggers a workflow that bypasses ICP scoring on the assumption that partners pre-qualify leads. In practice, partner quality control is highly variable — some partners send 90%+ qualified leads, others send mostly junk.Run after the ICP rubric has been live 60+ days with statistically meaningful lead volume. Trigger-based re-runs after major lead source changes (new content gating, new partner channel, event sponsorship changes).
The diagnosis requires the rubric to have been live with sufficient volume to produce statistically meaningful leakage signals. Below 60 days or below 200 leads/month, results are flagged as DIRECTIONAL — useful for direction but not for prioritization. For lower-volume accounts, extend the lookback window to 90+ days.
Run ICP Scoring Rubric Builder first →Edit the gold variables — your brand, ICP rubric thresholds, average ACV, cost per junk lead. Confirm your HubSpot source attribution categories match the standard taxonomy (organic, paid_google, paid_linkedin, partner, event, referral, direct, email). If your source taxonomy is non-standard, update the prompt's lead source taxonomy variable to match your HubSpot setup before running.
For an account with 200-1000 leads/month, the workflow takes 7-12 minutes. Claude pulls the lead dataset, classifies SCORED vs BYPASSED, groups by source × form, calculates per-leakage cost, and surfaces the top 3 leakage points with detailed fix memos. The output is the leakage table + 3 detailed fix memos — these are the action artifacts.
Hand fix memos to the right owners. Tier 1 fixes (form-level webhook reconfigurations) ship in 1-2 weeks via Marketing Ops. Tier 2 fixes (source attribution + partner SLA) take 4-8 weeks via RevOps + Channel Partner coordination. Tier 3 fixes (event marketing automation rebuilds) take 8-12+ weeks. Re-run the diagnosis at 30 days to verify Tier 1 closure and at 90 days to verify Tier 2 closure.
Same 5-leakage-category framework, different scope. Pick the one that matches your funnel architecture and operational priorities.
PLG companies have a fundamentally different leakage profile because free trial signups dominate inbound volume. The standard 5 categories collapse — almost everything routes through free trial. PLG variant adds activation-based leakage scoring (which trial signups bypass scoring AND don't activate) and self-serve vs sales-assisted segmentation.
Event-heavy funnels (conference exhibitors, webinar-heavy B2B brands) have event registrations as the dominant leakage point. Event variant runs deeper diagnosis on event-specific bypass — separates conference badge scans from webinar registrations from virtual event signups, and surfaces per-event source quality.
Partner-heavy funnels need per-partner accountability rather than aggregate partner channel diagnosis. Partner variant subdivides partner referrals by individual partner and surfaces per-partner conversion quality, allowing the channel team to either renegotiate SLAs or pause low-quality partners.
You can spend 6 months tuning the perfect ICP rubric. If 50% of leads bypass it through gated content forms, event registrations, and partner referrals, the tuning won't matter. Run the leakage diagnosis quarterly. Ship the Tier 1 form-level fixes in 1-2 weeks. Coordinate the longer-tail integration fixes across Marketing Ops + RevOps + Channel Partners. Or have senior GrowthSpree operators run the diagnosis quarterly, generate the fix memos, and coordinate cross-functional execution — the same operating motion run across 300+ B2B SaaS accounts.