Sixty thousand leads entered the SDR queue. Roughly 2,000 survived ICP qualification. The other 58,000 consumed enrichment credits, routing logic, and rep hours before anyone realized they were dead on arrival.
That 97% discard rate sounds brutal. It should. And the math behind it is what matters for anyone managing pipeline economics in B2B SaaS right now.
The Capacity Tax Nobody Budgets For
LeadSpot's 2026 B2B Lead Quality Report found that SDRs spend 9.4 hours per week on leads that prove unqualified, dead, or unreachable. Integrate's research from the same year pegs the number at 27% of total rep time going to poor-fit leads. Together, they describe a structural problem: the SDR queue is doing qualification work that should happen before a human touches the record.
Here's the cost translation. A $65 unverified lead converting to SQL at 10% runs $650 per SQL. A $90 human-verified lead converting at 25% drops you to $360 per SQL (LeadSpot Pipeline Trust Report, 2026). The cheaper lead costs more. This is the unit economics trap that CPL dashboards hide from leadership.
Fit and Intent Are Two Separate Questions
The instinct is to score everything on a single axis. Composite scores feel clean, fit neatly into a CRM field, and also hide the reason a lead was prioritized. Sales can't trust them. Marketing can't debug them.
A more defensible approach, echoed across Rework, Apollo, and Tomba's scoring frameworks, separates two dimensions: (1) ICP fit (right account, right role, right firmographics) and (2) intent/timing (is there a reason to act now). High intent from a poor-fit account is still a poor-fit account. High fit with no timing signal goes to nurture, not to the top of the dial list.
- High Fit + High Intent: Route to SDR immediately.
- High Fit + Low Intent: Nurture sequence, re-evaluate on signal change.
- Low Fit + High Intent: Suppress or deprioritize. This is the quadrant teams most often get wrong.
- Low Fit + Low Intent: Archive.
Keeping these dimensions in separate fields (not collapsed into one number) gives RevOps a debugging surface when conversion rates shift.
Negative Signals Do More Work Than Positive Ones
Most scoring models obsess over what makes a lead good. The faster lever is defining what makes a lead bad. Competitors, existing customers, opt-outs, invalid domains, roles outside the buying committee, companies below your minimum ACV threshold: these are explicit disqualifiers that can be automated before enrichment even runs.
Apollo and Involve.me's scoring guidance both emphasize this: suppression is a feature, not a failure. A negative-fit rule that removes 40% of records from the queue before any SDR sees them isn't losing pipeline. It's protecting capacity for records that can actually convert.
In the 60,000-to-2,000 reduction, negative-fit rules did the majority of the cutting. Industry mismatch, employee band below threshold, missing buying role, known competitor domains. Each rule is auditable. Each rule has an owner.
The Trust Gap Between Marketing and Sales Numbers
LeadSpot's 2026 Pipeline Trust Report surfaced a stat worth pausing on: marketing teams reported a 31% MQL-to-SQL conversion rate while sales teams reported 8% on the same pipeline. That's a stage-definition disagreement that erodes credibility with the CFO and the board.
The fix isn't picking one number. It's reconciling what counts as an MQL, what counts as an SQL, and who marks the transition. Teams that qualified leads before delivery to SDRs reported 28% MQL-to-SQL versus 9% for teams qualifying after delivery (LeadSpot, directional comparison, not causal proof). The upstream gate doesn't just improve conversion. It reduces the surface area for reporting disputes.
What This Actually Costs to Get Wrong
The MQL-to-SQL benchmark across B2B SaaS sits around 13% (Martal, Tomba, 2026; definitions vary). SQL-to-opportunity conversion runs 47-60% depending on the source. For every 100 MQLs, roughly 6-8 become opportunities. If 22% of leads entering the queue are bad or unusable (LeadSpot's marketing ops survey), the denominator is inflated before the funnel even starts.
Fully loaded cost per opportunity, including SDR labor, tooling, and overhead alongside media spend, is the metric that makes the case for an upstream gate. CPL alone will never justify cutting volume, because CPL makes 60,000 leads look like abundance. Cost per opportunity makes them look like what they are: mostly waste with a small, valuable core buried inside.
The 2,000 that survived weren't the cheapest leads. They were the ones where fit was verifiable, intent was present, and no disqualifier applied. The other 58,000 would have consumed the same SDR hours, the same sequences, the same domain reputation, and returned almost nothing. The queue was never the bottleneck. The absence of a gate was.
