Why Your ICP Scoring Model Works for Enterprise but Breaks in Mid-Market (and How to Fix It)

The Problem Nobody Wants to Name

Your scoring model was built when the company sold to one segment. Then leadership added a second motion — usually mid-market added on top of enterprise, sometimes the reverse — and the same scoring logic was quietly extended to cover both. Nobody rebuilt it. The weights, thresholds, and tier definitions stayed the same.

Now the SDR team is complaining that MQLs don't convert. The enterprise AEs love their inbound. The mid-market AEs are ignoring the queue and sourcing their own pipeline. Marketing is defending the model with dashboards. Sales leadership is asking for "better leads."

This is ICP misalignment, and it's structural. It won't be fixed by tuning weights or buying better data. The core issue is that enterprise and mid-market buying behaviors are different enough that a single scoring model cannot serve both without one of them silently failing.

Why One Model Can't Serve Both Motions

Enterprise and mid-market look similar on a firmographic surface — both are companies with employees, revenue, and tech stacks. But the buying behavior, signals that matter, and time-to-signal all diverge sharply.

Enterprise: Broad Signals, Long Cycles, Committee Buying

Enterprise deals typically involve 8–12 stakeholders, 9–18 month cycles, and formal procurement. The scoring model that works here rewards:

• Account-level engagement across multiple personas (not just one champion)
• Presence of triggers like leadership changes, funding rounds, or org restructuring
• Existing tech stack signals suggesting displacement or integration opportunity
• Intent data spread across multiple research categories over weeks or months

In enterprise, a single high-intent action from one contact means very little. What matters is pattern density across the account. Scoring needs to be forgiving on any single signal but strict on the composite.

Mid-Market: Narrow Signals, Compressed Cycles, Concentrated Buying

Mid-market deals typically run 30–90 days with 2–4 stakeholders. Buying is often driven by one operator with authority and budget. The scoring model that works here rewards:

• Fit precision on a tight firmographic band (headcount, revenue, industry vertical)
• High-intent behavioral signals from a single contact (pricing page, demo request, competitor comparison)
• Recency, not accumulation — a signal from 45 days ago is stale
• Tech stack matches that indicate readiness to buy now

Mid-market scoring needs to be strict on fit but responsive to individual intent. The opposite of enterprise.

If you use the same model for both, one of these happens:

1. Enterprise-tuned model on mid-market leads: You wait for account-level pattern density that will never come because mid-market buyers don't behave that way. Good leads sit unscored. SDRs skip them.
2. Mid-market-tuned model on enterprise leads: A single contact from a Fortune 500 fills out a form and immediately gets routed as an SQL. The AE calls, discovers no buying committee, and marks it disqualified. Real enterprise opportunities get buried.

Diagnosing Which Failure Mode You Have

Before rebuilding anything, run a diagnostic on your closed-won data from the last four quarters. This is the same discipline we apply in a GTM Audit — treat your scoring weights as hypotheses, not truths.

The Tier Conversion Test

For each scoring tier (A/B/C or 1-100 bands), calculate:

• Win rate by tier
• Average deal size by tier
• Sales cycle length by tier
• Split all three by segment (enterprise vs mid-market)

Your Tier A leads should convert at roughly 2x your baseline. If enterprise Tier A converts at 3x baseline but mid-market Tier A converts at 1.1x baseline, your model is calibrated for enterprise and failing for mid-market. And vice versa.

The Closed-Won Escape Test

Pull every closed-won deal from the last year. What percentage came in as Tier A or Tier B at the time of MQL? If more than 20% of closed-won deals were scored below your qualification threshold, the model is missing real opportunities. Look at the segment split — usually one segment dominates the "escapes."

The Rep Behavior Test

Ask AEs and SDRs: "Do you trust the score?" Then look at their behavior. If reps in one segment consistently work leads out of score order — hunting through the queue for accounts they think are better — the model is failing that segment. Reps in the segment where scoring works will generally work top-down.

The Fix: Segmented Scoring, Shared Definition

You don't need two entirely different systems. You need two calibrated scoring models sitting under one shared ICP definition, with the routing logic determined at the point of lead capture.

Step 1: Sign a Single ICP Document

Sales, marketing, and CS leadership all sign one document that defines:

• Segment boundaries (what is enterprise vs mid-market — usually by employee count, revenue, or contract value)
• Positive ICP criteria for each segment
• Negative ICP (accounts to explicitly not pursue — wrong size, regulatory mismatch, chronic churn risk)
• Ownership of updates and review cadence

The most common failure pattern in B2B is sales and marketing running different ICPs. The signed document is the starting point, not the finish line — but without it, nothing downstream works.

Step 2: Build Two Scoring Models Under One Routing Layer

In HubSpot, Salesforce, or wherever your scoring lives, the architecture should be:

1. Segmentation logic runs first: Based on firmographic data (enriched via Clay, Apollo, or ZoomInfo at capture), the record is tagged Enterprise or Mid-Market.
2. Scoring model runs second: Two separate scoring calculations, each with its own weights, thresholds, and tier definitions.
3. Routing runs third: Segment-appropriate SLA, owner, and sequence.

If you're building this from scratch or restructuring an existing CRM, this is where HubSpot Architecture work pays off — the segmentation-first logic has to be baked into the record structure, not bolted on.

Step 3: Calibrate Weights Per Segment

Here's a rough template. Adjust based on your closed-won data.

Enterprise scoring weights:

• Firmographic fit: 25%
• Account-level engagement (multiple contacts): 25%
• Trigger events (funding, leadership, restructuring): 20%
• Intent data breadth (multi-category, sustained): 20%
• Tech stack signals: 10%

Mid-market scoring weights:

• Firmographic fit (tight band): 35%
• Individual behavioral intent (recent, high-signal): 30%
• Tech stack match: 15%
• Trigger events: 10%
• Account engagement: 10%

Notice how the weights invert on account engagement and individual intent. That's not preference — that's how the two segments actually buy.

Step 4: Set Segment-Specific SLAs and Sequences

Mid-market Tier A should get contacted within 15 minutes. That window matters because mid-market buyers are often in an active evaluation with a short list. Enterprise Tier A can absorb a 24-hour SLA because the buying cycle is measured in months.

Similarly, sequences should differ. Enterprise cadences need multi-threading logic — sequences that pull in second and third contacts from the account. Mid-market cadences should compress touches and lean harder on the individual signal that triggered the MQL. If your team runs both motions from the same 12-touch sequence, one is being served badly. This is the kind of thing Outbound System Engineering addresses head-on — different motions require different sequence architectures, not just different copy.

What Breaks If You Skip This

Teams that try to shortcut segmented scoring usually run into one of these failure modes within a quarter or two:

• Marketing hits MQL targets, sales rejects the leads. The model is producing volume against a broken definition of quality.
• Attribution models look great, revenue doesn't follow. Because the scored MQLs aren't actually the deals closing. Revenue Intelligence work often surfaces this — the sourced-pipeline number and the closed-won reality don't reconcile.
• Reps stop trusting the queue. Once trust breaks, reps go rogue. Some do well, most don't. Forecasting becomes noise.
• CAC creeps up. You're spending marketing dollars generating leads that either don't convert or convert at the wrong deal size.

The Quarterly Review Discipline

A signed ICP and segmented scoring model aren't set-and-forget. The market moves. Your product moves. Your ideal customer moves.

Every quarter, run this review:

1. Closed-won ICP alignment: What percentage of closed-won deals matched the ICP for their segment? Target 80%+.
2. Tier conversion by segment: Is Tier A still converting at 2x baseline in each segment?
3. Escape analysis: What deals closed that weren't scored as high-priority? Are there patterns?
4. Negative ICP review: Are you still routing accounts to sales that should be filtered out?
5. Weight recalibration: Do the weights still match closed-won reality?

Most teams don't have the internal capacity to run this rigorously every quarter. It's the kind of work that fits into a GTM Operations Retainer — someone owning the calibration cadence so it actually happens.

The Uncomfortable Truth About Scoring Models

Scoring models are treated as marketing artifacts, but they're really the operational expression of what your entire go-to-market team believes about who buys. When enterprise and mid-market are collapsed into one model, you're implicitly saying they buy the same way. They don't.

The teams that get this right treat their scoring model like a product — versioned, tested, calibrated against reality, and owned by someone whose job depends on it working. The teams that get it wrong treat it like a report — configured once, referenced occasionally, defended when questioned.

If you're running two motions and one model, you're leaving pipeline on the table in whichever segment the model wasn't built for. Usually it's mid-market, because most B2B scoring frameworks were designed with enterprise complexity in mind. But it can go either way.

---

If your scoring model is producing the right volume but the wrong outcomes — or your sales team has quietly stopped trusting the queue — it's worth a diagnostic conversation. Book a strategy call with Revstek and we'll walk through your current scoring architecture, segment behavior, and closed-won data to identify where the model is misfiring and what a segmented rebuild would look like.