A one-page canvas to design a scoring model with marketing and sales in the room. Allocate 90 minutes the first time you run it. Print the canvas — do not run it as a screenshare. People agree more readily when they are looking at a shared physical artifact, not staring at a Zoom window.
Copy this file, rename it scoring-model-<YYYY-Q>.md, and fill in each section together.
Pick one — and only one — primary job:
- Routing: tell sales when to engage in-ICP people at the right moment.
- Prioritization: rank an existing sales queue.
- Suppression: tell marketing when not to send another email.
If the team wants the model to do all three, agree which is primary. The model will be calibrated against the primary job. The other jobs are secondary uses of the same score.
Primary job chosen:
Why this and not the others:
The model only scores people who pass the ICP gates. List the gates:
- Industry gate: ___ (in-ICP industries)
- Company size gate: ___ (in-ICP employee or revenue band)
- Geography gate: ___ (in-territory)
- Role/function gate (if applicable): ___
A person who fails any gate is excluded from the model entirely — they do not get a fit score, and behavior signals do not accumulate against an MQL alert.
Agree the gates with sales:
- Marketing leader sign-off: ____________________
- Sales leader sign-off: ____________________
We will maintain two separate scores:
-
BehaviorScore(dynamic, what they do) -
FitScore(static, who they are)
The routing rule combines them at the gate, not at the score. Sales sees both numbers on the lead record.
If anyone in the room argues for a combined single score, refer them to 01-what-scoring-is-actually-for.md, section "Two scores, not one." Resolve the disagreement before moving on.
| Signal | Type (primary/secondary/weak) | Weight | Decay half-life |
|---|---|---|---|
| Demo request | Primary (gate) | 100 | Never |
| Pricing page visit (2+ in 7 days) | |||
| Contact-sales form | |||
| Webinar attended | |||
| Multi-session site visit | |||
| High-value content download | |||
| General content download | |||
| Email click | |||
| Email open | (drop) | 0 | n/a |
| ___ | |||
| ___ |
Anti-signals (reduce score or exclude):
| Signal | Effect |
|---|---|
| Unsubscribed | Exclude from model |
| Hard bounce / complaint | Exclude from model |
| Marked competitor | Exclude from model |
| Out-of-ICP role change (where detected) | ___ |
Gates are already in Section 2. Within-ICP, fit weights:
| Dimension | Weight | Banded scoring (if applicable) |
|---|---|---|
| Role function (matches buyer persona) | ||
| Seniority (Director+ / Manager / IC) | ||
| Technographic (in-stack tech) | ||
| Buying-group size (multi-person at account) | ||
| ___ |
Maximum FitScore for a perfect-fit person: ___
The MQL alert fires when:
BehaviorScore >= ___ AND FitScore >= ___
Expected daily alert volume at these thresholds: ___ alerts/day
Sales capacity to act on alerts within 24 hours: ___ alerts/day
If the expected volume exceeds capacity, the thresholds need tightening. Do this in the room, with sales.
The model does not fire MQL alerts when:
- Person is attached to an open opportunity
- Person was MQL'd and rejected by sales within the last ___ days
- Person is a current customer (route through expansion model instead)
- Person has been unengaged from email for 90+ days (re-MQL only on new direct action)
- ___ (other team-specific rule)
We will review the model:
- Quarterly (recommended)
- Twice per year
- Other: ___
Review participants:
- Marketing operations lead
- Sales operations lead
- VP Marketing (or designee)
- VP Sales (or designee)
- RevOps lead (if separate)
| Change type | Sign-off required |
|---|---|
| Add a new signal | Marketing ops + RevOps |
| Change a weight | Marketing ops + RevOps |
| Change the MQL threshold | Marketing + Sales VPs |
| Change a fit gate (ICP change) | Marketing + Sales VPs |
| Retire a signal | Marketing ops + RevOps |
| Suppress an audience from the model | Marketing + Sales |
Sign-off log: maintain in the wiki, dated, with rationale.
This canvas becomes the source of truth for the scoring model. The platform configuration is the implementation of the canvas, not its definition. If they diverge, the canvas is authoritative and the platform gets reconciled to it.
Canvas owner: ____________________
Canvas storage location: ____________________
Date of this version: ____________________
Next scheduled review: ____________________
This canvas is the agreed scoring model for the period stated above.
- VP Marketing: ______________________________ Date: ___________
- VP Sales: ______________________________ Date: ___________
- Marketing Ops Lead: ______________________________ Date: ___________
- Sales Ops / RevOps Lead: ______________________________ Date: ___________