Building your own scoring model
Prospecting 11 min
0/28
Business problem
Generic scoring ('my eyes rate 1-100') doesn't work because every company looks similar. The model has to reflect your reality: what was common in your won deals (industry? size? signals?) and what eliminated them (misfits). Without that, scoring is decoration.
Proven method
Building a scoring model in 5 steps:
- Collect points from won deals (what was common) and lost ones (what eliminated them),
- Define criteria with weights: +20 ICP match, +15 target industry, +15 size in range, +10 identified decision maker, +15 buying signal, +10 growth (hiring/expansion), +10 technology in the stack, -20 wrong industry, -15 too small scale, -10 no decision maker,
- Set thresholds: HOT 80-100 (contact today), WARM 60-79 (sequence), COLD 40-59 (nurture), BAD <40 (never),
- Implement in a spreadsheet/CRM (columns + formula) or in AI prompts,
- Test on 20 historical leads and calibrate the weights. AI doesn't invent the model — AI APPLIES your model to leads and justifies the result.
Market example
- BAD
- Unstructured prospecting without ICP filters and generic outreach to random accounts.
- GOOD
- Your model: ICP match 20, industry 15, size 15, decision maker 10, signal 15, growth 10, technology 10, misfits -45. Lead: performance agency (industry +15), 18 people (size +15), ICP match (a CRM for agencies +20), Head of Sales found (+10), hiring 3 salespeople (signal +15, growth +10), uses HubSpot Starter (technology +10) = 95 → HOT. AI justifies: 'the lead meets 6 of 7 criteria, no misfits'.
Ready-to-use prompt
You are a lead scoring architect. Context: I sell [PRODUCT] for [SEGMENT]. My won deals shared: [WHAT WAS COMMON]. Lost deals shared: [WHAT ELIMINATED THEM]. Task: Design my scoring model: 1. Positive criteria (5-8) with weights (sum max ~100) 2. Negative criteria (3-5) with negative points 3. Thresholds: HOT/WARM/COLD/BAD 4. Example: score the lead [LEAD DESCRIPTION] by the model + justification 5. 3 calibration questions I must answer to fine-tune the weights Format: sections, ready to implement in a spreadsheet.
Model rekomendowany: Claude 3.5 Sonnet / GPT-4o Copy and paste into ChatGPT, Claude or Perplexity
ALTERNATYWNE SCENARIUSZE
Variations & alternative scenarios
Spreadsheet version: AI generates formulas (IF/SUM) ready to copy.
Team version: AI proposes who handles HOT/WARM/COLD.
Practical exercise
Write your scoring model: 5-8 criteria with weights + 3 negative ones. Score 5 leads from your list. Check whether the categories (HOT/WARM/COLD/BAD) match your intuition — if not, fix the weights.
Zapisano w przeglądarce