At a glance. Prioritise with evidence from your customers. Use scores to guide the next action and keep a person accountable for commercial qualification.
Keep fit and engagement separate
Fit describes how closely an account matches your ideal customer: sector, geography, company size, relevant role and business need. Engagement describes what the person has done: replied, requested information, attended a call or taken another meaningful step.
A strong-fit account with no response may deserve research and a relevant introduction. A highly active contact outside the target does not become a good opportunity just because the score is high. Show the two dimensions separately before combining them.
Start with customers you have actually won
- Review closed business from the last 12 months, including losses and opportunities that stalled.
- Identify the characteristics and actions that consistently mattered.
- Choose a small set of signals with reliable data.
- Assign provisional weights and define what each threshold triggers.
- Review false positives and missed opportunities with the sales team.
An illustrative scoring model
| Signal | Dimension | Illustrative points |
|---|---|---|
| Company matches the agreed size range | Fit | 25 |
| Contact has a relevant decision-making or influencing role | Fit | 20 |
| Company operates in a target sector | Fit | 10 |
| Contact replies to an introduction | Engagement | 20 |
| Contact visits a relevant service or pricing page | Engagement | 15 |
| Contact attends a discovery call or demo | Engagement | 10 |
These points are a worked example, not recommended universal weights. Website activity depends on your measurement setup and should not outweigh a substantive conversation. Choose MQL and SQL thresholds based on what your sales team can use, then test them against outcomes.
Manual tiers, CRM rules or an enrichment layer
| Approach | Best suited to | Watch for |
|---|---|---|
| Manual A/B/C tiers | Manageable volumes and an experienced reviewer. | Consistency between reviewers and clear reasons for a tier. |
| CRM scoring rules | Several sources and a repeatable qualification process. | Missing data, outdated rules and scores no one sees. |
| Enrichment and AI support | Research-heavy lists or large datasets needing classification. | Unverified facts, opaque decisions and ongoing data quality. |
There is no fixed monthly lead count at which automation becomes necessary. Measure how much time sorting takes and whether important contacts are missed. A simple A/B/C view can be more useful than a complicated score with no trusted owner.
Make the score lead to an action
Define an owner, response target and next step for each tier. Keep the score in the CRM view where the team works. If a high score does not change what happens next, it is only another field.
Review the model during the first 60–90 days and whenever the offer or target changes. Look at which leads became real opportunities, which were rejected and why. Keep a record of rule changes so comparisons remain meaningful.
Cost, data and governance
Check whether your existing CRM plan already includes the required features before buying another tool. Compare configuration, data enrichment, monitoring and team time alongside licence costs. Provider plans change; use the current official quote for purchasing decisions.
Document which data the score uses, how it is obtained and how people can correct it. Do not assume that data already held can automatically be used for every profiling purpose. Check the intended use and safeguards with the person responsible for data protection before launch.
Frequently asked questions
Is scoring the same as qualification?
No. Scoring helps prioritise review. Qualification confirms the commercial context through research and conversation. A high score cannot answer every question about budget, need, decision-makers or timing.
How often should the model change?
Review it frequently at the start, then when evidence or the target changes. Avoid changing weights after every isolated result; use a documented review cycle.
Do we need AI?
Only if it solves a real research or classification problem. Reliable CRM rules or manual tiers may be sufficient.