Treat scoring as a tested decision aid, not an automatic send decision. Use known positive and negative examples, change one rule at a time, and confirm that processing has finished before evaluating the result.
Before you start
- Collect several known-fit and known-not-fit companies and people.
- Write the offer, customer type, region, and business context each company segment represents.
- List the responsibilities and influence that matter for each persona, not only exact titles.
- Separate hard exclusions from preferences.
- Choose an owner for changes, tests, and controlled rescoring.
Understand the scoring layers
| Layer | Purpose | Common mistake |
|---|
| Target Company Segments | Describe the types of companies that fit the offer and use case. | Combining unrelated business models or regions into one vague segment. |
| Persona Segments | Describe the roles, responsibilities, and influence relevant within a target company. | Matching only exact job titles and excluding valid title variations. |
| DQ criteria and exclusions | Remove companies or people that are unsuitable for every applicable target segment. | Using a hard exclusion for a preference that has legitimate exceptions. |
| Lead result | Combines available company and persona evidence into a prioritised review input. | Treating the result as automatic approval without checking the underlying data. |
Write useful segments
Describe each company segment in concise, specific language. Include the offer, customer type, and context needed to distinguish it from another segment. If the current editor recommends a length, follow that guidance; the source brief suggests roughly 30–100 words as a practical target.
Persona segments should describe responsibilities and influence. Add representative titles as examples, but do not make one exact title the only evidence when equivalent titles are common.
- Keep segments distinct enough that a reviewer can explain the difference.
- Use observable company characteristics where possible.
- Include positive context and explicit non-fit patterns.
- Avoid overlapping segments that produce contradictory assignments.
- Review partner or optional segments; do not leave an enabled segment empty.
Steps
- **Open Lead Scoring. **Start with Target Company Segments and review the existing logic before adding new rules.
- **Define company segments. **Describe the offer, customer type, region, scale, and business context that make a company suitable.
- **Define persona segments. **Add a small number of roles based on responsibility, buying influence, and relevant title variations.
- **Add hard DQ criteria carefully. **Use headquarters, location, employee count, industry, or research-based exclusions only when they apply across the relevant target segments.
- **Use Save and Test. **Compare known positive and negative companies and people. Verify the source data as well as the resulting segment and priority.
- **Change one rule at a time. **Record what changed, why it changed, and which test cases should move or remain stable.
- **Confirm processing is complete. **Do not rely on a Done label alone if individual leads still lack scoring results.
- **Rescore in a controlled way. **Use the supported rescore flow, monitor the affected sample, and avoid broad changes during active outreach.
- **Document the outcome. **Record the approved rule, test examples, owner, and review date in the team change log.
Review and safety checks
- Known positive examples are not removed by hard exclusions.
- Known negative examples do not qualify through an overly broad segment.
- Persona logic covers responsibilities and realistic title variations.
- The underlying company and role data is current enough for the decision.
- Scoring has finished for the sample before results are compared.
- No active campaign will be changed unexpectedly by a rescore.
- A human reviewer still checks context, exclusions, and ownership before outreach.
Troubleshooting
| Problem | Solution |
|---|
| A known good lead is disqualified | Check hard company exclusions first, then the company segment, persona responsibility, and title logic. Test the lead again after one controlled change. |
| Too many leads qualify | Make company fit and persona responsibility more specific. Use DQ only for exclusions that are truly universal. |
| Changes cannot be saved | Check for enabled but empty segments or incomplete required fields, then retry. If the problem persists, preserve the draft and contact support. |
| A list looks complete before scoring results are ready | Open representative leads and confirm that company, persona, and priority results are populated before relying on the list status. |
| The reasoning does not match a manual change | Record the manual decision and verify the current displayed reasoning. Escalate inconsistent explanations rather than treating them as reliable. |
Frequently asked questions
What is the difference between a company segment and a persona segment?
A company segment describes the type of organisation that fits. A persona segment describes the relevant role or responsibility within that organisation.
When should I use DQ criteria?
Use DQ for hard exclusions that make a record unsuitable across the applicable target segments. Keep preferences in segment logic instead.
How many test records should I use?
Use enough representative positive, negative, and edge cases to expose the rule’s behaviour. Start with at least several known examples from each important segment.
Why was a good lead disqualified?
Check hard company exclusions first, then company-segment wording, persona responsibility, title variations, and the underlying data.
Can I rescore every list after a change?
Use the supported rescore flow carefully. Test first, document the effect, and avoid unexpected changes to active outreach.
Does a high score mean the lead can be contacted automatically?
No. The score is a review input. Confirm signal context, ownership, previous outreach, opt-outs, blacklists, sender readiness, and message quality.
Keep reading
Set up copywriting, prompts, and templates
Use Copywriting settings for shared style and delivery rules, and use Templates for reusable prompts, replies, and Deep Research instructions. Each campaign still needs the correct list, signal, persona, language, and sender context.
Review and approve the Knowledge DB
Review every Knowledge DB section before approving it for research or copywriting. Approval means the content is trusted enough to influence customer facing output, so persuasive wording is not a substitute for accurate evidence.
Write effective Deep Research prompts
Write Deep Research prompts around a decision: qualify an account, verify a signal, identify the relevant role, or find a factual conversation angle. A focused prompt produces more useful results than an open ended request to research everything.
