One sales signal is usually ambiguous.
A funding round does not prove a project exists. A relevant comment does not prove a budget exists. A pricing-page visit does not tell you who visited or why.
Signal stacking in sales improves the decision by combining several pieces of evidence. But a stack is only useful when the signals add independent information. Three alerts derived from the same press release do not create three times the confidence.
What is signal stacking?#
Signal stacking is the process of combining multiple recent indicators about a person or account to improve prioritisation and determine an appropriate next action.
A strong stack often combines different dimensions:
- Fit: the company and person match your market
- Change: something has altered priorities, capacity or timing
- Interest: behaviour suggests a relevant topic or category matters
- Relationship: there is a credible route or existing connection
- Readiness: first-party or commercial activity suggests evaluation
The stack is an evidence model, not a guarantee of intent.
Add information, not alert count#
Suppose a company announces a new VP Sales. You receive the event from a news feed, LinkedIn and a data provider. That is one underlying change observed three times.
Now suppose the same company:
- Hires a new VP Sales.
- Opens five SDR roles.
- Has two employees researching sales workflow automation.
- Comments on a detailed post about prospect-data quality.
Those events add different information: leadership change, team growth, topic interest and person-level context. Together they support a more specific hypothesis.
A practical signal taxonomy#
Give each event a family so the workflow can reason about independence.
| Family | Examples |
|---|---|
| Firmographic fit | Industry, size, region, business model |
| Person fit | Function, seniority, responsibility |
| Organisational change | New leader, restructure, expansion |
| Hiring | Open roles, hiring velocity, new function |
| First-party behaviour | Site activity, trial usage, content engagement |
| Third-party research | Topic surge, review activity, event participation |
| Social activity | Post, substantive comment, relevant engagement |
| Technology | Adoption, removal, integration or migration |
| Relationship | Previous customer, mutual contact, existing owner |
| Negative evidence | Opt-out, recent renewal, project paused, layoffs |
Signals in different families are more likely to be complementary, though they can still share one source event.
A transparent scoring framework#
Start with a small model that reviewers can understand.
Base fit#
- Company fit: 0–3
- Person fit: 0–3
If either is zero, the record normally should not reach outbound regardless of activity.
Event evidence#
- Direct problem or evaluation behaviour: +3
- Strong change tied to your category: +2
- Relevant person-level social evidence: +2
- Broad company change: +1
- Weak engagement: +0.5
Independence bonus#
- Two qualifying families: +1
- Three qualifying families: +2
- Four or more: +3, capped
Negative evidence#
- Active opportunity with another owner: route, do not prospect
- Opt-out: suppress
- Recent irrelevant or rejected contact: -2 or suppress
- Event outside its useful window: remove from active score
- Contradictory business event: -1 to -3
The numbers are deliberately coarse. False precision makes a new model harder to question without making it more accurate.
Use time windows by signal type#
Signals decay at different speeds.
| Signal | Illustrative active window |
|---|---|
| Direct recommendation request | Days |
| Pricing or trial activity | Days to weeks |
| Substantive social comment | One to several weeks |
| New executive | One to several months |
| Hiring plan | Several months while active |
| Funding or expansion | Several months, weak without detail |
These are starting assumptions, not universal rules. Measure your sales cycle and how quickly each event loses predictive value.
Store event time separately from capture time. An article published today about a change that happened six months ago is not a fresh event.
Detect correlated signals#
Use an underlying event key where possible. It might combine:
- Company identifier
- Event family
- Normalised event subject
- Effective date
- Source entity
Then link observations from several providers to the same event rather than scoring them separately.
Also watch for causal chains. A funding round may produce a company announcement, several news articles and a hiring burst. The hiring is genuinely new evidence, but its relationship to the funding should remain visible.
Person-level and account-level stacks#
Do not silently mix entities.
An account may show research activity while a particular executive shows no observable behaviour. That can justify researching the buying group; it does not prove the executive performed the research.
Maintain:
- Account score
- Person score
- Relationship between the person and account
- Confidence in identity resolution
When routing, explain which evidence applies to which entity.
Example stacks#
Weak stack: duplicated attention#
- Target account matches size filter
- One employee likes a relevant post
- The same like arrives from two collection systems
Action: deduplicate and monitor. There is little evidence for direct contact.
Useful stack: change plus topic evidence#
- Clear company fit
- New RevOps leader started 30 days ago
- Company is hiring SDR operations roles
- Leader comments on a discussion about enrichment quality
Action: human review with the original comment, job descriptions and role-change evidence.
High-readiness stack: first-party evaluation#
- Existing ICP account
- Known stakeholder returns to pricing and security pages
- Second stakeholder starts a trial
- Open opportunity has been dormant for 60 days
Action: notify the account owner with the first-party evidence and ownership history.
Negative stack: apparent interest, wrong timing#
- Strong topic research
- Relevant executive in place
- Company announced a broad hiring freeze
- Prospect requested no contact last quarter
Action: suppress. Positive activity must not override explicit negative rules.
How to route stacks#
Translate score bands into different actions rather than treating one threshold as “send”.
| Evidence band | Action |
|---|---|
| Fit only | Monitor or nurture |
| One relevant event | Research |
| Two independent events | Human review |
| Strong person-level plus account evidence | Prioritised review or owner alert |
| Direct first-party evaluation | Immediate account-owner action |
| Any hard suppression | No outreach |
Keep a reviewer in the loop for ambiguous social and third-party evidence.
Learn weights from outcomes carefully#
Track:
- Acceptance rate by signal family
- Positive replies and meetings by combination
- Time from first event to outcome
- False positives and rejection reasons
- Duplicate-event rate
- Percentage of high scores created by one provider
- Negative-signal saves: actions correctly suppressed
You need sufficient samples before changing weights. One large deal influenced by a rare signal does not prove that signal deserves maximum priority everywhere.
Common signal-stacking mistakes#
Rewarding volume from one source#
Ten page views or ten likes may show intensity, but they should not automatically equal ten independent signals.
Forgetting fit gates#
Strong activity from an account you cannot serve is not a high-priority lead.
Hiding the components#
A total score without its events cannot be reviewed or debugged.
Never expiring evidence#
Historical events are useful context. They should not accumulate forever in the active score.
Treating negative evidence as a small deduction#
Some conditions are hard stops. An opt-out is not “minus two points”.
A stack should tell a coherent story#
The purpose of signal stacking is not to manufacture certainty. It is to combine independent evidence into a better, explainable reason for attention.
The final record should let a reviewer say:
This company fits, this relevant change happened, this person showed related interest, and the evidence is current enough to justify this next step.
If the system cannot tell that story without hiding behind a score, the stack needs more work.
TWL Signals
Turn market attention into a qualified prospect feed.
Monitor the people and companies your buyers already follow, keep the original context, enrich the profiles and route only the strongest matches into outreach.
Start building your signal workflow