EVIDENCE NOTE · CLAIM RIGHTS
What buying signals let you say in outbound
A hiring announcement can support a sentence about hiring. It cannot establish that a company is replacing its software. Check each claim against the source before using it in outreach.
THE SHORT VERSION
Key takeaways
- A signal decides what to investigate; evidence decides what the message may claim.
- Check each factual claim against the source passage that supports it.
- Unsupported claims should be blocked, qualified as inference, or rewritten at the problem level.
A company hiring a VP Sales and seven SDRs may be expanding its sales team. A leader’s public complaint about stale prospecting data tells you something more specific. Each source supports different language in the first message.
Attaching a source URL to a generated message does not establish that the message is accurate. Check whether the source supports each claim before approving the draft.
The hidden jump inside bad personalization
Consider this draft:
I saw you’re replacing your sales stack and thought this might be relevant.
I saw your comment about stale prospecting data and thought this might be relevant.
EVIDENCE-BACKED ✓The first sentence sounds plausible because the surrounding signals point in that direction. But plausibility is not evidence. The source supports a concern about data freshness; it does not support a vendor replacement, an active evaluation, a budget, or a buying decision.
The rewrite refers to the concern the buyer expressed and leaves their purchasing plans open.
Four evidence states
They said it.
A specific person made a relevant public statement and the source can be preserved.
Specific attribution may be allowed.It happened.
A company event is confirmed by a reliable source, such as a careers page or company announcement.
Describe the event, not its imagined motive.We inferred it.
Several facts make an interpretation reasonable, but no source states the interpretation directly.
Use for prioritization or qualify the language.Do not pretend.
The source is missing, stale, contradictory, or attached to the wrong entity.
Block the claim or stay at the problem level.Check each claim against its source
A source can be relevant to an account and still fail to support a sentence. Verification therefore needs to happen at claim level.
| Evidence available | Allowed | Not supported |
|---|---|---|
| Company careers page lists seven SDR openings. | “I noticed your team is hiring seven SDRs.” | “You’re struggling to scale outbound.” |
| VP Sales wrote about stale prospecting data. | “I saw your comment about stale prospecting data.” | “Your current vendor is failing.” |
| Company launched an enterprise tier. | “I saw the enterprise launch.” | “You’re preparing for SOC 2.” |
| Strong ICP fit, no fresh event. | “Teams in your category often face…” | “I noticed you are currently dealing with…” |
20 observed, inferred, and blocked claim examples
Use this library as a calibration set. The middle column can guide prioritization, but it must not be promoted into a fact. The final two columns show the narrowest useful claim and the language that should be blocked.
| Observed | Reasonable inference | Allowed in outreach | Blocked as fact |
|---|---|---|---|
| A new VP Sales joined. | Priorities may be reassessed. | “Congratulations on the new role.” | “You are replacing your sales stack.” |
| Seven SDR roles are open. | The SDR team is expanding. | “I noticed seven open SDR roles.” | “Your outbound cannot scale.” |
| A leader wrote that prospecting data becomes stale. | Freshness matters to that leader. | Quote or paraphrase the stated concern. | “Your current vendor is failing.” |
| The company raised a Series B. | Growth investment may follow. | Reference the announced round. | “You now have budget for us.” |
| An enterprise tier launched. | Enterprise operations may become more important. | Reference the launch. | “You need SOC 2 immediately.” |
| SOC 2 work was mentioned publicly. | Trust work is active. | Reference the exact mention. | “You failed an audit.” |
| The pricing page was visited repeatedly. | The category may be under consideration. | Use it to prioritize research. | “You are evaluating vendors.” |
| A competitor comparison page was viewed. | Alternatives may be relevant. | Discuss the problem without citing surveillance. | “You are switching from Competitor X.” |
| The company added a RevOps role. | Revenue process ownership is formalizing. | Mention the published role. | “Your CRM data is a mess.” |
| A regional expansion was announced. | New go-to-market coordination may be needed. | Reference the announced market. | “Your current process cannot support Europe.” |
| A new integration was released. | That ecosystem matters to the roadmap. | Reference the release. | “Customers demanded this integration.” |
| A product was discontinued. | Resources or strategy changed. | Reference the public notice. | “The product failed.” |
| Support response-time complaints are public. | Service operations may be under pressure. | Cite the exact public statement when appropriate. | “Your support team is understaffed.” |
| The website now targets a new segment. | Positioning or acquisition is changing. | Reference the visible positioning change. | “Your old ICP stopped converting.” |
| A leadership post asks for tool recommendations. | The category is being explored. | Respond to the stated request. | “You selected a budget and timeline.” |
| A webinar covered the exact problem. | The topic matters to the brand or audience. | Reference the webinar topic. | “This is your top internal priority.” |
| A technology was removed from the public site. | The stack may have changed. | Use as a research prompt and verify elsewhere. | “You churned from Vendor X.” |
| Headcount fell across two quarters. | Efficiency may matter more. | Avoid sensitive personalization; use broad efficiency language. | “You are cutting costs because revenue is down.” |
| A customer case study names a workflow. | That workflow exists and matters externally. | Reference the published workflow. | “The workflow is broken.” |
| Strong ICP fit; no current event. | The role may recognize a common problem. | “Teams like yours often…” | “I noticed you are dealing with…” |
A verification pipeline needs six checks
- 01
Source
Can the underlying material be opened, preserved, and attributed? A summary without a source is a research lead, not evidence.
- 02
Entity
Does the source refer to the correct company and person? Similar company names, reposts, and stale profiles create confident-looking errors.
- 03
Claim
Does the relevant passage directly support the sentence being considered?
- 04
Recency
Was the event recent enough to justify a “why now” message?
- 05
Contradiction
Is there stronger or newer evidence that changes the interpretation?
- 06
Language
Can the claim be written as fact, must it be qualified as inference, or should it be removed?
What a reviewer needs to see
A score compresses uncertainty. That can be useful for ranking, but it is a weak explanation for a person reviewing a message. “Confidence: 87” does not tell the reviewer which part is observed, which part is inferred, or why a particular sentence is allowed.
A useful review object exposes the chain:
- Source
- Original URL and preserved excerpt
- Entity
- Company and person the source refers to
- Observed fact
- What happened or what was stated
- Evidence class
- Behavioral, attributive, inference, or none
- Freshness
- When the observation stops supporting “why now”
- Language right
- The specific sentence pattern the source permits
Save rejected claims and the reasons for rejection so future drafts can follow the same standard. The full Nolina verification workflow keeps that decision attached to the source and outbound message.
When there is no event, use the audience lane
Evidence-backed outbound does not require a fresh signal for every prospect. It requires honesty about whether a signal exists.
A well-matched account can receive problem-level outreach based on its role, category, or operating model. The message simply cannot imply that a specific event occurred. We call this the audience lane.
Keep the source beside the draft so a reviewer can check the claim.
Automation should inherit evidence rights
Human review is valuable during early learning, but “a person clicked approve” is not a complete control system. Write review criteria that can be applied consistently:
- Specific personal attribution requires a preserved first-party source.
- Observed company events require a reliable source and date.
- Inference must remain labeled or be translated into problem-level language.
- Missing or contradictory evidence blocks the claim.
- Channel execution must respect the policies of the platform being used.
For example, LinkedIn’s published policies prohibit unauthorized tools that scrape its services or automate activity on the website. Research, contact resolution, drafting, and supported workflows should not be presented as permission for unrestricted bot sending.
This is also the architectural difference highlighted in the Nolina vs GoJiberry comparison: multichannel execution and claim-level evidence governance solve related but different jobs.
Review the reason for writing
Good evidence does not make every email persuasive. It makes the reason for writing inspectable. The offer still has to matter, the buyer still has to own the problem, and the message still has to earn a response.
Use the same source checks when reviewing later drafts. Repeated factual corrections are a reason to pause automation and revise the message rules.
COMMON QUESTIONS
Frequently asked questions
What is evidence-backed outbound?
Evidence-backed outbound is a prospecting approach in which specific message claims are constrained by preserved sources. Facts, observed events, and inferences remain separate so a seller can see why a statement is allowed before it reaches a buyer.
What is the difference between a signal and evidence?
A signal determines what may be worth investigating. Evidence is the source material used to check a specific statement. One signal can produce several possible interpretations, but only some may be safe to claim as fact.
Can inference be used in outbound messaging?
Yes, but it should not be disguised as an observed fact. Inference can guide prioritization, be qualified in the message, or be rewritten as a broader problem-level statement when direct support is missing.
What should happen to an unsupported outbound claim?
The system should block it, qualify it as inference, or rewrite it using language the evidence supports. A confidence score alone is not sufficient because it does not explain which part of the sentence failed.
SOURCE NOTES
Sources and further reading
- Intent Signals Are Not a Strategy. They’re a Multiplier.Common Room ↗
- How to Identify and Act on Buying SignalsClay ↗
- Prohibited Software and ExtensionsLinkedIn Help ↗
- LinkedIn User AgreementLinkedIn ↗
Vendor sources describe definitions, workflows, and platform policies. They do not independently establish product performance.
TRY A PILOT